# Physical AI

Physical AI is a field involving hardware and software systems that perceive, reason about, learn from, and act in the physical world using sensors and actuators.

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

## SteerTrace: Tracing How Target Attacks Steer World Action Models

DevFeed: [SteerTrace: Tracing How Target Attacks Steer World Action Models](<https://devfeed.tech/articles/steertrace-tracing-how-target-attacks-steer-world-action-models-145575.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.25719>)

Author: Yukun Dai, Tianshi Wang, Fengling Li, Jingjing Li, Xin Li, Lei Zhu

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

Content type: article

Language: en

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

Topics: [World models](<https://devfeed.tech/topics/world-models.md>), [Computer vision](<https://devfeed.tech/topics/computer-vision.md>), [Neural Network](<https://devfeed.tech/topics/neural-network.md>), [Physical AI](<https://devfeed.tech/topics/physical-ai.md>)

Tags: [a-safety](<https://devfeed.tech/tags/a-safety.md>), [actions](<https://devfeed.tech/tags/actions.md>), [attack](<https://devfeed.tech/tags/attack.md>), [attacks](<https://devfeed.tech/tags/attacks.md>), [bounded](<https://devfeed.tech/tags/bounded.md>), [budget](<https://devfeed.tech/tags/budget.md>), [camera](<https://devfeed.tech/tags/camera.md>), [cs-ai](<https://devfeed.tech/tags/cs-ai.md>), [cs-cv](<https://devfeed.tech/tags/cs-cv.md>), [effect](<https://devfeed.tech/tags/effect.md>), [experts](<https://devfeed.tech/tags/experts.md>), [ffn](<https://devfeed.tech/tags/ffn.md>), [frame](<https://devfeed.tech/tags/frame.md>), [future](<https://devfeed.tech/tags/future.md>), [inside](<https://devfeed.tech/tags/inside.md>), [layers](<https://devfeed.tech/tags/layers.md>), [model](<https://devfeed.tech/tags/model.md>), [models](<https://devfeed.tech/tags/models.md>), [noise](<https://devfeed.tech/tags/noise.md>), [target](<https://devfeed.tech/tags/target.md>), [tracing](<https://devfeed.tech/tags/tracing.md>), [world](<https://devfeed.tech/tags/world.md>)

### AI overview

SteerTrace reports that a single per-model perturbation bounded at 4/255, shared across frames, tasks, and camera views, steers actions toward an attacker-chosen target across eight world action models on LIBERO and RoboTwin, bringing task success close to zero. Same-budget noise and BadWAM cause little shift toward the target. Tracing the effect shows different dependencies: under the standard attack, restoring future tokens removes 61-94% of the targeted effect in five models whose action experts attend to them. When predicted futures are kept close to clean predictions, current-frame tokens become more important in three models: restoring them removes 87-99% of the effect, versus 9-15% for future tokens. In six models, restoring 0.36% of video-expert FFN neurons in layers 10-29 removes 72-96% of the effect. The abstract also reports that residual directions 2-16 contain only 11-18% of shift energy, yet removing them cancels most of the effect.

### Source excerpt

arXiv:2611.25719v1 Announce Type: cross Abstract: World action models (WAMs) predict future videos and actions with one backbone, and their imagined future is often seen as a safety signal. We introduce SteerTrace, a targeted universal attack on WAMs that also traces its effect inside the model. One $\ell_\infty$-bounded (4/255) perturbation per model, shared across frames, tasks, and camera views, pushes the executed action toward an attacker-chosen target. On eight WAMs from four families (FastWAM, Faster-WAM, OpenWAM and LingBot-VA) on LIBERO and RoboTwin, it drives task success close to zero and moves the attacked action toward the target. Same-budget noise and BadWAM produce little action shift in the target direction. We use the signed targeted effect to measure how much of the effect each intervention removes or reproduces. Under the standard attack, restoring future tokens removes 61-94% of the targeted effect in the five models whose action experts attend to them. When the attack is constrained to keep the predicted future close to the clean prediction, the targeted effect can become more dependent on current-frame tokens. In three of these five models, restoring current-frame tokens removes 87-99% of the effect, compared with 9-15% for future tokens. In six models, restoring just 0.36% of the FFN neurons in layers 10-29 of the video expert removes 72-96% of the targeted effect. Residual directions 2-16 hold only 11-18% of the shift energy, yet removing them cancels most of the targeted effect.

