# Machine Intelligence

Machine intelligence is a concept concerning engineered computational systems that perform cognitive capabilities such as learning, reasoning, perception, and planning.

This is one page of public article previews, not the complete archive. Follow Next page to continue. Summaries are not the original full articles.

## AI GameStore: Scalable, Open-Ended Evaluation of Machine General Intelligence with Human Games

DevFeed: [AI GameStore: Scalable, Open-Ended Evaluation of Machine General Intelligence with Human Games](<https://devfeed.tech/articles/ai-gamestore-scalable-open-ended-evaluation-of-machine-general-intelligence-with-human-games-145317.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2602.17594>)

Author: Lance Ying, Ryan Truong, Prafull Sharma, Kaiya Ivy Zhao, Nathan Cloos, Kelsey R. Allen, Thomas L. Griffiths, Katherine M. Collins, Jos\\'e Hern\\'andez-Orallo, Phillip Isola, Samuel J. Gershman, Joshua

Published: 2026-11-20T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Artificial Intelligence](<https://devfeed.tech/sources/arxiv-artificial-intelligence.md>)

Topics: [Machine Intelligence](<https://devfeed.tech/topics/machine-intelligence.md>), [Artificial Intelligence](<https://devfeed.tech/topics/ai.md>), [Frontier AI](<https://devfeed.tech/topics/frontier-ai.md>), [Ground truth / benchmark quality](<https://devfeed.tech/topics/ground-truth-benchmark-quality.md>), [AI Development](<https://devfeed.tech/topics/ai-development.md>)

Tags: [ai](<https://devfeed.tech/tags/ai.md>), [ai-benchmarks](<https://devfeed.tech/tags/ai-benchmarks.md>), [ai-systems](<https://devfeed.tech/tags/ai-systems.md>), [all](<https://devfeed.tech/tags/all.md>), [app](<https://devfeed.tech/tags/app.md>), [app-store](<https://devfeed.tech/tags/app-store.md>), [apple](<https://devfeed.tech/tags/apple.md>), [apple-app-store](<https://devfeed.tech/tags/apple-app-store.md>), [arxiv](<https://devfeed.tech/tags/arxiv.md>), [automatically](<https://devfeed.tech/tags/automatically.md>), [based](<https://devfeed.tech/tags/based.md>), [benchmarks](<https://devfeed.tech/tags/benchmarks.md>), [best](<https://devfeed.tech/tags/best.md>), [capabilities](<https://devfeed.tech/tags/capabilities.md>), [cs-ai](<https://devfeed.tech/tags/cs-ai.md>), [evaluation](<https://devfeed.tech/tags/evaluation.md>), [games](<https://devfeed.tech/tags/games.md>), [general](<https://devfeed.tech/tags/general.md>), [intelligence](<https://devfeed.tech/tags/intelligence.md>), [machine](<https://devfeed.tech/tags/machine.md>), [open](<https://devfeed.tech/tags/open.md>)

### AI overview

AI GameStore proposes evaluating machine general intelligence through human-designed games, comparing AI and people with similar experience, time, or other resources. Its platform uses LLMs and human input to create games by sourcing and adapting standardized, containerized environments from popular digital gaming platforms. In an initial study of 100 games drawn from Apple App Store and Steam charts, seven frontier vision-language models played short episodes; even the best scored below 10% of the human average on most games, with particular difficulty in world-model learning, memory, and planning. The authors present the platform as an early step toward scalable, open-ended evaluation and outline further development as future work.

### Source excerpt

arXiv:2602.17594v2 Announce Type: replace Abstract: Rigorously evaluating machine intelligence against the broad spectrum of human general intelligence has become increasingly important and challenging in this era of rapid technological advance. Conventional AI benchmarks typically assess only narrow capabilities in a limited range of human activity. Most are also static, quickly saturating as model capabilities improve. We propose that a more promising way to evaluate human-like general intelligence in AI systems is through a particularly strong form of general game playing: studying how and how well they play and learn to play all conceivable human games, in comparison to human players with the same level of experience, time, or other resources. We define a "human game" to be a game designed by humans for humans, and argue for the evaluative suitability of this space of all such games people can imagine and enjoy -- the "Multiverse of Human Games". Taking a first step towards this vision, we introduce the AI GameStore, a scalable and open-ended platform that uses LLMs with humans-in-the-loop to synthesize new representative human games, by automatically sourcing and adapting standardized and containerized variants of game environments from popular human digital gaming platforms. We generated 100 such games based on the top charts of Apple App Store and Steam, and evaluated seven frontier vision-language models (VLMs) on short episodes of play. The best models achieved less than 10% of the human average score on the majority of the games, and especially struggled with games that challenge world-model learning, memory and planning. We conclude with a set of next steps for building out the AI GameStore as a practical way to measure and drive progress toward human-like general intelligence in machines.

## GuardAgent: A Memory-Augmented Agentic Framework for Adaptive Detection of Adversarial Promotions in LLM-Based Recommendation

DevFeed: [GuardAgent: A Memory-Augmented Agentic Framework for Adaptive Detection of Adversarial Promotions in LLM-Based Recommendation](<https://devfeed.tech/articles/guardagent-a-memory-augmented-agentic-framework-for-adaptive-detection-of-adversarial-promotions-in-llm-based-recommendation-141482.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.22274>)

Author: Shiyi Yang, Xinshu Li, Chen Wang, Xiwei Xu, Liming Zhu, Lina Yao

Published: 2026-11-18T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Cryptography and Security](<https://devfeed.tech/sources/arxiv-cryptography-and-security.md>)

Topics: [threat detection](<https://devfeed.tech/topics/threat-detection.md>), [Large Language Model](<https://devfeed.tech/topics/llm.md>), [Adversarial attacks](<https://devfeed.tech/topics/adversarial-attacks.md>), [LLM security](<https://devfeed.tech/topics/llm-security.md>), [Machine Intelligence](<https://devfeed.tech/topics/machine-intelligence.md>)

Tags: [adaptive](<https://devfeed.tech/tags/adaptive.md>), [adversarial](<https://devfeed.tech/tags/adversarial.md>), [agentic](<https://devfeed.tech/tags/agentic.md>), [attack](<https://devfeed.tech/tags/attack.md>), [cs-cr](<https://devfeed.tech/tags/cs-cr.md>), [cs-ir](<https://devfeed.tech/tags/cs-ir.md>), [datasets](<https://devfeed.tech/tags/datasets.md>), [detection](<https://devfeed.tech/tags/detection.md>), [framework](<https://devfeed.tech/tags/framework.md>), [llm](<https://devfeed.tech/tags/llm.md>), [memory](<https://devfeed.tech/tags/memory.md>), [promotion](<https://devfeed.tech/tags/promotion.md>), [reasoning](<https://devfeed.tech/tags/reasoning.md>), [recommendation](<https://devfeed.tech/tags/recommendation.md>), [recommender-systems](<https://devfeed.tech/tags/recommender-systems.md>), [reflection](<https://devfeed.tech/tags/reflection.md>), [robust](<https://devfeed.tech/tags/robust.md>)

### AI overview

GuardAgent targets promotion-driven attacks on LLM-based recommender systems, where text drives recommendation through semantic understanding and reasoning, creating a harder-to-detect attack surface. The proposed LLM-powered detector loops through Perception, Action, Reflection, and Memory, using continual learning to adapt to unseen threat patterns and producing explainable reasoning traces and outcomes. The abstract reports state-of-the-art generalization in experiments on real-world datasets; it provides no specific metrics or implementation details.

### Source excerpt

arXiv:2611.22274v1 Announce Type: cross Abstract: In traditional recommender systems (RSs), textual information typically serves as auxiliary signals for representation learning. In contrast, in LLM-based RSs, textual information becomes the primary medium for recommendation generation through semantic understanding and reasoning. This shift fundamentally changes the attack surface, making promotion-driven manipulations more feasible, effective, and harder to detect. However, effectively detecting such attacks remains largely unexplored. To fill this gap, we present the first systematic study on detecting promotion-driven attacks against LLM-based RSs, aiming to safeguard next-generation recommendation platforms. Specifically, we propose GuardAgent, a novel LLM-powered agentic detection framework that operates through a loop of Perception, Action, Reflection and Memory. In contrast to conventional LLM-based detectors that function in a static, one-off, and opaque fashion, GuardAgent continually learns and self-evolves to adapt to unseen threat patterns while generating explainable reasoning traces and outcomes. Extensive experiments on real-world datasets demonstrate that GuardAgent achieves state-of-the-art generalization performance, laying the foundation for robust and trustworthy recommendation in practical settings.

## Dsat: A Native SAT Solver for Discrete Logic

DevFeed: [Dsat: A Native SAT Solver for Discrete Logic](<https://devfeed.tech/articles/dsat-a-native-sat-solver-for-discrete-logic-137252.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2605.09347>)

Author: Yaofang Zhang, Ken Zhou, Adnan Darwiche

Published: 2026-11-17T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Artificial Intelligence](<https://devfeed.tech/sources/arxiv-artificial-intelligence.md>)

Topics: [Artificial Intelligence](<https://devfeed.tech/topics/ai.md>), [Machine Intelligence](<https://devfeed.tech/topics/machine-intelligence.md>)

Tags: [ai](<https://devfeed.tech/tags/ai.md>), [arxiv](<https://devfeed.tech/tags/arxiv.md>), [boolean](<https://devfeed.tech/tags/boolean.md>), [cs-ai](<https://devfeed.tech/tags/cs-ai.md>), [cs-lo](<https://devfeed.tech/tags/cs-lo.md>), [csp](<https://devfeed.tech/tags/csp.md>), [hybrid](<https://devfeed.tech/tags/hybrid.md>), [logic](<https://devfeed.tech/tags/logic.md>), [native](<https://devfeed.tech/tags/native.md>), [planning](<https://devfeed.tech/tags/planning.md>), [probabilistic](<https://devfeed.tech/tags/probabilistic.md>), [reasoning](<https://devfeed.tech/tags/reasoning.md>), [symbolic-reasoning](<https://devfeed.tech/tags/symbolic-reasoning.md>)

### AI overview

Dsat is a SAT solver for discrete logic, where variables can take arbitrary values rather than only true or false. It applies techniques akin to unit resolution and clause learning directly to discrete variables, avoiding the usual step of encoding them as Boolean variables--a step the authors say can pose computational and semantic challenges. The abstract says Dsat is evaluated against CSP solvers on discrete CNFs, Boolean SAT solvers on binarized CNFs, and hybrid solvers, but provides no results or performance details.

### Source excerpt

arXiv:2605.09347v2 Announce Type: replace Abstract: Discrete variables are common in many applications, such as probabilistic reasoning, planning and explainable AI. When symbolic reasoning techniques are brought in to bear on these applications, a standard technique for handling discrete variables is to binarize them into Boolean variables to allow the use of Boolean computational machinery such as SAT solvers. This technique can face both computational and semantical challenges though. In this work, we develop a native SAT solver for discrete logic, which is a direct extension of Boolean logic in which variables can take arbitrary values. Our proposed solver has a similar design to Boolean SAT solvers, with ingredients such as unit resolution and clause learning but ones that operate natively on discrete variables. We illustrate the merits of the developed SAT solver by comparing it empirically to CSP solvers applied to discrete CNFs, to Boolean SAT solver applied to binarized CNFs, and to some hybrid solvers.