## ActMem: From Latent Memory Reasoning to Grounded Robotic Action

DevFeed: [ActMem: From Latent Memory Reasoning to Grounded Robotic Action](<https://devfeed.tech/articles/actmem-from-latent-memory-reasoning-to-grounded-robotic-action-145059.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.25461>)

Author: Linfeng Dong, Jianwei Niu, Jiuyi Zhao, Ruibing Feng, Xinghao Wu, Wu Yang, Shuai Tao, Wulong Liu, Mu Li

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

Content type: article

Language: en

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

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

Tags: [attention](<https://devfeed.tech/tags/attention.md>), [chain-of-thought](<https://devfeed.tech/tags/chain-of-thought.md>), [cs-cv](<https://devfeed.tech/tags/cs-cv.md>), [cs-lg](<https://devfeed.tech/tags/cs-lg.md>), [flow-matching](<https://devfeed.tech/tags/flow-matching.md>), [grounded](<https://devfeed.tech/tags/grounded.md>), [grounding](<https://devfeed.tech/tags/grounding.md>), [hidden-states](<https://devfeed.tech/tags/hidden-states.md>), [history](<https://devfeed.tech/tags/history.md>), [horizon](<https://devfeed.tech/tags/horizon.md>), [inference](<https://devfeed.tech/tags/inference.md>), [memory](<https://devfeed.tech/tags/memory.md>), [reasoning](<https://devfeed.tech/tags/reasoning.md>), [robotic](<https://devfeed.tech/tags/robotic.md>)

### AI overview

ActMem separates memory reasoning from grounding actions in current perception within one vision-language backbone. It infers a subtask from timestamped interaction history, retains the resulting hidden states as a Latent Memory Chain-of-Thought to condition a flow-matching action expert, and lets that expert attend directly to the current view. Efficient Memory CoT Inference reuses computation across consecutive decisions without changing the output. The abstract reports 71.1% task success and 80.8% completion success on RoboMemArena, respectively 7.5 and 8.7 percentage points above the strongest baseline; 73.8% average success on MIKASA-Robo; and 68.8% average success on real-world memory tasks, 10 points above the strongest memory-augmented VLA. These are reported results from the abstract; further experimental details are not provided.

### Source excerpt

arXiv:2611.25461v1 Announce Type: cross Abstract: Vision-Language-Action (VLA) models have shown promising results in real-world long-horizon manipulation, where memory plays a crucial role. However, existing memory-augmented VLAs either retrieve past observations without reasoning about what they imply, or reason over the full history in a shared context where it crowds out current perception. Cognitive science treats them as distinct systems: memory reconstructs past experience to decide what to do, while perception guides how to act. We therefore propose ActMem, an in-context memory framework that decouples memory reasoning from perceptual grounding within a single vision-language backbone. Built on a vision-language model with native video understanding, ActMem infers the current subtask from a timestamped interaction history and keeps the hidden states formed during this inference as a Latent Memory Chain-of-Thought (Memory CoT), which conditions a flow-matching action expert. Perception Attention further lets the expert attend directly to the current view. To keep inference fast under long histories, Efficient Memory CoT Inference reuses computation across consecutive decisions without changing the output. ActMem achieves 71.1% task success and 80.8% completion success on RoboMemArena, surpassing the strongest baseline by 7.5 and 8.7 percentage points, and 73.8% average success on MIKASA-Robo, the best among all compared methods. In the real world, it reaches 68.8% average success on exogenous, endogenous, and traceless memory tasks, 10.0 percentage points above the strongest memory-augmented VLA. Code and checkpoints will be released at https://github.com/LigphiDonk/ActMem and https://huggingface.co/donk111/ActMem.

## Cyber-Physical RSI for AI-RAN: A Case Study

DevFeed: [Cyber-Physical RSI for AI-RAN: A Case Study](<https://devfeed.tech/articles/cyber-physical-rsi-for-ai-ran-a-case-study-145554.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.25619>)

Author: Binghan Wu, Kai Zeng, Zhiqi Li, Zhigang Wang, Yang Liu, Ya-Qin Zhang, Joseph Sifakis, Ye Ouyang

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

Content type: article

Language: en

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

Topics: [Physical AI](<https://devfeed.tech/topics/physical-ai.md>), [Fine-tuning](<https://devfeed.tech/topics/fine-tuning.md>), [Omniverse](<https://devfeed.tech/topics/omniverse.md>), [ReAct agent Python](<https://devfeed.tech/topics/react-agent-python.md>)

Tags: [ai](<https://devfeed.tech/tags/ai.md>), [artificial-intelligence](<https://devfeed.tech/tags/artificial-intelligence.md>), [case-study](<https://devfeed.tech/tags/case-study.md>), [computer](<https://devfeed.tech/tags/computer.md>), [cs-ai](<https://devfeed.tech/tags/cs-ai.md>), [experiment](<https://devfeed.tech/tags/experiment.md>), [feedback](<https://devfeed.tech/tags/feedback.md>), [framework](<https://devfeed.tech/tags/framework.md>), [ping-pong](<https://devfeed.tech/tags/ping-pong.md>), [radio](<https://devfeed.tech/tags/radio.md>), [recursive-self-improvement](<https://devfeed.tech/tags/recursive-self-improvement.md>), [rsi](<https://devfeed.tech/tags/rsi.md>), [validation](<https://devfeed.tech/tags/validation.md>)

### AI overview

The paper presents a cyber-physical recursive self-improvement framework for AI-RAN that uses physical measurements to adapt a student model and guide later experiments. It links the student with an experiment-planning coach, a calibratable virtual environment, and a physical host agent. In a ping-pong handover case study, 12 physical executions were organized into six baseline-candidate pairs, and the revised model was evaluated on reserved physical configurations. In short-window tests, raising the time-to-trigger from 320 to 1280 ms reduced the ping-pong count from three to zero. The updated student's grouped-validation MAE was 2.673 round trips/min; on a common three-configuration physical evaluation, pre-feedback and post-feedback MAEs were 1.240 and 1.601 round trips/min, respectively. The results show that measured outcomes inform the feedback loop, while generalization varies with physical configuration.

### Source excerpt

arXiv:2611.25619v1 Announce Type: new Abstract: Learned models for radio access networks (RANs) are often developed using simulation, yet their predictions can shift when exposed to physical propagation conditions, protocol behavior, and equipment implementations. This paper presents a cyber-physical recursive self-improvement (RSI) framework for AI-RAN that connects a task-oriented student model, an experiment-planning coach agent, a calibratable virtual environment, and a physical host agent. Physical measurements are organized as evidence for both model adaptation and subsequent experiment planning. We examine an implementation of the physical-feedback pathway using ping-pong handover as a case study. The system completes 12 physical executions organized into six baseline--candidate pairs, updates the student from measured outcomes, and evaluates the revised model on reserved physical configurations. In the observed short-window tests, increasing the time-to-trigger from 320 to 1280\,ms changes the ping-pong count from three to zero. The updated student achieves a grouped-validation mean absolute error (MAE) of 2.673 round trips/min, while a common three-configuration physical evaluation yields MAEs of 1.240 and 1.601 round trips/min for the pre-feedback and post-feedback models, respectively. These results characterize both the operational feedback loop and the parameter-dependent generalization behavior revealed by new physical experiments.

## Topology-Driven Anti-Entanglement Control for Soft Robots

DevFeed: [Topology-Driven Anti-Entanglement Control for Soft Robots](<https://devfeed.tech/articles/topology-driven-anti-entanglement-control-for-soft-robots-145326.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2605.05236>)

Author: Haoyang Le, Shengxuan Wang, Mohan Chen, Shuo Feng

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

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>), [Physical AI](<https://devfeed.tech/topics/physical-ai.md>), [Distributed Systems & Parallel Computing](<https://devfeed.tech/topics/distributed-systems-parallel-computing.md>)

Tags: [control](<https://devfeed.tech/tags/control.md>), [cs-ai](<https://devfeed.tech/tags/cs-ai.md>), [cs-ro](<https://devfeed.tech/tags/cs-ro.md>), [entanglement](<https://devfeed.tech/tags/entanglement.md>), [robots](<https://devfeed.tech/tags/robots.md>), [topology](<https://devfeed.tech/tags/topology.md>)

### AI overview

The paper proposes topology-driven multi-agent reinforcement learning (TD-MARL) to coordinate soft robots unwinding in constrained environments. It combines centralized learning, where agents share topological state to account for one another's strategies and reduce training instability, with distributed execution that removes the need for inter-robot communication. A topological safety layer uses invariants to assess and mitigate entanglement risk. The abstract reports that full simulation experiments outperform current deep reinforcement learning methods in convergence and anti-entanglement performance; it provides no quantitative results or experimental details.

### Source excerpt

arXiv:2605.05236v3 Announce Type: replace-cross Abstract: In the field of precision manufacturing in complex constrained environments, the role of soft robots is increasingly prominent, and the realization of anti-winding control based on multi-intelligent body reinforcement learning has become a research hotspot. One of the core problems at present is to coordinate multiple robots to complete the unwinding operation in a highly constrained environment. The existing distributed training framework faces some observability challenges in high-density barrier and unstable environments, resulting in poor learning results. This paper proposes a topology-driven Multi-Agent Reinforcement Learning (TD-MARL) framework to coordinate multi-robot systems to avoid entanglement. Specifically, the critical network adopts centralized learning, so that each intelligent body can perceive the strategies of other intelligent bodies by sharing the topological state, thus alleviating the training instability caused by complex interactions; eliminating the demand for communication resources between robots through distributed execution, Upgrade system reliability; the integrated topological security layer uses topological invariants to accurately assess and mitigate the risk of entanglement to avoid the strategy from falling into local difficulties. Finally, the full simulation experiments carried out in the real simulation environment show that the method is better than the current advanced deep reinforcement learning (DRL) method in terms of convergence and anti-winding effect.