## Isolating Perception for Multi-View Planning under Occlusion-Induced Partial Observability

DevFeed: [Isolating Perception for Multi-View Planning under Occlusion-Induced Partial Observability](<https://devfeed.tech/articles/isolating-perception-for-multi-view-planning-under-occlusion-induced-partial-observability-137412.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.19179>)

Author: Kevin Nguyen, Ceyhun Efe Kayan, Feng Liu, Li Zhang

Published: 2026-11-17T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Artificial Intelligence](<https://devfeed.tech/sources/arxiv-artificial-intelligence.md>)

Topics: [Machine Intelligence](<https://devfeed.tech/topics/machine-intelligence.md>), [LLM evaluation / benchmarking](<https://devfeed.tech/topics/llm-evaluation-benchmarking.md>), [llm-reasoning](<https://devfeed.tech/topics/llm-reasoning.md>)

Tags: [benchmark](<https://devfeed.tech/tags/benchmark.md>), [cs-ai](<https://devfeed.tech/tags/cs-ai.md>), [cs-cl](<https://devfeed.tech/tags/cs-cl.md>), [cs-cv](<https://devfeed.tech/tags/cs-cv.md>), [gemini-3-1-pro](<https://devfeed.tech/tags/gemini-3-1-pro.md>), [gemma-4-31b](<https://devfeed.tech/tags/gemma-4-31b.md>), [performance](<https://devfeed.tech/tags/performance.md>), [planning](<https://devfeed.tech/tags/planning.md>), [qwen3-5](<https://devfeed.tech/tags/qwen3-5.md>), [vision-language-models-vlms](<https://devfeed.tech/tags/vision-language-models-vlms.md>)

### AI overview

The paper reports that separating local visual perception from planning and cross-view evidence reconciliation improves performance on BW-PO, a Blocksworld benchmark designed around occlusion-induced partial observability. In the proposed VLM-as-perceiver setup, external solvers handle reconciliation and plan generation; on model-specific challenging subsets, success rises from 1% to 97% for Gemma-4-31B, 0% to 95% for Qwen3.5-122B-A10B, and 3% to 84% for Gemini-3.1-Pro. The authors frame depth-ambiguity reconciliation and completeness of visual extraction in dense scenes as distinct bottlenecks. The supplied text does not describe implementation details or broader evaluation.

### Source excerpt

arXiv:2611.19179v1 Announce Type: cross Abstract: Vision-language models (VLMs) are beginning to saturate visual planning benchmarks, yet they fail to resolve depth ambiguity and degrade sharply with problem size under occlusion-induced partial observability (PO). We introduce BW-PO based on the Blocksworld domain, a controlled benchmark that reduces existing VLM planners and formalizers to near-zero performance, and propose VLM-as-perceiver, a paradigm that narrows the VLM's role to local perception while delegating cross-view evidence reconciliation and plan generation to external solvers. Across model-specific challenging subsets of BW-PO, this raises plan success from 1% to 97% for Gemma-4-31B, 0% to 95% for Qwen3.5-122B-A10B, and 3% to 84% for Gemini-3.1-Pro. Isolating perception in this way separates two bottlenecks that are otherwise entangled: reconciliation under depth ambiguity, and completeness of visual extraction under scene density.

## Neural Simulation-Based Inference for Computational Cognitive Models: A Tutorial

DevFeed: [Neural Simulation-Based Inference for Computational Cognitive Models: A Tutorial](<https://devfeed.tech/articles/neural-simulation-based-inference-for-computational-cognitive-models-a-tutorial-135806.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.17608>)

Author: Ti-Fen Pan, Yi Cheng, Anne GE Collins

Published: 2026-11-16T05:00:00Z

Content type: tutorial

Language: en

Sources: [arXiv: Machine Learning](<https://devfeed.tech/sources/arxiv-machine-learning.md>)

Topics: [Artificial Intelligence](<https://devfeed.tech/topics/ai.md>), [Neural Network](<https://devfeed.tech/topics/neural-network.md>), [Machine Intelligence](<https://devfeed.tech/topics/machine-intelligence.md>)

Tags: [cs-lg](<https://devfeed.tech/tags/cs-lg.md>), [inference](<https://devfeed.tech/tags/inference.md>), [models](<https://devfeed.tech/tags/models.md>), [neural](<https://devfeed.tech/tags/neural.md>), [q-bio-nc](<https://devfeed.tech/tags/q-bio-nc.md>), [simulation](<https://devfeed.tech/tags/simulation.md>), [tutorial](<https://devfeed.tech/tags/tutorial.md>)

### AI overview

Neural simulation-based inference (SBI) trains neural networks on simulated data to learn the relationship between observations and cognitive models, enabling parameter estimation and model comparison without computing likelihoods. The tutorial presents a workflow for applying neural SBI to cognitive modeling and practical recommendations for reducing simulation and computational burden. The supplied abstract does not detail the workflow or use cases.

### Source excerpt

arXiv:2611.17608v1 Announce Type: new Abstract: Computational cognitive modeling is an increasingly essential methodology to elucidate the mechanisms underlying human and animal cognition. By fitting models to empirical data, researchers can perform parameter estimation, model comparison, and model-based neural analyses to assess competing theories, quantify individual differences, and investigate neural representations. Typically, fitting these models relies on likelihood-dependent methods (e.g., Maximum likelihood estimation or hierarchical Bayesian methods). However, for many reasonable cognitive models, computing the likelihood is computationally intractable, in which case these methods become inadequate. This substantially limits the range of theories that researchers can investigate. Simulation-based inference with neural networks (neural SBI) has emerged as a powerful toolkit to overcome this hurdle. By training deep neural networks on datasets generated through model simulations, neural SBI learns the relation between observable data and cognitive models. This approach enables all the applications of standard modeling while bypassing the likelihood computation, without constraints on empirical data size. Despite its successful adoption across diverse scientific fields, neural SBI remains underutilized in cognitive modeling, partly due to a lack of accessible, domain-tailored guidance. This tutorial addresses this gap by providing a clear workflow for applying neural SBI to cognitive models. We also provide practical use cases and recommendations for reducing the computational and simulation burden, facilitating efficient implementation in real-world research settings. By lowering the technical barriers to neural SBI, this tutorial aims to empower cognitive scientists to investigate richer theory spaces and ultimately propel scientific discovery.

## Courteous Anticipation: Improving Long-Lived Task Planning in Persistent Shared Environments

DevFeed: [Courteous Anticipation: Improving Long-Lived Task Planning in Persistent Shared Environments](<https://devfeed.tech/articles/courteous-anticipation-improving-long-lived-task-planning-in-persistent-shared-environments-134997.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2607.20289>)

Author: Md Ridwan Hossain Talukder, Elizabeth Phillips, Gregory J. Stein

Published: 2026-11-16T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Artificial Intelligence](<https://devfeed.tech/sources/arxiv-artificial-intelligence.md>)

Topics: [Machine Intelligence](<https://devfeed.tech/topics/machine-intelligence.md>)

Tags: [candidate](<https://devfeed.tech/tags/candidate.md>), [cost](<https://devfeed.tech/tags/cost.md>), [cs-ai](<https://devfeed.tech/tags/cs-ai.md>), [cs-ro](<https://devfeed.tech/tags/cs-ro.md>), [domains](<https://devfeed.tech/tags/domains.md>), [estimator](<https://devfeed.tech/tags/estimator.md>), [future](<https://devfeed.tech/tags/future.md>), [home](<https://devfeed.tech/tags/home.md>), [independent](<https://devfeed.tech/tags/independent.md>), [model](<https://devfeed.tech/tags/model.md>), [modular](<https://devfeed.tech/tags/modular.md>), [planning](<https://devfeed.tech/tags/planning.md>), [reduce](<https://devfeed.tech/tags/reduce.md>), [restaurant](<https://devfeed.tech/tags/restaurant.md>), [robots](<https://devfeed.tech/tags/robots.md>), [tasks](<https://devfeed.tech/tags/tasks.md>)

### AI overview

Courteous anticipatory planning lowers cumulative task costs by accounting for how a robot's current plan affects future tasks for all robots sharing a persistent environment. A model-based planner compares candidate plans using immediate cost plus expected future costs estimated by separate learned estimators for each robot, avoiding joint rollouts; adding a robot requires training its own estimator. In PDDL evaluations, it reduced total cost versus myopic and selfish anticipatory planning by 10.4% and 4.5% in a two-robot home domain, and by 20.5% and 6.7% in a three-robot restaurant domain. The supplied abstract reports these results for the evaluated settings.

### Source excerpt

arXiv:2607.20289v2 Announce Type: replace-cross Abstract: We consider a task planning scenario in which robots sharing a persistent environment are assigned tasks one at a time from a held-out sequence. Standard task planners, lacking foresight of future tasks and inconsiderate of others' constraints, solve each task in isolation, leaving terminal states that increase future cost for all; side effects that compound over lengthy task sequences. To reduce cost over the sequence, a robot must anticipate how its actions now may impact performance on future tasks for all robots sharing the environment. Therefore, we present courteous anticipatory planning, wherein a model-based planner proposes candidate plans and selects the one that jointly minimizes immediate cost and aggregated expected future cost across all robots, estimated via independent per-robot learned estimators. This factored formulation avoids combinatorial joint rollouts and supports modular deployment: adding a robot requires only training its own estimator. We evaluate in two persistent PDDL domains, a home environment with robots that have similar capabilities but different responsibilities, and a restaurant environment where robots' distinct capabilities create states that other robots lack the capability to resolve. During lengthy task sequences, our planner reduces total cost by 10.4% versus myopic and 4.5% versus selfish anticipatory planning in a two-robot home environment and by 20.5% and 6.7%, respectively, in a three-robot restaurant.

## Marginal-Preserving Coupled Diffusion Composition for Long-Horizon Planning

DevFeed: [Marginal-Preserving Coupled Diffusion Composition for Long-Horizon Planning](<https://devfeed.tech/articles/marginal-preserving-coupled-diffusion-composition-for-long-horizon-planning-132147.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.15585>)

Author: Hengrui Zhang, Qiyu Wang, Yuhu Cheng, Brooks Paige, Xuesong Wang

Published: 2026-11-13T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Artificial Intelligence](<https://devfeed.tech/sources/arxiv-artificial-intelligence.md>)

Topics: [Machine Intelligence](<https://devfeed.tech/topics/machine-intelligence.md>)

Tags: [composition](<https://devfeed.tech/tags/composition.md>), [coupling](<https://devfeed.tech/tags/coupling.md>), [cs-ai](<https://devfeed.tech/tags/cs-ai.md>), [diffusion](<https://devfeed.tech/tags/diffusion.md>), [graph](<https://devfeed.tech/tags/graph.md>), [horizon](<https://devfeed.tech/tags/horizon.md>), [inference](<https://devfeed.tech/tags/inference.md>), [multimodal](<https://devfeed.tech/tags/multimodal.md>), [noise](<https://devfeed.tech/tags/noise.md>), [overlap](<https://devfeed.tech/tags/overlap.md>), [planning](<https://devfeed.tech/tags/planning.md>), [sampling](<https://devfeed.tech/tags/sampling.md>), [search](<https://devfeed.tech/tags/search.md>), [training](<https://devfeed.tech/tags/training.md>), [trajectory-planning](<https://devfeed.tech/tags/trajectory-planning.md>)

### AI overview

MCDC coordinates long-horizon trajectory chunks by positively correlating neighboring chunks' initial diffusion noise while preserving each chunk's exact standard-Gaussian marginal, as required by the pretrained denoiser. This training-free inference method aims to make adjacent chunks choose compatible overlap modes without adding separate search or task-specific machinery. The authors prove marginal preservation and, under a simple multimodal overlap model, show that positive coupling raises the probability of compatible mode selection. On OGBench long-horizon stitch tasks, they report consistent improvement over independent sampling at nearly the same inference time and a favorable success-cost trade-off against search-based composition methods.