## Mind2Act World: A Diagnostic Benchmark for Embodied Mind, Physical Action, and Mind-Action Composition in Robotic Manipulation

DevFeed: [Mind2Act World: A Diagnostic Benchmark for Embodied Mind, Physical Action, and Mind-Action Composition in Robotic Manipulation](<https://devfeed.tech/articles/mind2act-world-a-diagnostic-benchmark-for-embodied-mind-physical-action-and-mind-action-composition-in-robotic-manipulation-145673.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.26444>)

Author: Yuzhi Huang, Jie Wu, Fanding Huang, Junchao Lin, Weichen Zhang, Ziyi Xiong, Yang Li, Zhengwei Xie, Zhimin Wang, Zhen Liu, Jiazhen Huang, Yuzheng Zhang, Yuche Liu, Hongyi Yang, Chen Tang, Shixiang Tang

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

Content type: article

Language: en

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

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

Tags: [act](<https://devfeed.tech/tags/act.md>), [benchmark](<https://devfeed.tech/tags/benchmark.md>), [composition](<https://devfeed.tech/tags/composition.md>), [cs-ai](<https://devfeed.tech/tags/cs-ai.md>), [cs-cv](<https://devfeed.tech/tags/cs-cv.md>), [cs-ro](<https://devfeed.tech/tags/cs-ro.md>), [diagnostic](<https://devfeed.tech/tags/diagnostic.md>), [manipulation](<https://devfeed.tech/tags/manipulation.md>), [mind](<https://devfeed.tech/tags/mind.md>), [robotic](<https://devfeed.tech/tags/robotic.md>), [robotic-manipulation](<https://devfeed.tech/tags/robotic-manipulation.md>), [world](<https://devfeed.tech/tags/world.md>)

### AI overview

Mind2Act World is a benchmark for diagnosing whether robotic manipulation failures stem from reasoning, physical execution, or combining the two through environmental feedback. It contains 30 simulated tasks across five application domains, with three difficulty levels and Clean or Random scenes (180 configurations). Its Mind view tests decisions from past observations, Act tests precise and timely execution, and Mind2Act tests closed-loop tasks where action outcomes inform later decisions; a diagnosis harness attributes failures to these capabilities or their composition. In evaluations of multimodal agents, vision-language-action models, and world-action models, success in individual views did not transfer to combined tasks: the strongest agent succeeded in 52.7% of Mind episodes and 39.0% of Act episodes, but only 12.0% of Mind2Act episodes. The abstract reports that this agent retained task state but struggled with sustained contact, while end-to-end policies performed practiced motions but retained little task state.

### Source excerpt

arXiv:2611.26444v1 Announce Type: cross Abstract: Embodied intelligence aims to let robots perform complex physical tasks with human-like adaptability and reliability, which requires three complementary capabilities: history-based cognitive reasoning, precise and timely physical execution, and the composition of reasoning and execution through environmental feedback. Yet existing benchmarks either evaluate reasoning and execution within the same tasks or assess them separately without their composition, so they reveal little about whether a failure comes from reasoning, execution, or their composition. We introduce Mind2Act World, a diagnostic benchmark of 30 simulated manipulation tasks from five real-world application domains. Each task has three difficulty levels and two scene settings (Clean and Random), yielding 180 configurations. The tasks form three views: Mind tests decisions based on past observations, Act tests precise and timely execution, and Mind2Act combines both in closed-loop tasks where action outcomes guide later decisions. Comparing the three views shows whether a system lacks a capability or fails to compose capabilities it already has. A diagnosis harness further attributes the failed episodes of three representative systems to Mind, Act, or their composition and summarizes each system in a profile card. Our evaluation of multimodal agents, vision-language-action models (VLAs), and world-action models (WAMs) shows that competence in a single view does not carry over to Mind2Act tasks. Even the strongest agent succeeds in 52.7% of Mind and 39.0% of Act episodes but in only 12.0% of Mind2Act episodes. The strongest agent retains task state but struggles with sustained contact, whereas end-to-end policies execute practiced motions but retain little task state. These findings point to composing reasoning and execution through environmental feedback as the main open problem.

## MAWAM: A Multi-Agent World-Action Model for Coordinated Manipulation

DevFeed: [MAWAM: A Multi-Agent World-Action Model for Coordinated Manipulation](<https://devfeed.tech/articles/mawam-a-multi-agent-world-action-model-for-coordinated-manipulation-145467.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.25105>)

Author: Xianzhe Fan, Chi-Pin Huang, Fu-En Yang, Jing Wang, Yecheng Wu, Yu-Cheng Chou, Johan Bjorck, Linxi "Jim" Fan, Jan Kautz, Yu-Chiang Frank Wang, Zhiding Yu, Hengshuang Zhao, Zhiqi Li

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

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>), [Robot Manipulation](<https://devfeed.tech/topics/robot-manipulation.md>), [World models](<https://devfeed.tech/topics/world-models.md>), [Physical AI](<https://devfeed.tech/topics/physical-ai.md>)