### Source excerpt

arXiv:2611.15585v1 Announce Type: new Abstract: Compositional diffusion planners extend short-horizon trajectory models to long-horizon goal-conditioned planning by generating and stitching multiple local trajectory chunks. However, in multimodal environments, local plausibility is not sufficient for global consistency: independently initialized chunks may denoise toward incompatible overlap modes, producing plans that are individually feasible but incoherent after stitching. Existing remedies typically add search, guidance, pruning, or graph-based structure at inference time, increasing computational cost or relying on task-specific machinery. We instead investigate graph-structured dependence among chunk-wise initial diffusion noises as a lightweight coordination mechanism for compositional trajectory planning. We propose \textbf{M}arginal-Preserving \textbf{C}oupled \textbf{D}iffusion \textbf{C}omposition (MCDC), a training-free inference method for pretrained compositional diffusion planners. Instead of sampling chunk noises independently, MCDC constructs coupled Gaussian initializations whose per-chunk marginals remain exactly standard Gaussian, as required by the pretrained denoiser, while neighboring chunks receive positively correlated noise components. This preserves the standard-Gaussian initialization marginal of each chunk while biasing adjacent chunks toward compatible overlap-mode choices. We prove exact marginal preservation and show, under a simple multimodal overlap model, that positive coupling increases the probability of compatible mode selection. On OGBench long-horizon stitch tasks, MCDC consistently improves over independent compositional diffusion sampling while retaining nearly the same inference time, and achieves a favorable success--cost trade-off compared with search-based composition methods.

## Robust Task-Aware State Estimation from Pre-trained Perception

DevFeed: [Robust Task-Aware State Estimation from Pre-trained Perception](<https://devfeed.tech/articles/robust-task-aware-state-estimation-from-pre-trained-perception-132050.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.14781>)

Author: Guy Azran, Sarah Keren

Published: 2026-11-13T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Artificial Intelligence](<https://devfeed.tech/sources/arxiv-artificial-intelligence.md>)

Topics: [Artificial Intelligence](<https://devfeed.tech/topics/ai.md>), [Machine Intelligence](<https://devfeed.tech/topics/machine-intelligence.md>), [AI Models](<https://devfeed.tech/topics/ai-models.md>), [Inference](<https://devfeed.tech/topics/inference.md>), [computational complexity](<https://devfeed.tech/topics/computational-complexity.md>)

Tags: [agent](<https://devfeed.tech/tags/agent.md>), [agents](<https://devfeed.tech/tags/agents.md>), [algorithm](<https://devfeed.tech/tags/algorithm.md>), [arxiv](<https://devfeed.tech/tags/arxiv.md>), [complexity](<https://devfeed.tech/tags/complexity.md>), [confidence](<https://devfeed.tech/tags/confidence.md>), [construction](<https://devfeed.tech/tags/construction.md>), [cs-ai](<https://devfeed.tech/tags/cs-ai.md>), [decision](<https://devfeed.tech/tags/decision.md>), [decision-making](<https://devfeed.tech/tags/decision-making.md>), [distribution](<https://devfeed.tech/tags/distribution.md>), [distributions](<https://devfeed.tech/tags/distributions.md>), [efficiently](<https://devfeed.tech/tags/efficiently.md>), [estimator](<https://devfeed.tech/tags/estimator.md>), [feature](<https://devfeed.tech/tags/feature.md>), [full](<https://devfeed.tech/tags/full.md>), [guarantees](<https://devfeed.tech/tags/guarantees.md>), [improvement](<https://devfeed.tech/tags/improvement.md>), [language-models](<https://devfeed.tech/tags/language-models.md>), [robust](<https://devfeed.tech/tags/robust.md>), [state](<https://devfeed.tech/tags/state.md>)

### AI overview

The paper proposes retaining a threshold-sized set of plausible states from uncertain pretrained perception instead of committing to one best guess or representing the full state distribution. Its θ-confidence Minimal Belief Set contains the smallest set of states whose combined probability meets a user-selected threshold. A three-layer belief system approximates state probabilities from per-feature predictions and task constraints while keeping a factored representation; PRaVaSS builds the set with correctness guarantees and polynomial complexity in the unconstrained case. With a VLM-based estimator, the authors report validating theoretical properties, scalability, and better downstream task-planning decisions. The supplied text is an abstract and does not give quantitative results or experimental details.

### Source excerpt

arXiv:2611.14781v1 Announce Type: new Abstract: Agents that perform state estimation using pretrained perception models, e.g., Vision-Language Models (VLMs), face the fundamental challenge that predictions are uncertain. Still, most approaches determinize these predictions into a single best-guess state and make decisions as if it were ground truth. This is brittle because when the guess is wrong, the agent's decisions may be unsafe or infeasible. At the other extreme, maintaining a full probability distribution over symbolic states is often intractable. We propose a middle path, drawing on the underlying probabilities the perception model already produces (e.g., a VLM's output-layer token distributions) to construct a $\theta$-confidence Minimal Belief Set (MBS$_\theta$), a minimal subset of states whose cumulative probability meets a user-specified belief-mass threshold. The minimality of an MBS$_\theta$ ensures that the agent considers only the most likely states, enhancing robustness in downstream decision-making while maintaining computational tractability. Using a novel three-layered belief system, we efficiently approximate individual state probabilities from per-feature predictions and task-level constraints. On this belief construction, we base our MBS$_\theta$ construction, while maintaining only a lean factored representation of the belief. Our algorithm, Probability-Ranked Valid State Search (PRaVaSS), takes advantage of the unique structure of this belief system to construct an MBS$_\theta$ with correctness guarantees and polynomial complexity in the unconstrained case. Using a VLM as the backbone of our state estimator, we validate our method's theoretical properties, scalability with problem size, and improvement of downstream decision-making when used in a task-planning pipeline.

## Rational Use of Time for Social Coordination in Humans and AI

DevFeed: [Rational Use of Time for Social Coordination in Humans and AI](<https://devfeed.tech/articles/rational-use-of-time-for-social-coordination-in-humans-and-ai-132047.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.14766>)

Author: Xiulin Yang, Sirisha Gudavalli, Jennifer Hu

Published: 2026-11-13T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Artificial Intelligence](<https://devfeed.tech/sources/arxiv-artificial-intelligence.md>)

Topics: [Artificial Intelligence](<https://devfeed.tech/topics/ai.md>), [Multi Agent Systems](<https://devfeed.tech/topics/multi-agent-systems.md>), [Machine Intelligence](<https://devfeed.tech/topics/machine-intelligence.md>), [Machine learning](<https://devfeed.tech/topics/machine-learning.md>)

Tags: [ai](<https://devfeed.tech/tags/ai.md>), [bayesian](<https://devfeed.tech/tags/bayesian.md>), [card-game](<https://devfeed.tech/tags/card-game.md>), [communication](<https://devfeed.tech/tags/communication.md>), [computer-use](<https://devfeed.tech/tags/computer-use.md>), [cs-ai](<https://devfeed.tech/tags/cs-ai.md>), [cs-cl](<https://devfeed.tech/tags/cs-cl.md>), [frontier-models](<https://devfeed.tech/tags/frontier-models.md>), [learning](<https://devfeed.tech/tags/learning.md>), [llms](<https://devfeed.tech/tags/llms.md>), [probabilistic](<https://devfeed.tech/tags/probabilistic.md>), [real-time](<https://devfeed.tech/tags/real-time.md>), [signal](<https://devfeed.tech/tags/signal.md>), [social](<https://devfeed.tech/tags/social.md>), [strategy](<https://devfeed.tech/tags/strategy.md>), [temporal](<https://devfeed.tech/tags/temporal.md>), [time](<https://devfeed.tech/tags/time.md>), [timing](<https://devfeed.tech/tags/timing.md>)

### AI overview

In The Mind, a cooperative card game that allows no explicit communication, researchers compared 150 humans with computer-use agents and text-prompted LLMs to study coordination through timing. Human behavior was best captured by a Bayesian strategy that inferred a partner's beliefs from their timing; frontier models were described similarly. However, none of the tested models appeared to adapt efficiently to their partners over repeated play. The abstract suggests humans can reason probabilistically about continuous-time social cues, and identifies interpreting and adapting to implicit signals as a challenge for AI coordination.

### Source excerpt

arXiv:2611.14766v1 Announce Type: new Abstract: All social interactions are situated in time. Previous studies have shown that humans are able to use implicit temporal cues, such as the duration of a pause, to infer others' mental states. But how do humans use temporal cues for dynamic, fine-grained social coordination? And can AI systems do the same, without ``experiencing'' continuous real-time as humans do? We study these questions using a cooperative card game task called The Mind, which requires players to achieve a shared goal across repeated interactions but forbids explicit communication: the only available signal is the passage of time. We evaluate 150 humans, and two classes of AI models which reflect different ways of operationalizing time in artificial systems: computer-use agents (CUAs), and standard text-prompted LLMs. We find that human behavior is best captured by a Bayesian strategy that infers a partner's belief from their timing---and similarly for frontier models. However, unlike humans, none of the tested models seem to efficiently adapt their behaviors to their partners' over the course of the game. From a cognitive perspective, our findings suggest that humans perform implicit probabilistic reasoning about continuous-time signals in dynamic social settings, and that such reasoning mechanisms may emerge without real-time experience. Practically, our study highlights new challenges for AI systems to coordinate efficiently with other agents by learning to interpret and adapt to their implicit signals in dynamic settings.

## From Foundation VLM to Versatile Driving VLM: A Triplet Incremental Training Approach

DevFeed: [From Foundation VLM to Versatile Driving VLM: A Triplet Incremental Training Approach](<https://devfeed.tech/articles/from-foundation-vlm-to-versatile-driving-vlm-a-triplet-incremental-training-approach-132277.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.16727>)

Author: Shengtao Wu, Zichao Dong, Weixuan Qiao, Xiaolu Liu, Xin Zhan, Junbo Chen, Jianke Zhu

Published: 2026-11-13T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Artificial Intelligence](<https://devfeed.tech/sources/arxiv-artificial-intelligence.md>)

Topics: [autonomous vehicles](<https://devfeed.tech/topics/autonomous-vehicles.md>), [AI Foundation Models](<https://devfeed.tech/topics/ai-foundation-models.md>), [Machine Intelligence](<https://devfeed.tech/topics/machine-intelligence.md>)

Tags: [3d](<https://devfeed.tech/tags/3d.md>), [benchmarks](<https://devfeed.tech/tags/benchmarks.md>), [capabilities](<https://devfeed.tech/tags/capabilities.md>), [cs-ai](<https://devfeed.tech/tags/cs-ai.md>), [cs-cv](<https://devfeed.tech/tags/cs-cv.md>), [driving](<https://devfeed.tech/tags/driving.md>), [foundation-models](<https://devfeed.tech/tags/foundation-models.md>), [framework](<https://devfeed.tech/tags/framework.md>), [general](<https://devfeed.tech/tags/general.md>), [grounding](<https://devfeed.tech/tags/grounding.md>), [incremental](<https://devfeed.tech/tags/incremental.md>), [inference](<https://devfeed.tech/tags/inference.md>), [language-models](<https://devfeed.tech/tags/language-models.md>), [multimodal](<https://devfeed.tech/tags/multimodal.md>), [planning](<https://devfeed.tech/tags/planning.md>), [reasoning](<https://devfeed.tech/tags/reasoning.md>), [robust](<https://devfeed.tech/tags/robust.md>), [semantic](<https://devfeed.tech/tags/semantic.md>), [spatial](<https://devfeed.tech/tags/spatial.md>), [stream](<https://devfeed.tech/tags/stream.md>), [training](<https://devfeed.tech/tags/training.md>), [vlm](<https://devfeed.tech/tags/vlm.md>)

### AI overview

The paper argues that general VLMs can adapt to varied autonomous-driving tasks, with precise metric depth as a key gap. Its Triplet Incremental Training framework uses image, depth, and question-answer triplets in three streams: an unfrozen VLM adapts to driving tasks, an auxiliary depth stream applies geometric constraints, and a frozen twin VLM supplies logit-level KL distillation to limit semantic drift. The depth branch is removed at deployment, so it adds no inference overhead. Across four benchmarks in 3D perception, driving language reasoning, end-to-end planning, and general multimodal understanding, the authors report outperforming state-of-the-art driving VLMs, matching domain-specific experts in 3D perception, and retaining general reasoning. These are abstract-level claims; detailed benchmark results are not supplied.