Tags: [cs-ai](<https://devfeed.tech/tags/cs-ai.md>), [cs-cv](<https://devfeed.tech/tags/cs-cv.md>), [cs-ro](<https://devfeed.tech/tags/cs-ro.md>), [manipulation](<https://devfeed.tech/tags/manipulation.md>), [model](<https://devfeed.tech/tags/model.md>), [multi-agent](<https://devfeed.tech/tags/multi-agent.md>), [robotics](<https://devfeed.tech/tags/robotics.md>)

### AI overview

MAWAM predicts a robot team's physical future alongside its actions: it jointly generates each robot's future views and executable action chunks, using cross-agent attention while keeping separate control interfaces. On three real-robot tasks, it reports 67.8% mean success, 18.9 percentage points above MA-VLA trained on the same demonstrations and cameras. In simulation, it reports 90.3% mean success across eight tasks with teams of up to four arms. Ablations indicate both cross-agent attention and team-stage future-video supervision matter: removing attention reduced four-arm relay success from 95.0% to 39.0%, while removing that video supervision reduced Place Food success from 78.0% to 47.0%.

### Source excerpt

arXiv:2611.25105v1 Announce Type: cross Abstract: Every move in a robot team changes the world in which its teammates must act. Coordinated manipulation therefore calls for predicting the team's actions together with the physical future they jointly create. We introduce MAWAM, a multi-agent world-action model that jointly generates each robot's future views and executable action chunks in a single sample. Built from a pretrained single-robot world-action backbone, MAWAM jointly denoises all robots' action and future-view tokens through cross-agent attention, while preserving a separate control interface for each robot. MAWAM achieves the highest success on all three real-robot tasks (laundry-bag collection, bath-towel folding, and umbrella handover), with 67.8% mean success, 18.9 percentage points above MA-VLA trained on the same demonstrations and cameras. In simulation, MAWAM reaches 90.3% mean success on eight tasks with teams of up to four arms. Ablations show that cross-agent attention and team-stage future-video supervision both improve coordinated control: without attention, a four-arm relay falls from 95.0% to 39.0%, and without this supervision, Place Food falls from 78.0% to 47.0%.

## GroundingPI: A Grounding Foundation Model towards Physical Intelligence with Visual Primitives

DevFeed: [GroundingPI: A Grounding Foundation Model towards Physical Intelligence with Visual Primitives](<https://devfeed.tech/articles/groundingpi-a-grounding-foundation-model-towards-physical-intelligence-with-visual-primitives-144951.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2609.39601>)

Author: Qize Yu, Lianrui Fan, Boyu Chen, Jiaqi Liang, Xini Ding, Yue Chen, Zetian Song, Yuran Wang, Yi Zou, Kaixuan Wang, Tianxing Chen, Wenxuan Song, Bohan Zhou, Mingleyang Li, Siqiao Huang, Yuqi Ye, Caigao

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

Content type: article

Language: en

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

Topics: [AI Foundation Models](<https://devfeed.tech/topics/ai-foundation-models.md>), [Physical AI](<https://devfeed.tech/topics/physical-ai.md>), [Robot Manipulation](<https://devfeed.tech/topics/robot-manipulation.md>), [Reinforcement learning](<https://devfeed.tech/topics/reinforcement-learning.md>), [grpo](<https://devfeed.tech/topics/grpo.md>), [Fine-tuning](<https://devfeed.tech/topics/fine-tuning.md>), [scaling laws](<https://devfeed.tech/topics/scaling-laws.md>)

Tags: [autonomous-driving](<https://devfeed.tech/tags/autonomous-driving.md>), [benchmarks](<https://devfeed.tech/tags/benchmarks.md>), [cs-ai](<https://devfeed.tech/tags/cs-ai.md>), [cs-cv](<https://devfeed.tech/tags/cs-cv.md>), [cs-lg](<https://devfeed.tech/tags/cs-lg.md>), [cs-ro](<https://devfeed.tech/tags/cs-ro.md>), [fine-tuning](<https://devfeed.tech/tags/fine-tuning.md>), [foundation-model](<https://devfeed.tech/tags/foundation-model.md>), [grounding](<https://devfeed.tech/tags/grounding.md>), [grpo](<https://devfeed.tech/tags/grpo.md>), [multimodal](<https://devfeed.tech/tags/multimodal.md>), [ocr](<https://devfeed.tech/tags/ocr.md>), [pretraining](<https://devfeed.tech/tags/pretraining.md>), [robotic-manipulation](<https://devfeed.tech/tags/robotic-manipulation.md>)

### AI overview

GroundingPI is a 4B model that represents points and boxes as quantized coordinates in a shared vocabulary, trained with multimodal and spatial pretraining, supervised fine-tuning, and GRPO reinforcement learning. The abstract reports an average of 73.68% across 34 grounding benchmarks, ahead of the larger GPT-6 Astra at 71.54%. As a visual backbone, it outperformed evaluated alternatives in all four out-of-distribution settings across RoboTwin 2.0 and RoboCasa-GR1, with gains up to 24.8%; on RoboCasa-GR1, using 50% of demonstrations beat baselines using 75%. Its mean open-loop L2 error on nuScenes was 0.296 m. The authors report that greater pretraining scale improved grounding and manipulation, dense grounding benefited both, and OCR may aid perceptual learning.

### Source excerpt

arXiv:2609.39601v3 Announce Type: replace-cross Abstract: Precise grounding matters. It specifies which object is the target and where that object is, even in clutter and for tiny objects, and it has to be fast enough for closed-loop control. Yet vision-language-action (VLA) and world-action models (WAMs) take perception from general-purpose vision-language and video-generation backbones, which still fail in these settings. We introduce GroundingPI, a 4B grounding foundation model that generates points and boxes as quantized coordinates in a shared vocabulary. Training combines multimodal and spatial pretraining, supervised fine-tuning, and reinforcement learning with GRPO, using supervision from public datasets and dedicated data engines. Against 44 baselines across 34 grounding benchmarks spanning 11 perceptual capabilities, GroundingPI establishes a new state of the art, averaging 73.68%, above the larger GPT-6 Astra (71.54%). As a downstream visual backbone, GroundingPI improves performance on robotic manipulation and autonomous driving. Across RoboTwin 2.0 and RoboCasa-GR1, it outperforms all evaluated backbones in all four out-of-distribution settings, with relative gains of up to 24.8% over the strongest baseline. On RoboCasa-GR1, GroundingPI trained with 50% of the demonstrations outperforms those baselines trained with 75%. Its mean open-loop L2 error on nuScenes is 0.296 m. We systematically analyze GroundingPI's pretraining in scale and data composition. Grounding quality and downstream manipulation improve with pretraining scale, with generalization tracking perceptual gains most closely. Analyzing the data recipe across these 11 perceptual capabilities shows dense grounding's substantial benefits for both, and OCR's potential as a catalyst for perceptual learning. These results support grounding as a perceptual foundation, and dedicated perceptual pretraining as a promising direction for foundation models of physical intelligence.

## Meta's new Muse Charm device may cost $300

DevFeed: [Meta's new Muse Charm device may cost $300](<https://devfeed.tech/articles/meta-s-new-muse-charm-device-may-cost-300-144387.md>)

Original publisher: [Read original article](<https://techcrunch.com/2026/11/19/metas-new-muse-charm-device-may-cost-300/>)

Author: Lucas Ropek

Published: 2026-11-19T21:58:06Z

Content type: news

Language: en

Sources: [TechCrunch | AI News & Artificial Intelligence](<https://devfeed.tech/sources/techcrunch-ai-news-artificial-intelligence.md>)

Topics: [Physical AI](<https://devfeed.tech/topics/physical-ai.md>)

Tags: [ai](<https://devfeed.tech/tags/ai.md>), [ai-wearable](<https://devfeed.tech/tags/ai-wearable.md>), [ai-wearables](<https://devfeed.tech/tags/ai-wearables.md>), [cost](<https://devfeed.tech/tags/cost.md>), [device](<https://devfeed.tech/tags/device.md>), [hardware](<https://devfeed.tech/tags/hardware.md>), [in-brief](<https://devfeed.tech/tags/in-brief.md>), [meta](<https://devfeed.tech/tags/meta.md>), [muse](<https://devfeed.tech/tags/muse.md>), [muse-charm](<https://devfeed.tech/tags/muse-charm.md>), [reporting](<https://devfeed.tech/tags/reporting.md>), [wearable](<https://devfeed.tech/tags/wearable.md>)

### AI overview

The Information reports that Meta plans to price its pocket-sized Muse Charm AI device at $300; Meta did not deny the report and said it would share more soon. The device is intended as a physical companion for Meta's Muse assistant, letting users talk to it, ask questions, and give commands, potentially while wearing it around the neck. The excerpt notes that the price is close to Friend's $250 wearable but much higher than Amazon's Bee Frontier, listed at $75 (then on sale for $52.49); Bee's optional app integration costs $99 per year. Whether customers will accept Muse Charm's price remains uncertain, amid a difficult market for AI wearables.

### Source excerpt

Meta hasn't denied the reporting from The Information, saying only that it will "have more to share soon" about its upcoming AI wearable.