### Source excerpt

arXiv:2611.16727v1 Announce Type: cross Abstract: Adapting general Vision-Language Models (VLMs) to Autonomous Driving (AD) is thought to require heavy 3D spatial priors or intrusive architectural overhauls to compensate for a lack of spatial grounding. In this work, we present a counterintuitive pilot study: off-the-shelf general VLMs already exhibit remarkable adaptability across diverse driving tasks, while precise metric depth remains the primary missing link, a bottleneck readily addressable via onboard sensors and depth foundation models. To this end, we introduce the Triplet Incremental Training Framework, a non-intrusive paradigm that seamlessly instills 3D geometric awareness into a general VLM backbone while fully preserving its open-world capabilities. Driven by automatically constructed (image, depth, QA pair) triplets, our framework operates via a unified tri-stream training pipeline: an unfrozen primary VLM stream to adapt to driving tasks, an auxiliary depth stream that guides the visual backbone to implicitly imagine 3D scene geometry via explicit geometric constraints, and a frozen twin VLM stream employing logit-level KL distillation to prevent semantic drift. Crucially, the auxiliary depth branch is discarded during deployment, ensuring the enhanced geometric awareness comes with zero inference overhead. We evaluate our framework across four major benchmarks targeting four diverse capabilities spanning 3D spatial perception, driving-oriented language reasoning, end-to-end planning, and general multimodal understanding. Our approach outperforms state-of-the-art driving VLMs, rivals domain-specific expert models in 3D perception, and successfully retains its original general reasoning capabilities. This demonstrates that Triplet Incremental Training offers a unified, deployment-ready paradigm for transforming general foundation models into robust driving intelligences.

## Scaling Intelligence: A landmark commission and exhibition

DevFeed: [Scaling Intelligence: A landmark commission and exhibition](<https://devfeed.tech/articles/scaling-intelligence-a-landmark-commission-and-exhibition-133572.md>)

Original publisher: [Read original article](<https://modal.com/blog/scaling-intelligence>)

Author: Peri Langlois

Published: 2026-11-13T00:00:00Z

Content type: article

Language: en

Sources: [Modal](<https://devfeed.tech/sources/modal.md>)

Topics: [Artificial Intelligence](<https://devfeed.tech/topics/ai.md>), [Machine Intelligence](<https://devfeed.tech/topics/machine-intelligence.md>), [AI Infrastructure](<https://devfeed.tech/topics/ai-infrastructure.md>)

Tags: [agent](<https://devfeed.tech/tags/agent.md>), [agents](<https://devfeed.tech/tags/agents.md>), [ai](<https://devfeed.tech/tags/ai.md>), [ai-generated-images](<https://devfeed.tech/tags/ai-generated-images.md>), [ai-infrastructure](<https://devfeed.tech/tags/ai-infrastructure.md>), [ai-model](<https://devfeed.tech/tags/ai-model.md>), [ai-models](<https://devfeed.tech/tags/ai-models.md>), [arm](<https://devfeed.tech/tags/arm.md>), [art](<https://devfeed.tech/tags/art.md>), [budget](<https://devfeed.tech/tags/budget.md>), [cats](<https://devfeed.tech/tags/cats.md>), [cloud](<https://devfeed.tech/tags/cloud.md>), [compute](<https://devfeed.tech/tags/compute.md>), [engineering](<https://devfeed.tech/tags/engineering.md>), [exhibition](<https://devfeed.tech/tags/exhibition.md>), [intersection](<https://devfeed.tech/tags/intersection.md>), [machine-intelligence](<https://devfeed.tech/tags/machine-intelligence.md>), [nonprofit](<https://devfeed.tech/tags/nonprofit.md>), [san-francisco](<https://devfeed.tech/tags/san-francisco.md>), [technology](<https://devfeed.tech/tags/technology.md>)

### AI overview

Scaling Intelligence presents 11 artworks that use AI models, agents, sensors, and cloud compute as creative materials, often making their limits and effects part of the work. A satellite-image loop searches endlessly for a nonexistent Arctic island; creek sensors feed a model that builds a changing state for an LED installation; and a choir of language models learns from visitors but falls silent when its compute budget runs out. Other works let visitors create glyphs with a model trained on ancient Maya writing, retrain a conversational agent nightly on public interactions, or observe groups of agents forming rules and institutions. Across the exhibition, AI translates otherwise imperceptible signals, responds to participation, or runs open-ended processes, while several pieces foreground dependence on people, data, or finite computation.

### Source excerpt

Last week, Modal and Gray Area, the San Francisco nonprofit that has supported artists working at the intersection of art and technology for nearly two decades, opened *Scaling Intelligence*, an exhibition of 11 new artworks made with machine intelligence and cloud-scale compute. Over the past few months, we've worked closely with the artists, supporting them with engineering help and compute credits as they built everything from scratch.

## How to Win Back Your Agency from Cognitive Surrender

DevFeed: [How to Win Back Your Agency from Cognitive Surrender](<https://devfeed.tech/articles/how-to-win-back-your-agency-from-cognitive-surrender-128948.md>)

Original publisher: [Read original article](<https://spin.atomicobject.com/preserve-shared-understanding/>)

Author: Matt Sweda

Published: 2026-11-11T13:00:22Z

Content type: opinion

Language: en

Sources: [Atomic Object](<https://devfeed.tech/sources/atomic-object.md>)

Topics: [engineering-culture](<https://devfeed.tech/topics/engineering-culture.md>), [Machine Intelligence](<https://devfeed.tech/topics/machine-intelligence.md>), [Pull Request](<https://devfeed.tech/topics/pull-request.md>), [AI Context Management](<https://devfeed.tech/topics/ai-context-management.md>)

Tags: [agentic-software](<https://devfeed.tech/tags/agentic-software.md>), [ai](<https://devfeed.tech/tags/ai.md>), [ai-agent](<https://devfeed.tech/tags/ai-agent.md>), [ai-assistance](<https://devfeed.tech/tags/ai-assistance.md>), [ai-for-developers](<https://devfeed.tech/tags/ai-for-developers.md>), [artificial-intelligence](<https://devfeed.tech/tags/artificial-intelligence.md>), [cognitive-surrender](<https://devfeed.tech/tags/cognitive-surrender.md>), [coherence](<https://devfeed.tech/tags/coherence.md>), [conversations](<https://devfeed.tech/tags/conversations.md>), [craft](<https://devfeed.tech/tags/craft.md>), [debugging](<https://devfeed.tech/tags/debugging.md>), [development](<https://devfeed.tech/tags/development.md>), [discovery](<https://devfeed.tech/tags/discovery.md>), [documentation](<https://devfeed.tech/tags/documentation.md>), [feedback](<https://devfeed.tech/tags/feedback.md>), [knowledge-transfer](<https://devfeed.tech/tags/knowledge-transfer.md>), [llm](<https://devfeed.tech/tags/llm.md>)

### AI overview

The author argues that AI-assisted coding can weaken shared understanding when developers use prompts in place of conversations with teammates or clients. LLMs can help produce code quickly, but that speed may shift the effort of understanding to reviewers and maintainers, while different assumptions surface later as incompatible implementations. Specs help document context but cannot ensure people interpret it alike or resolve ambiguity. The author favors writing implementations oneself and using LLMs to find holes, generate counterexamples, and answer questions, while keeping human judgment central. They recommend questioning unclear output, clarifying assumptions with others, and holding AI-generated changes to the same standards of explanation and review as other code.

### Source excerpt

If you haven't yet read The Decoherence Trap by Kristen Maevyn, I highly suggest you do, because in a way this is my extension of those great ideas. Maevyn describes how individually reasonable changes can accumulate into a system that becomes increasingly difficult to change. What I want to explore is how we're losing the [...] The post How to Win Back Your Agency from Cognitive Surrender appeared first on Atomic Spin.

## Cognitive Multi-Agent Search and Rescue: Language-Grounded Hierarchical Reinforcement Learning for Collaborative Robotics

DevFeed: [Cognitive Multi-Agent Search and Rescue: Language-Grounded Hierarchical Reinforcement Learning for Collaborative Robotics](<https://devfeed.tech/articles/cognitive-multi-agent-search-and-rescue-language-grounded-hierarchical-reinforcement-learning-for-collaborative-robotics-128263.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.13161>)

Author: Sepideh Ravanbakhsh, Hamidreza Moradi

Published: 2026-11-11T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Artificial Intelligence](<https://devfeed.tech/sources/arxiv-artificial-intelligence.md>)

Topics: [Multi Agent Systems](<https://devfeed.tech/topics/multi-agent-systems.md>), [Reinforcement learning](<https://devfeed.tech/topics/reinforcement-learning.md>), [Artificial Intelligence](<https://devfeed.tech/topics/ai.md>), [Large Language Model](<https://devfeed.tech/topics/llm.md>), [Agentic AI](<https://devfeed.tech/topics/what-is-agentic-ai.md>), [Language Engineering](<https://devfeed.tech/topics/language-engineering.md>), [Machine Intelligence](<https://devfeed.tech/topics/machine-intelligence.md>)

Tags: [ablation](<https://devfeed.tech/tags/ablation.md>), [abstract](<https://devfeed.tech/tags/abstract.md>), [adapting](<https://devfeed.tech/tags/adapting.md>), [agent](<https://devfeed.tech/tags/agent.md>), [agent-behavior](<https://devfeed.tech/tags/agent-behavior.md>), [agent-framework](<https://devfeed.tech/tags/agent-framework.md>), [agents](<https://devfeed.tech/tags/agents.md>), [arxiv](<https://devfeed.tech/tags/arxiv.md>), [autonomous](<https://devfeed.tech/tags/autonomous.md>), [autonomous-systems](<https://devfeed.tech/tags/autonomous-systems.md>), [based](<https://devfeed.tech/tags/based.md>), [behavior](<https://devfeed.tech/tags/behavior.md>), [coherence](<https://devfeed.tech/tags/coherence.md>), [combined](<https://devfeed.tech/tags/combined.md>), [cs-ai](<https://devfeed.tech/tags/cs-ai.md>), [cs-ce](<https://devfeed.tech/tags/cs-ce.md>), [language](<https://devfeed.tech/tags/language.md>), [learning](<https://devfeed.tech/tags/learning.md>), [multi-agent](<https://devfeed.tech/tags/multi-agent.md>), [reinforcement-learning](<https://devfeed.tech/tags/reinforcement-learning.md>), [robotics](<https://devfeed.tech/tags/robotics.md>), [search](<https://devfeed.tech/tags/search.md>)

### AI overview

The proposed LG-HRL framework coordinates UAV and UGV teams in simulated search and rescue by having an LLM turn human instructions into semantic cues for strategy selection and role assignment, alongside hierarchical policies and a shared cognitive workspace. A formal LG-MAHDP models how language grounding, hierarchical control, and distributed coordination interact. In disaster simulations, the authors report 2.3x faster learning, 37% greater coordination coherence, 46% higher subgoal accuracy, 31% better hazard avoidance, and 20-30% higher victim rescue rates. The system showed emergent role specialization and adapted to verbal updates; ablations indicate that both language grounding and the shared workspace contribute to performance. Results are from simulation.