## Robots can do backflips but can't hold a gaze or press piano keys, and scientists think they've fixed that

DevFeed: [Robots can do backflips but can't hold a gaze or press piano keys, and scientists think they've fixed that](<https://devfeed.tech/articles/robots-can-do-backflips-but-can-t-hold-a-gaze-or-press-piano-keys-and-scientists-think-they-ve-fixed-that-143166.md>)

Original publisher: [Read original article](<https://www.techspot.com/news/114610-robots-can-do-backflips-but-cant-hold-gaze.html>)

Author: Skye Jacobs

Published: 2026-11-19T11:51:00Z

Content type: news

Language: en

Sources: [TechSpot](<https://devfeed.tech/sources/techspot.md>)

Topics: [Robot Manipulation](<https://devfeed.tech/topics/robot-manipulation.md>), [Physical AI](<https://devfeed.tech/topics/physical-ai.md>), [Humanoid Robots](<https://devfeed.tech/topics/humanoid-robots.md>), [3D](<https://devfeed.tech/topics/3d.md>)

Tags: [ai](<https://devfeed.tech/tags/ai.md>), [computer-vision](<https://devfeed.tech/tags/computer-vision.md>), [easy](<https://devfeed.tech/tags/easy.md>), [embedded](<https://devfeed.tech/tags/embedded.md>), [engineer](<https://devfeed.tech/tags/engineer.md>), [engineers](<https://devfeed.tech/tags/engineers.md>), [explore](<https://devfeed.tech/tags/explore.md>), [eye](<https://devfeed.tech/tags/eye.md>), [eyeballs](<https://devfeed.tech/tags/eyeballs.md>), [eyes](<https://devfeed.tech/tags/eyes.md>), [feedback](<https://devfeed.tech/tags/feedback.md>), [fiber](<https://devfeed.tech/tags/fiber.md>), [flexible](<https://devfeed.tech/tags/flexible.md>), [folding](<https://devfeed.tech/tags/folding.md>), [gauge](<https://devfeed.tech/tags/gauge.md>), [graphene](<https://devfeed.tech/tags/graphene.md>), [groups](<https://devfeed.tech/tags/groups.md>), [hand](<https://devfeed.tech/tags/hand.md>), [hardware](<https://devfeed.tech/tags/hardware.md>), [materials](<https://devfeed.tech/tags/materials.md>), [muscular](<https://devfeed.tech/tags/muscular.md>), [research](<https://devfeed.tech/tags/research.md>), [robots](<https://devfeed.tech/tags/robots.md>), [science](<https://devfeed.tech/tags/science.md>)

### AI overview

University of Pennsylvania engineers built a soft artificial muscle fiber that can hold a precise position by sensing its own temperature and correcting heat loss. A liquid crystal elastomer contracts when heated, graphene makes it softer and tougher, and an embedded copper coil serves as both heater and temperature-feedback sensor. The system adjusts heating in real time to counter the fiber's tendency to cool and drift back, without a separate sensor. In lab demonstrations, a robot hand varied the force and speed of piano-key presses, while paired fibers moved robotic eyes with controlled, repeatable gaze. The hand was not a skilled pianist. The researchers' next step is combining many fibers into larger systems; coordinating them with AI computer vision is a longer-term goal.

### Source excerpt

The work, published in Advanced Functional Materials, is modeled on the small muscles that move human fingers and eyes. Those muscles make fine, controlled movements, and that kind of control has been hard to get from the hardware inside most robots today. Read Entire Article

## Adaptive Action Horizon Selection via Flow-Field Uncertainty

DevFeed: [Adaptive Action Horizon Selection via Flow-Field Uncertainty](<https://devfeed.tech/articles/adaptive-action-horizon-selection-via-flow-field-uncertainty-143490.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.24178>)

Author: Idan Ben-Ami, Arnon Netzer, Sarah Keren, Idit Diamant

Published: 2026-11-19T05:00:00Z

Content type: article

Language: en

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

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

Tags: [adaptive](<https://devfeed.tech/tags/adaptive.md>), [cost](<https://devfeed.tech/tags/cost.md>), [covariance](<https://devfeed.tech/tags/covariance.md>), [cs-lg](<https://devfeed.tech/tags/cs-lg.md>), [cs-ro](<https://devfeed.tech/tags/cs-ro.md>), [decision](<https://devfeed.tech/tags/decision.md>), [decision-making](<https://devfeed.tech/tags/decision-making.md>), [dispersion](<https://devfeed.tech/tags/dispersion.md>), [diversity](<https://devfeed.tech/tags/diversity.md>), [entropy](<https://devfeed.tech/tags/entropy.md>), [execution](<https://devfeed.tech/tags/execution.md>), [field](<https://devfeed.tech/tags/field.md>), [flow](<https://devfeed.tech/tags/flow.md>), [horizon](<https://devfeed.tech/tags/horizon.md>), [inference](<https://devfeed.tech/tags/inference.md>), [information](<https://devfeed.tech/tags/information.md>), [jacobian](<https://devfeed.tech/tags/jacobian.md>), [language](<https://devfeed.tech/tags/language.md>), [latency](<https://devfeed.tech/tags/latency.md>), [less](<https://devfeed.tech/tags/less.md>), [selection](<https://devfeed.tech/tags/selection.md>), [uncertainty](<https://devfeed.tech/tags/uncertainty.md>)

### AI overview

DUAC estimates trajectory-conditioned uncertainty for adaptive action horizons in pretrained vision-language-action models from a single generated action chunk, using Jacobian-vector products and a closed-form flow-posterior covariance formulation. It requires neither retraining nor multiple endpoint samples, reducing estimation latency compared with AAC's multiple sampled chunks. The methods capture different signals: AAC measures global dispersion across strategies, while DUAC measures local uncertainty around the selected trajectory and is less sensitive to diversity among distinct valid strategies. Across LIBERO, LIBERO-Pro, LIBERO-Plus, and RoboCasa, DUAC-based horizon selection performed favorably among evaluated methods, with strong task success and low uncertainty-estimation latency.