### Source excerpt

arXiv:2611.13161v1 Announce Type: new Abstract: Effective search and rescue (SAR) operations require autonomous systems capable of coordinating multiple heterogeneous agents, adapting to dynamic hazards, and aligning with human mission intent. Existing language-guided reinforcement learning (RL) methods primarily focus on single-agent instruction following and lack mechanisms for team-level coordination in complex, partially observable environments. This paper presents a Cognitive Multi-Agent Framework based on Language-Grounded Hierarchical Reinforcement Learning (LG-HRL) that integrates natural-language understanding, hierarchical policy decomposition, and a shared cognitive workspace to achieve cohesive multi-agent behavior in SAR scenarios. Human instructions are interpreted by a large language model (LLM) and transformed into structured semantic cues that guide high-level strategy selection and role assignment across UAV and UGV agents. A formal Language-Grounded Multi-Agent Hierarchical Decision Process (LG-MAHDP) is introduced to model the interplay between semantic grounding, hierarchical control, and distributed coordination. Comprehensive experiments in a simulated disaster environment show that LG-HRL improves learning speed by 2.3 times, increases coordination coherence by 37%, and enhances subgoal accuracy and hazard avoidance by 46% and 31%, respectively. The system achieves 20-30% higher victim rescue rates, demonstrates emergent role specialization, and adapts immediately to human verbal updates. Ablation studies confirm that both the shared cognitive workspace and language grounding are critical to overall system performance. Overall, this work establishes a new paradigm for language-guided multi-agent SAR robotics, demonstrating that semantic grounding combined with hierarchical reasoning enables safer, more efficient, and more interpretable autonomous rescue operations.

## LandslideFM: A Foundation Model Towards Unified Landslide Perception and Understanding

DevFeed: [LandslideFM: A Foundation Model Towards Unified Landslide Perception and Understanding](<https://devfeed.tech/articles/landslidefm-a-foundation-model-towards-unified-landslide-perception-and-understanding-128205.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.12936>)

Author: Shanwen Wang, Sirui Wang, Xin Sun, Jungong Han

Published: 2026-11-11T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Artificial Intelligence](<https://devfeed.tech/sources/arxiv-artificial-intelligence.md>)

Topics: [AI Foundation Models](<https://devfeed.tech/topics/ai-foundation-models.md>), [Artificial Intelligence](<https://devfeed.tech/topics/ai.md>), [vlm](<https://devfeed.tech/topics/vlm.md>), [Fine-tuning](<https://devfeed.tech/topics/fine-tuning.md>), [Machine Intelligence](<https://devfeed.tech/topics/machine-intelligence.md>), [Transformers](<https://devfeed.tech/topics/transformers.md>), [Robot Manipulation](<https://devfeed.tech/topics/robot-manipulation.md>)

Tags: [benchmark](<https://devfeed.tech/tags/benchmark.md>), [cs-ai](<https://devfeed.tech/tags/cs-ai.md>), [detection](<https://devfeed.tech/tags/detection.md>), [disaster](<https://devfeed.tech/tags/disaster.md>), [experiments](<https://devfeed.tech/tags/experiments.md>), [fine-tuning](<https://devfeed.tech/tags/fine-tuning.md>), [foundation-model](<https://devfeed.tech/tags/foundation-model.md>), [framework](<https://devfeed.tech/tags/framework.md>), [landslide](<https://devfeed.tech/tags/landslide.md>), [performance](<https://devfeed.tech/tags/performance.md>), [reasoning](<https://devfeed.tech/tags/reasoning.md>), [spatial](<https://devfeed.tech/tags/spatial.md>), [training](<https://devfeed.tech/tags/training.md>), [vision-language-model](<https://devfeed.tech/tags/vision-language-model.md>)

### AI overview

LandslideFM combines landslide visual perception with contextual scene understanding. Its Source-Agreement Graduated Trust semi-supervised framework uses limited labeled data alongside abundant unlabeled data for multiple visual tasks. Predicted landslide regions then guide a parameter-efficiently fine-tuned vision-language model in landslide reasoning. The accompanying LandslideVL benchmark pairs imagery with textual prompts for training and evaluation. The abstract reports that LandslideFM outperforms general-purpose vision-language models on specialized landslide reasoning and scene analysis, but provides no specific metrics or experimental details.

### Source excerpt

arXiv:2611.12936v1 Announce Type: new Abstract: Intelligent landslide detection is essential for disaster prevention. However, existing models primarily focus on isolated visual tasks, require extensive labeled data, and lack deep analytical and scene-understanding capabilities. To address these limitations, we propose LandslideFM, the first unified foundation model that connects visual perception tasks with contextual interpretation of landslides. Specifically, LandslideFM adapts to multiple visual tasks via a Source-Agreement Graduated Trust semi-supervised learning framework that leverages limited labeled data alongside abundant unlabeled data. Furthermore, predicted landslide regions provide explicit spatial guidance to a Vision-Language Model adapted through parameter-efficient fine-tuning, enabling comprehensive landslide reasoning. To support training and evaluation in this domain, we also present LandslideVL, the first benchmark for intelligent landslide analysis, combining visual imagery and newly introduced textual prompts. Comprehensive experiments across public datasets and LandslideVL demonstrate that LandslideFM achieves superior quantitative and qualitative performance. Most importantly, LandslideFM surpasses general-purpose VLMs in specialized landslide reasoning and scene analysis. Our code is released at https://github.com/wangshanwen001/LandslideFM.

## Ceci n'est pas une pipe -- Toward a Symbolic Solver for the Abstraction and Reasoning Corpus Using Evolutionary Computation

DevFeed: [Ceci n'est pas une pipe -- Toward a Symbolic Solver for the Abstraction and Reasoning Corpus Using Evolutionary Computation](<https://devfeed.tech/articles/ceci-n-est-pas-une-pipe-toward-a-symbolic-solver-for-the-abstraction-and-reasoning-corpus-using-evolutionary-computation-128360.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.13831>)

Author: Luca Favole, Giorgio Mongardi, Giovanni Squillero, Alberto Tonda

Published: 2026-11-11T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Artificial Intelligence](<https://devfeed.tech/sources/arxiv-artificial-intelligence.md>)

Topics: [Artificial Intelligence](<https://devfeed.tech/topics/ai.md>), [Automated reasoning](<https://devfeed.tech/topics/automated-reasoning.md>), [Machine Intelligence](<https://devfeed.tech/topics/machine-intelligence.md>), [Combinatorial optimization](<https://devfeed.tech/topics/combinatorial-optimization.md>)

Tags: [ai](<https://devfeed.tech/tags/ai.md>), [arc](<https://devfeed.tech/tags/arc.md>), [benchmark](<https://devfeed.tech/tags/benchmark.md>), [cognitive-psychology](<https://devfeed.tech/tags/cognitive-psychology.md>), [cs-ai](<https://devfeed.tech/tags/cs-ai.md>), [domain-specific-language](<https://devfeed.tech/tags/domain-specific-language.md>), [fitness](<https://devfeed.tech/tags/fitness.md>), [grid](<https://devfeed.tech/tags/grid.md>), [object](<https://devfeed.tech/tags/object.md>), [pixel](<https://devfeed.tech/tags/pixel.md>), [synthesis](<https://devfeed.tech/tags/synthesis.md>), [transformer](<https://devfeed.tech/tags/transformer.md>)

### AI overview

The paper presents a fully symbolic evolutionary program synthesis approach submitted to the ARC-AGI-2 main contest. To address evolutionary search's difficulty choosing a domain-specific representation, it evolves the grid parser and output projector alongside the transforming program, allowing a task-specific language to emerge. It grounds parser symbols in core knowledge theory from cognitive psychology and uses object-aware fitness to score partial solutions beyond pixel-level differences. The excerpt identifies these as design contributions but provides no performance results.

### Source excerpt

arXiv:2611.13831v1 Announce Type: new Abstract: The Abstraction and Reasoning Corpus (ARC) is a benchmark designed to assess the capability of Artificial Intelligence (AI) systems to generalize starting from a few samples, a task humans excel at. The ARC-AGI contests held since 2020 have been the focus of intense competition, with participants proposing approaches ranging from traditional symbolic AI to Transformer architectures. Despite a long history of successes, evolutionary algorithms (EAs) were notably absent from the approaches appearing in the main contest. As score-and-search program synthesis approaches, EAs suffer from issues related to choosing the proper representation in the Domain-Specific Language applied to the problem (for example, pixel-level or object-level), and distinguishing between two candidate solutions producing the same outputs. In this work, we present the first fully symbolic, evolutionary-based program synthesis approach submitted to the ARC-AGI-2 main contest. The main contributions are: evolving the grid parser and the output projector together with the program that transforms the grid, so that a task-specific language emerges during program synthesis; grounding the symbols the parser can extract in core knowledge theory from cognitive psychology; and an object-aware fitness that grades partially correct outputs beyond pixel-level discrepancies.

## A computational framework for understanding embodied plans

DevFeed: [A computational framework for understanding embodied plans](<https://devfeed.tech/articles/a-computational-framework-for-understanding-embodied-plans-128340.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.13669>)

Author: Aalap D. Shah, Basil Saeed, Joshua B. Tenenbaum, Tobias Gerstenberg, Hyowon Gweon, Ilker Yildirim

Published: 2026-11-11T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Artificial Intelligence](<https://devfeed.tech/sources/arxiv-artificial-intelligence.md>)

Topics: [Artificial Intelligence](<https://devfeed.tech/topics/ai.md>), [Machine Intelligence](<https://devfeed.tech/topics/machine-intelligence.md>), [Physical AI](<https://devfeed.tech/topics/physical-ai.md>)

Tags: [actions](<https://devfeed.tech/tags/actions.md>), [agent](<https://devfeed.tech/tags/agent.md>), [architecture](<https://devfeed.tech/tags/architecture.md>), [artificial-intelligence](<https://devfeed.tech/tags/artificial-intelligence.md>), [cs-ai](<https://devfeed.tech/tags/cs-ai.md>), [cs-hc](<https://devfeed.tech/tags/cs-hc.md>), [cs-ro](<https://devfeed.tech/tags/cs-ro.md>), [distribution](<https://devfeed.tech/tags/distribution.md>), [embedding](<https://devfeed.tech/tags/embedding.md>), [framework](<https://devfeed.tech/tags/framework.md>), [generative](<https://devfeed.tech/tags/generative.md>), [goals](<https://devfeed.tech/tags/goals.md>), [horizon](<https://devfeed.tech/tags/horizon.md>), [humans](<https://devfeed.tech/tags/humans.md>), [interaction](<https://devfeed.tech/tags/interaction.md>), [joint-probability](<https://devfeed.tech/tags/joint-probability.md>), [language-model](<https://devfeed.tech/tags/language-model.md>), [manipulation](<https://devfeed.tech/tags/manipulation.md>), [model](<https://devfeed.tech/tags/model.md>), [models](<https://devfeed.tech/tags/models.md>), [motion-planning](<https://devfeed.tech/tags/motion-planning.md>), [navigation](<https://devfeed.tech/tags/navigation.md>), [object](<https://devfeed.tech/tags/object.md>)

### AI overview

GenEA models multi-step, long-horizon embodied plans by combining a multilevel state-space for articulated-body biomechanics and object interactions with simulation-based intuitive physics, extending robotics task-and-motion planning. Embedded in a generative model, this representation produces a joint probability distribution over goals and plans. In experiments on object manipulation and navigation, transformations of that distribution reportedly reproduced aspects of human reasoning and consistently outperformed tested heuristics, standard task-and-motion planning, and a state-of-the-art vision-language model. The evidence provided is an abstract, so it does not specify experimental details or the size of the performance gains.

### Source excerpt

arXiv:2611.13669v1 Announce Type: new Abstract: Reasoning about people's plans and goals based on their actions is fundamental to human social interaction and essential for cooperative, embodied artificial intelligence. Formalizing this capacity in naturalistic, embodied settings has been a longstanding challenge. Here, we present GenEA, a novel agent representation capable of reasoning over multi-step, long-horizon embodied plans. At the heart of GenEA lies a multilevel state-space that couples the coarse biomechanics of an articulated body and its interactions with nearby objects by extending the task-and-motion planning framework from robotics with simulation-based intuitive physics. By embedding this state-space in a generative model, GenEA yields a joint probability distribution over goals and embodied plans. Across several experiments spanning the qualitatively distinct domains of object manipulation and navigation, featuring diverse tasks and multiple behavioral measures, we show that GenEA, through simple transformations of this joint distribution, recapitulates how humans reason about embodied plans. Crucially, GenEA consistently outperforms strong alternative models, including plausible heuristics, standard task-and-motion planning, and a state-of-the-art vision-language model. GenEA thereby offers a new perspective on the architecture of human reasoning and social cognition, suggesting that embodied artificial intelligence systems seeking to seamlessly interact with humans should adopt GenEA-like generative models of action planning.

## BODYinTRANSIT-Lab: A Multimodal Dataset for Investigating Sensory-Driven Transformations in Body Perception and Affective Experiences

DevFeed: [BODYinTRANSIT-Lab: A Multimodal Dataset for Investigating Sensory-Driven Transformations in Body Perception and Affective Experiences](<https://devfeed.tech/articles/bodyintransit-lab-a-multimodal-dataset-for-investigating-sensory-driven-transformations-in-body-perception-and-affective-experiences-128798.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.14093>)

Author: Mohammad Mahdi Dehshibi, Tom\\'{a}s Mart\\'{i}nez Cort\\'{e}s, Amar D'Adamo, Daniel de la Prida, Luis-Antonio Azpicueta-Ru\\'{i}z, Laia Turmo Vidal, Marte Roel Lesur, Joaqu\\'{i}n R. D\\'{i}az-Dur\\'{a}n, Al

Published: 2026-11-11T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Machine Learning](<https://devfeed.tech/sources/arxiv-machine-learning.md>)

Topics: [Artificial Intelligence](<https://devfeed.tech/topics/ai.md>), [Machine Intelligence](<https://devfeed.tech/topics/machine-intelligence.md>), [User Interfaces](<https://devfeed.tech/topics/user-interfaces.md>)

Tags: [3d](<https://devfeed.tech/tags/3d.md>), [benchmark](<https://devfeed.tech/tags/benchmark.md>), [computing](<https://devfeed.tech/tags/computing.md>), [creative](<https://devfeed.tech/tags/creative.md>), [cs-hc](<https://devfeed.tech/tags/cs-hc.md>), [cs-lg](<https://devfeed.tech/tags/cs-lg.md>), [dataset](<https://devfeed.tech/tags/dataset.md>), [deep-learning](<https://devfeed.tech/tags/deep-learning.md>), [health](<https://devfeed.tech/tags/health.md>), [interactive](<https://devfeed.tech/tags/interactive.md>), [kinematics](<https://devfeed.tech/tags/kinematics.md>), [multimodal](<https://devfeed.tech/tags/multimodal.md>), [neuroscience](<https://devfeed.tech/tags/neuroscience.md>), [real-time](<https://devfeed.tech/tags/real-time.md>), [recordings](<https://devfeed.tech/tags/recordings.md>), [resource](<https://devfeed.tech/tags/resource.md>), [sensor](<https://devfeed.tech/tags/sensor.md>), [social](<https://devfeed.tech/tags/social.md>), [weight](<https://devfeed.tech/tags/weight.md>)

### AI overview

BODYinTRANSIT-Lab is a multimodal dataset of 101 participants' responses to the Footsteps Illusion, in which manipulated real-time audio of participants' footsteps during walking is intended to alter perceived body weight. It synchronizes full-body movement with muscle, cardiac, breathing, and electrodermal signals, alongside immediate perceived-weight ratings and post-trial reports, avatar visualizations, and body maps. Participants were stratified by eating-disorder symptoms and physical activity, with additional psychological, behavioral, demographic, and social measures. A within-subject deep-learning benchmark uses walking kinematics to classify relative affective valence and perceived weight; on within-session segment-level splits, mean balanced accuracy was 45.9% and 44.5%, respectively. The dataset is presented as a resource for embodied-experience research and computational modeling.

### Source excerpt

arXiv:2611.14093v1 Announce Type: cross Abstract: Multimodal datasets that combine synchronised behavioural and physiological signals with subjective affective and body perception measures are still rare. This limits both the understanding of embodied experience and the development of computational models that infer subjective states from sensor data. We introduce BODYinTRANSIT-Lab, a multimodal dataset that addresses this gap by capturing affective, perceptual, and behavioural responses to the Footsteps Illusion. In this closed-loop paradigm, participants perform walking tasks while receiving real-time manipulated auditory feedback of their own footsteps intended to alter perceived body weight. The dataset comprises synchronised recordings from 101 participants recruited through systematic stratification by eating-disorder symptomatology and physical activity level. BODYinTRANSIT-Lab combines full-body kinematics with electromyography, electrocardiography, respiration, and electrodermal activity. Its annotations capture both immediate and post-trial experience, including within-trial perceived-weight ratings, post-trial assessments of body feelings and affective experience, 3D avatar body visualisations, and quantitative body maps. Psychological, behavioural, demographic, and social measures provide further information about factors that may influence individual responses. We provide a deep learning benchmark that uses walking kinematics to classify relative affective valence and perceived weight within subjects. On within-session, segment-level splits, the baseline model achieved mean balanced accuracies of 45.9% for affective valence and 44.5% for perceived weight. The BODYinTRANSIT-Lab dataset provides an experimentally controlled resource for cognitive neuroscience and affective computing research, with potential applications in affect-aware interactive systems, health and well-being technologies, and creative embodied experiences.

## ZoomGain: Mining Global-to-Zoom Perception Gaps for Fine-Grained Visual Supervision

DevFeed: [ZoomGain: Mining Global-to-Zoom Perception Gaps for Fine-Grained Visual Supervision](<https://devfeed.tech/articles/zoomgain-mining-global-to-zoom-perception-gaps-for-fine-grained-visual-supervision-128394.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.14076>)

Author: Haozhan Shen, Liangbo He, Jun Lan, Huijia Zhu, Weiqiang Wang, Tiancheng Zhao, Jianwei Yin

Published: 2026-11-11T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Artificial Intelligence](<https://devfeed.tech/sources/arxiv-artificial-intelligence.md>)

Topics: [Machine Intelligence](<https://devfeed.tech/topics/machine-intelligence.md>), [Fine-tuning](<https://devfeed.tech/topics/fine-tuning.md>), [image tagging](<https://devfeed.tech/topics/image-tagging.md>)

Tags: [benchmark](<https://devfeed.tech/tags/benchmark.md>), [cs-ai](<https://devfeed.tech/tags/cs-ai.md>), [cs-cv](<https://devfeed.tech/tags/cs-cv.md>), [evaluation](<https://devfeed.tech/tags/evaluation.md>), [fine](<https://devfeed.tech/tags/fine.md>), [pipeline](<https://devfeed.tech/tags/pipeline.md>), [supervision](<https://devfeed.tech/tags/supervision.md>), [training](<https://devfeed.tech/tags/training.md>), [vision](<https://devfeed.tech/tags/vision.md>), [zoom](<https://devfeed.tech/tags/zoom.md>)

### AI overview

ZoomGain uses a global-to-zoom perception gap to create fine-grained visual supervision: questions about a full image are derived from details recognized in a zoomed region. Multiple models verify answers on the zoomed view, while additional checks target ambiguity in the full image and answers inferable from language priors; the pipeline also generates distractors for multiple-choice questions. With human screening but no manual QA editing, the authors build a bilingual benchmark of 1,000 questions, where the best of 28 tested VLMs scores 61.08. They report that training on 9K samples improves their 4B and 9B models across perception benchmarks; the 9B model scores 90.25 on HR-Bench 4K and is reported to outperform several much larger models on various benchmarks. The abstract also notes a recurring distractor preference in incorrect multiple-choice answers.

### Source excerpt

arXiv:2611.14076v1 Announce Type: cross Abstract: Fine-grained visual perception requires Vision-Language Models (VLMs) to locate and interpret subtle evidence within complex scenes. Constructing supervision for this ability is challenging: questions must have correct, unambiguous answers that depend on actual fine-grained visual details rather than coarse cues or language priors. Verifying these properties makes manual annotation labor-intensive and automatic generation difficult for models that themselves struggle with fine-grained perception. Yet the same models can often recognize such details once the region is zoomed in. We exploit this global-to-zoom perception gap and introduce ZoomGain, a data construction pipeline that constructs questions about the full image from visual information gained only by zooming in. It verifies each answer on the zoomed view with multiple models, ensures that each question remains unambiguous in the full image and cannot be answered from language priors, and specifically generates high-quality distractors for multiple-choice questions. With this pipeline and human-only screening (no manual QA editing), we build ZoomGainBench, a 1,000-question bilingual benchmark on which the best of 28 VLMs achieves only 61.08. Failure analysis reveals that models favor a particular type of distractor when answering multiple-choice questions incorrectly. Beyond evaluation, training on just 9K ZoomGain samples substantially improves the visual perception capabilities of our 4B/9B models, enabling them to outperform several hundred-billion-parameter models across various perception benchmarks. The trained 9B model achieves SoTA performance on several benchmarks, such as 90.25 on HR-Bench 4K. These results establish global-to-zoom perception gaps as a practical basis for scaling fine-grained visual supervision for both evaluation and training. Our code is available at https://github.com/inclusionAI/ZoomGain.

## Evaluating LLM-Generated World Models for Neuro-Symbolic Reasoning

DevFeed: [Evaluating LLM-Generated World Models for Neuro-Symbolic Reasoning](<https://devfeed.tech/articles/evaluating-llm-generated-world-models-for-neuro-symbolic-reasoning-128431.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.14419>)

Author: Souhaila Djaffal, Amna Jamal, Gagana Madhugiri Devaraja, Anidipta Pal, Jobst Heitzig

Published: 2026-11-11T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Artificial Intelligence](<https://devfeed.tech/sources/arxiv-artificial-intelligence.md>)

Topics: [llm-reasoning](<https://devfeed.tech/topics/llm-reasoning.md>), [Large Language Model](<https://devfeed.tech/topics/llm.md>), [Artificial Intelligence](<https://devfeed.tech/topics/ai.md>), [Automated reasoning](<https://devfeed.tech/topics/automated-reasoning.md>), [Machine Intelligence](<https://devfeed.tech/topics/machine-intelligence.md>)

Tags: [benchmark](<https://devfeed.tech/tags/benchmark.md>), [cs-ai](<https://devfeed.tech/tags/cs-ai.md>), [llm](<https://devfeed.tech/tags/llm.md>), [multi-agent](<https://devfeed.tech/tags/multi-agent.md>), [planning](<https://devfeed.tech/tags/planning.md>), [structure](<https://devfeed.tech/tags/structure.md>), [symbolic-reasoning](<https://devfeed.tech/tags/symbolic-reasoning.md>), [verification](<https://devfeed.tech/tags/verification.md>), [world-models](<https://devfeed.tech/tags/world-models.md>)

### AI overview

Across a benchmark of 44 game, situational, and cultural scenarios, five LLMs produced 410 first-pass world models represented as partially observable stochastic games. The evaluation measured mechanical validity, agreement with authored structure, and semantic criteria. In the reference-backed game track, 134 of 135 models passed a strict mechanical audit, but mean permutation-invariant structural F1 was 0.366, and only 28 of 134 mechanically valid models exactly matched the reference. In the cultural track, naming the country changed most models' structural representations while shifting automated semantic scores only slightly. The findings indicate that executability alone does not establish faithful representation for downstream reasoning and planning.

### Source excerpt

arXiv:2611.14419v1 Announce Type: new Abstract: Formal world models can support planning, verification, and safety reasoning, but constructing them from open-ended natural-language situations is difficult, particularly in multi-agent settings where several formalizations may be plausible. We study whether five LLMs can generate such world models as partially observable stochastic games and whether mechanical validity is sufficient to establish that the resulting model faithfully represents the intended decision problem. We introduce a benchmark of 44 scenarios spanning game, situational, and cultural settings and evaluate generated world models along three dimensions: mechanical validity, reference-relative structural agreement, and semantic criteria. Across three empirical tracks, we analyze 410 first-pass generations. In the reference-backed game track, 134/135 generated world models pass a strict mechanical audit, yet mean permutation-invariant structural F1 is only 0.366, with just 28/134 mechanically valid models exactly matching the authored reference. In the cultural track, explicitly naming the country changes the generated structural representation for most LLMs while producing only small changes in automated semantic scores. These results show that executability is a necessary but insufficient criterion for LLM-generated world models: models that are mechanically valid can still differ substantially in the structure used by downstream neuro-symbolic reasoning and planning.

## Perception, Not Reasoning, Limits Visual Theory of Mind

DevFeed: [Perception, Not Reasoning, Limits Visual Theory of Mind](<https://devfeed.tech/articles/perception-not-reasoning-limits-visual-theory-of-mind-125994.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.10694>)

Author: Mohamed Rayan Barhdadi, Syed Talal Wasim, Juergen Gall, Erchin Serpedin, Hasan Kurban

Published: 2026-11-10T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Artificial Intelligence](<https://devfeed.tech/sources/arxiv-artificial-intelligence.md>)

Topics: [llm-reasoning](<https://devfeed.tech/topics/llm-reasoning.md>), [Artificial Intelligence](<https://devfeed.tech/topics/ai.md>), [Machine Intelligence](<https://devfeed.tech/topics/machine-intelligence.md>)

Tags: [2d](<https://devfeed.tech/tags/2d.md>), [3d](<https://devfeed.tech/tags/3d.md>), [agent](<https://devfeed.tech/tags/agent.md>), [agents](<https://devfeed.tech/tags/agents.md>), [api](<https://devfeed.tech/tags/api.md>), [bias](<https://devfeed.tech/tags/bias.md>), [computer-science](<https://devfeed.tech/tags/computer-science.md>), [computer-vision](<https://devfeed.tech/tags/computer-vision.md>), [cs-ai](<https://devfeed.tech/tags/cs-ai.md>), [cs-cl](<https://devfeed.tech/tags/cs-cl.md>), [cs-cv](<https://devfeed.tech/tags/cs-cv.md>), [evaluation](<https://devfeed.tech/tags/evaluation.md>), [inference](<https://devfeed.tech/tags/inference.md>), [limits](<https://devfeed.tech/tags/limits.md>), [mind](<https://devfeed.tech/tags/mind.md>), [protocol](<https://devfeed.tech/tags/protocol.md>), [q-bio-nc](<https://devfeed.tech/tags/q-bio-nc.md>), [ratio](<https://devfeed.tech/tags/ratio.md>), [reasoning](<https://devfeed.tech/tags/reasoning.md>), [structured](<https://devfeed.tech/tags/structured.md>), [theory](<https://devfeed.tech/tags/theory.md>), [vision-language-models](<https://devfeed.tech/tags/vision-language-models.md>), [vlms](<https://devfeed.tech/tags/vlms.md>)

### AI overview

Across seven vision-language models, visual theory-of-mind performance appears limited mainly by scene perception: models often fail to track which agents witnessed events, producing "reality bias" in 60% of false-belief failures. In a five-condition protocol spanning abstract 2D panels and naturalistic 3D game scenes, ground-truth scene-state scaffolds substantially narrowed the visual-to-text performance gap, matching or exceeding text-only reasoning on the panels and MuMA-ToM. Automatically generated scaffolds, including those from frontier VLMs, failed to do so. The findings support structured visual scene-state extraction as the dominant bottleneck under this evaluation, rather than belief inference itself.

### Source excerpt

arXiv:2611.10694v1 Announce Type: cross Abstract: Vision-language models pass text-based false-belief probes at near-human levels but drop by over twenty percentage points on abstract panels the moment the same problem is conveyed visually. The consensus diagnosis is conceptual, prescribing new belief-reasoning modules. We argue the dominant bottleneck is perceptual: not what the model cannot reason about, but what it cannot see. We propose a five-condition dissociation protocol that holds the reasoning task constant while varying perceptual fidelity from visual only to text only, and quantify the result with the Perceptual Contribution Ratio, the share of the visual-to-text gap a given fidelity closes. We test this on two kinds of scenes: abstract 2D panels in the Heider--Simmel tradition, and naturalistic 3D game scenes from MuMA-ToM and MindPower. The evaluation covers seven VLMs and roughly 159,000 API calls. Ground-truth scaffolds substantially close the gap in every setting, and on the abstract panels and MuMA-ToM they match or exceed the text-only reasoning. Whereas scaffolds built by automatic parsers, including frontier VLMs, fail. The most common error in the visual-only condition is \emph{reality bias}, which accounts for 60\% of false-belief failures. Models report where an object actually is, not where an absent agent believes it is. This points to a perceptual front-end that fails to track which agents were present at each moment. Under the evaluated protocol, structured visual scene-state extraction, rather than belief inference, is the dominant bottleneck for visual ToM.

## The Lagrangian Brain: A Constraint Mechanics of Mind and Machine

DevFeed: [The Lagrangian Brain: A Constraint Mechanics of Mind and Machine](<https://devfeed.tech/articles/the-lagrangian-brain-a-constraint-mechanics-of-mind-and-machine-126001.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.10737>)

Author: Yuhang He, Si Wu

Published: 2026-11-10T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Artificial Intelligence](<https://devfeed.tech/sources/arxiv-artificial-intelligence.md>)

Topics: [Artificial Intelligence](<https://devfeed.tech/topics/ai.md>), [Machine learning](<https://devfeed.tech/topics/machine-learning.md>), [Neural Network](<https://devfeed.tech/topics/neural-network.md>), [Machine Intelligence](<https://devfeed.tech/topics/machine-intelligence.md>), [Transformers](<https://devfeed.tech/topics/transformers.md>)

Tags: [abstract](<https://devfeed.tech/tags/abstract.md>), [all](<https://devfeed.tech/tags/all.md>), [area](<https://devfeed.tech/tags/area.md>), [arxiv](<https://devfeed.tech/tags/arxiv.md>), [balance](<https://devfeed.tech/tags/balance.md>), [bead](<https://devfeed.tech/tags/bead.md>), [brain](<https://devfeed.tech/tags/brain.md>), [circuit](<https://devfeed.tech/tags/circuit.md>), [constraint](<https://devfeed.tech/tags/constraint.md>), [cortex](<https://devfeed.tech/tags/cortex.md>), [cost](<https://devfeed.tech/tags/cost.md>), [cs-ai](<https://devfeed.tech/tags/cs-ai.md>), [experience](<https://devfeed.tech/tags/experience.md>), [fail](<https://devfeed.tech/tags/fail.md>), [feedback](<https://devfeed.tech/tags/feedback.md>), [follow](<https://devfeed.tech/tags/follow.md>), [future](<https://devfeed.tech/tags/future.md>), [hypothesis](<https://devfeed.tech/tags/hypothesis.md>), [lagrange](<https://devfeed.tech/tags/lagrange.md>), [machine](<https://devfeed.tech/tags/machine.md>), [mind](<https://devfeed.tech/tags/mind.md>)

### AI overview

The paper proposes that brains and machines can be understood as single evolving states governed by constraints, with separate multipliers that accumulate constraint violations and act over different timescales. It links inhibitory interneurons, adaptation, synapses and apical feedback to different kinds of constraints, and says this framework accounts for several neural properties and recasts belief, learning and planning as dynamics over trajectories. A machine using these principles is reported to train without gradients on 1,000 solved puzzles, solve all 422,786 Sudoku-Extreme test puzzles in one trajectory each, rediscover Sudoku rules through sleep-and-wake cycles, find shortest paths in mazes 64 times larger than its training area, and track permutations on strings 25 times longer than training, where the abstract says Transformers fail.

### Source excerpt

arXiv:2611.10737v1 Announce Type: new Abstract: A brain is in one state at a time, yet most of what decides its next state is absent from it: the patterns its world never produces, the statistics of a lifetime, a goal that lies ahead. In mechanics a wire acts on a bead only through a force that appears when the bead presses on it, and Lagrange computed such forces with one multiplier for each constraint. We propose that a brain follows the same mechanics. Everything a circuit knows, is told or wants is a constraint on its single state and acts through its own multiplier, which accumulates the violation and pushes the state back. Neurons and synapses carry the multipliers, each with its own time constant: inhibitory interneurons carry what never happens, adaptation the cost of staying in one state, synapses the statistics of experience, and feedback to apical dendrites the future. From this hypothesis follow Dale's law, the reciprocal wiring and balance of excitation and inhibition, the paradoxical responses of cortex and the frequencies of brain rhythms. Belief becomes the time one trajectory spends in each state, learning becomes sleep proposing laws and waking refuting the false ones, and planning becomes a wave from the goal meeting a wave from the present. Connectionism and symbolism are the limits of frozen and of instantaneous multipliers, and between them the time a circuit runs becomes the depth of its computation. A machine built on these principles, trained without gradients on 1000 solved puzzles, solves all 422,786 Sudoku-Extreme test puzzles with one trajectory each. It rediscovers the rules of Sudoku by sleeping and waking, plans shortest paths in mazes of 64 times the training area, and tracks permutations exactly on strings 25 times longer than in training, where Transformers fail. A mind, on this view, is life that learns its own constraints.

## Look Ahead: Search-Driven Reasoning in Embodied AI

DevFeed: [Look Ahead: Search-Driven Reasoning in Embodied AI](<https://devfeed.tech/articles/look-ahead-search-driven-reasoning-in-embodied-ai-126009.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.10769>)

Author: Ahmed Zeer, Osama Shbib, Muzaffer Kaan Yuce, Himmet Toprak Kesgin, Mehmet Fatih Amasyali

Published: 2026-11-10T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Artificial Intelligence](<https://devfeed.tech/sources/arxiv-artificial-intelligence.md>)

Topics: [Artificial Intelligence](<https://devfeed.tech/topics/ai.md>), [Physical AI](<https://devfeed.tech/topics/physical-ai.md>), [Robotics Simulation](<https://devfeed.tech/topics/robotics-simulation.md>), [Machine Intelligence](<https://devfeed.tech/topics/machine-intelligence.md>)

Tags: [act](<https://devfeed.tech/tags/act.md>), [ai](<https://devfeed.tech/tags/ai.md>), [beam](<https://devfeed.tech/tags/beam.md>), [coverage](<https://devfeed.tech/tags/coverage.md>), [cs-ai](<https://devfeed.tech/tags/cs-ai.md>), [embodied-ai](<https://devfeed.tech/tags/embodied-ai.md>), [evaluation](<https://devfeed.tech/tags/evaluation.md>), [performance](<https://devfeed.tech/tags/performance.md>), [planning](<https://devfeed.tech/tags/planning.md>), [policy](<https://devfeed.tech/tags/policy.md>), [reasoning](<https://devfeed.tech/tags/reasoning.md>), [reward](<https://devfeed.tech/tags/reward.md>), [runtime](<https://devfeed.tech/tags/runtime.md>), [search](<https://devfeed.tech/tags/search.md>), [simulator](<https://devfeed.tech/tags/simulator.md>), [test](<https://devfeed.tech/tags/test.md>), [transformers](<https://devfeed.tech/tags/transformers.md>)

### AI overview

In PushT simulation with a frozen ACT policy, beam search over perturbed action chunks improved success across 50 episodes from 0% to 18% and average cumulative reward from 39.44 to 103.2. The proposed One Step Ahead gate simulates the nominal chunk and triggers full search when predicted coverage drops beyond a threshold; reported ablations show it preserves or improves success while cutting runtime by about 3x versus full beam search. Larger search budgets did not consistently improve performance. The evidence is limited to this setup; language-conditioned VLAs and physical robots remain unevaluated.

### Source excerpt

arXiv:2611.10769v1 Announce Type: new Abstract: Search allows a robotic policy to evaluate candidate futures before committing to an action. Motivated by test-time planning for vision-language action (VLA) models, we study selective search over action chunks in PushT using a frozen Action Chunking with Transformers (ACT) policy. Beam search generates perturbed alternatives around the policy prediction and scores them through rollouts from cloned simulator states. Across 50 evaluation episodes, beam search raises task success from 0% to 18% and increases average cumulative reward from 39.44 to 103.2. We introduce One Step Ahead (OSA), a gate that simulates the nominal policy chunk and invokes full search only when its predicted coverage drop exceeds a threshold. In the reported ablations, OSA preserves or improves success while reducing runtime by approximately 3x relative to full beam search. Ablations over tree depth, branching factor, beam width, and replanning frequency show that larger search budgets do not consistently improve performance. The results support selective test-time search for action-chunking policies; evaluation on language-conditioned VLAs and physical robots remains future work.

## CID: Measuring Cognitive Work Through Semantic State Change

DevFeed: [CID: Measuring Cognitive Work Through Semantic State Change](<https://devfeed.tech/articles/cid-measuring-cognitive-work-through-semantic-state-change-124699.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.09542>)

Author: Ron Fulbright

Published: 2026-11-09T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Artificial Intelligence](<https://devfeed.tech/sources/arxiv-artificial-intelligence.md>)

Topics: [AI Research](<https://devfeed.tech/topics/ai-research.md>), [Embeddings](<https://devfeed.tech/topics/embeddings.md>), [Artificial Intelligence](<https://devfeed.tech/topics/ai.md>), [Machine Intelligence](<https://devfeed.tech/topics/machine-intelligence.md>)

Tags: [ai](<https://devfeed.tech/tags/ai.md>), [ai-collaboration](<https://devfeed.tech/tags/ai-collaboration.md>), [change](<https://devfeed.tech/tags/change.md>), [covariance](<https://devfeed.tech/tags/covariance.md>), [cs-ai](<https://devfeed.tech/tags/cs-ai.md>), [embedding](<https://devfeed.tech/tags/embedding.md>), [foundation](<https://devfeed.tech/tags/foundation.md>), [framework](<https://devfeed.tech/tags/framework.md>), [generation](<https://devfeed.tech/tags/generation.md>), [human-ai-collaboration](<https://devfeed.tech/tags/human-ai-collaboration.md>), [implementation](<https://devfeed.tech/tags/implementation.md>), [information](<https://devfeed.tech/tags/information.md>), [information-theory](<https://devfeed.tech/tags/information-theory.md>), [knowledge](<https://devfeed.tech/tags/knowledge.md>), [language-models](<https://devfeed.tech/tags/language-models.md>), [large-language-models](<https://devfeed.tech/tags/large-language-models.md>), [learning](<https://devfeed.tech/tags/learning.md>), [measuring](<https://devfeed.tech/tags/measuring.md>), [semantic](<https://devfeed.tech/tags/semantic.md>), [state](<https://devfeed.tech/tags/state.md>), [work](<https://devfeed.tech/tags/work.md>)

### AI overview

The paper proposes Contextual Information Differential (CID) to quantify semantic change as movement between vector representations in a model's semantic space. CID uses Mahalanobis distance to measure displacement while accounting for covariance in the embedding structure. The authors present it as a way to study cognitive work and information processing, including learning, reasoning, knowledge generation, and human-AI collaboration. The supplied text describes a proposed framework and its intended applications; it does not provide evidence of validation or results.

### Source excerpt

arXiv:2611.09542v1 Announce Type: new Abstract: Traditional information theory provides powerful tools for analyzing statistical properties of information but offers limited insight into meaning and semantic transformation. It has never been possible to quantify the meaning of an arbitrary piece of information. Large language models represent information as vectors embedded within high-dimensional semantic spaces, enabling meaning to be analyzed geometrically. This paper defines semantic state as the vector representation of information within a model's semantic space and defines cognitive work as the transformation between semantic states, movement through semantic space. To quantify this transformation, we introduce the Contextual Information Differential (CID), a covariance-aware metric based on Mahalanobis distance measuring semantic displacement relative to the structure of the embedding manifold. By treating meaning as a geometric property of semantic space, CID provides a practical method for measuring semantic change, cognitive transformation, and information processing. This framework establishes a foundation for analyzing cognitive processing, learning, reasoning, knowledge generation, synthetic expertise, and human/AI collaboration through implementation-independent measurement of semantic state change in both human and artificial systems.

## Hide to See: Reasoning-prefix Masking for Visual-anchored Thinking in VLM Distillation

DevFeed: [Hide to See: Reasoning-prefix Masking for Visual-anchored Thinking in VLM Distillation](<https://devfeed.tech/articles/hide-to-see-reasoning-prefix-masking-for-visual-anchored-thinking-in-vlm-distillation-124414.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2605.11651>)

Author: Seonghoon Yu, Dongjun Nam, Byung-Kwan Lee, Jeany Son

Published: 2026-11-09T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Artificial Intelligence](<https://devfeed.tech/sources/arxiv-artificial-intelligence.md>)

Topics: [llm-reasoning](<https://devfeed.tech/topics/llm-reasoning.md>), [Post-training optimization](<https://devfeed.tech/topics/post-training-optimization.md>), [Chain-of-thought](<https://devfeed.tech/topics/chain-of-thought.md>), [Machine Intelligence](<https://devfeed.tech/topics/machine-intelligence.md>), [Artificial Intelligence](<https://devfeed.tech/topics/ai.md>)

Tags: [benchmarks](<https://devfeed.tech/tags/benchmarks.md>), [budget](<https://devfeed.tech/tags/budget.md>), [causal-mask](<https://devfeed.tech/tags/causal-mask.md>), [cs-ai](<https://devfeed.tech/tags/cs-ai.md>), [cs-cl](<https://devfeed.tech/tags/cs-cl.md>), [cs-cv](<https://devfeed.tech/tags/cs-cv.md>), [distillation](<https://devfeed.tech/tags/distillation.md>), [distributions](<https://devfeed.tech/tags/distributions.md>), [mask](<https://devfeed.tech/tags/mask.md>), [masking](<https://devfeed.tech/tags/masking.md>), [multimodal](<https://devfeed.tech/tags/multimodal.md>), [next-token-prediction](<https://devfeed.tech/tags/next-token-prediction.md>), [open-source](<https://devfeed.tech/tags/open-source.md>), [qwen3](<https://devfeed.tech/tags/qwen3.md>), [reasoning](<https://devfeed.tech/tags/reasoning.md>), [scheduling](<https://devfeed.tech/tags/scheduling.md>), [student](<https://devfeed.tech/tags/student.md>), [thinking](<https://devfeed.tech/tags/thinking.md>), [tokens](<https://devfeed.tech/tags/tokens.md>), [vlm](<https://devfeed.tech/tags/vlm.md>), [vlms](<https://devfeed.tech/tags/vlms.md>)

### AI overview

The proposed distillation framework aims to reduce visual forgetting in compact think-answer VLMs by training them to rely on image evidence throughout their reasoning. It selectively masks salient reasoning-prefix tokens for each next-token prediction, then gradually increases the masking budget according to teacher-student distribution discrepancy. During distillation, its mask blocks future tokens and salient reasoning cues, replacing the standard causal mask. The abstract reports stronger results than recent open-source VLMs, distillation, and self-distillation methods on multimodal reasoning benchmarks, and analyses indicating increased visual use during the student's reasoning. It does not provide benchmark names or quantitative results.

### Source excerpt

arXiv:2605.11651v5 Announce Type: replace-cross Abstract: Recent think-answer approaches in VLMs, such as Qwen3-VL-Thinking, boost reasoning performance by leveraging intermediate thinking steps before the final answer, but their computational cost becomes substantial, especially for larger VLMs. To distill such capabilities into compact think-answer VLMs, a primary objective is to improve the student's ability to utilize visual evidence throughout its reasoning trace, as long think-answer traces suffer from visual forgetting issues. To this end, we introduce a novel think-answer distillation framework that encourages the student to anchor its thinking on visual information by masking the student's salient reasoning prefixes. To compensate for such masked textual cues, the student is encouraged to rely more on visual evidence as an alternative source of information during distillation. Our masking strategies include: 1) token-wise salient reasoning-prefix masking, which masks high-influence reasoning prefixes selectively for each next-token prediction, and 2) self-paced masking budget scheduling, which gradually increases the masking scale according to distillation difficulty, measured by the discrepancy between teacher--student distributions. In the distillation phase, the student is guided by our salient reasoning-prefix mask, which blocks both future tokens and salient reasoning cues, in place of the standard causal mask used for auto-regressive language modeling. Experimental results show that our approach outperforms recent open-source VLMs, VLM distillation, and self-distillation methods on multimodal reasoning benchmarks, while further analyzes confirm enhanced visual utilization along the student thinking process.

[Next page](<https://devfeed.tech/topics/machine-intelligence.md?cursor=WyIyMDI2LTExLTA5VDA1OjAwOjAwKzAwOjAwIiwgImM3ZDUwMTQ4LTU5YjAtNDk5OC1hNDI2LTliNjRjNzMxOTEzZiJd>)