### Source excerpt

arXiv:2611.24178v1 Announce Type: new Abstract: Reliable uncertainty estimates are important for adaptive execution in vision-language-action (VLA) models. Existing approaches such as Adaptive Action Chunking (AAC) estimate action entropy from multiple sampled action chunks, introducing additional inference cost and capturing both uncertainty within an action trajectory and diversity across alternative strategies. We introduce Divergence-based Uncertainty-aware Adaptive Action Chunking (DUAC), a post-hoc method that extracts trajectory-conditioned predictive uncertainty from the vector field of a pretrained VLA. Building on a closed-form flow-posterior covariance formulation, DUAC derives action-level uncertainty scores from a single generated chunk using Jacobian-vector products, without retraining or multiple endpoint samples. DUAC strongly correlates with AAC, while capturing distinct information: AAC measures global dispersion across sampled chunks, whereas DUAC measures local uncertainty around the trajectory selected for execution. Consequently, DUAC is less sensitive to diversity among distinct but valid action strategies. Across LIBERO, LIBERO-Pro, LIBERO-Plus, and RoboCasa, DUAC-based adaptive horizon selection performs favorably among the evaluated methods, combining strong task success with low uncertainty-estimation latency. These results establish the flow vector field as an efficient, trajectory-conditioned source of predictive uncertainty for robot decision making.

## Autonomous Video Generation for Self-Evolving World Models

DevFeed: [Autonomous Video Generation for Self-Evolving World Models](<https://devfeed.tech/articles/autonomous-video-generation-for-self-evolving-world-models-143281.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2606.24152>)

Author: Xin Wang, Wenxuan Liu, Tongtong Feng, Wenwu Zhu

Published: 2026-11-19T05:00:00Z

Content type: article

Language: en

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

Topics: [World models](<https://devfeed.tech/topics/world-models.md>), [Physical AI](<https://devfeed.tech/topics/physical-ai.md>)

Tags: [arxiv](<https://devfeed.tech/tags/arxiv.md>), [autonomous](<https://devfeed.tech/tags/autonomous.md>), [bind](<https://devfeed.tech/tags/bind.md>), [consistency](<https://devfeed.tech/tags/consistency.md>), [cs-cv](<https://devfeed.tech/tags/cs-cv.md>), [cs-lg](<https://devfeed.tech/tags/cs-lg.md>), [decision-making](<https://devfeed.tech/tags/decision-making.md>), [delay](<https://devfeed.tech/tags/delay.md>), [distillation](<https://devfeed.tech/tags/distillation.md>), [distribution](<https://devfeed.tech/tags/distribution.md>), [drones](<https://devfeed.tech/tags/drones.md>), [efficiency](<https://devfeed.tech/tags/efficiency.md>), [environments](<https://devfeed.tech/tags/environments.md>), [evaluation](<https://devfeed.tech/tags/evaluation.md>), [feedback](<https://devfeed.tech/tags/feedback.md>), [framework](<https://devfeed.tech/tags/framework.md>), [generation](<https://devfeed.tech/tags/generation.md>), [models](<https://devfeed.tech/tags/models.md>), [video](<https://devfeed.tech/tags/video.md>), [video-generation](<https://devfeed.tech/tags/video-generation.md>), [world](<https://devfeed.tech/tags/world.md>), [world-models](<https://devfeed.tech/tags/world-models.md>)

### AI overview

The article argues that world models should be judged by whether their generated futures support valid action for a particular embodied agent, not by visual plausibility alone. It proposes autonomous video generation evaluated through counterfactual controllability: generating intervention-conditioned futures, binding them to embodiment constraints, checking them under distribution shifts, and distilling surviving branches into compact decision variables. Its closed loop has four stages--Generation, Binding, Verification, and Distillation--with corresponding metrics of novelty, consistency, out-of-distribution performance, and efficiency. Drones and manipulators are proposed as testbeds, where factors such as wind, sensing limits, actuation delay, contact dynamics, and recovery constraints can be perturbed and verified. The supplied text presents this as a framework and examples, not evidence of demonstrated performance.

### Source excerpt

arXiv:2606.24152v3 Announce Type: replace-cross Abstract: Large-scale video generation models are increasingly described as world models because they can learn rich spatiotemporal regularities from visual data. We argue that an ideal world model should benefit in a self-evolving generative character. As such, a self-evolving world model can be realized via a generative model conditioned by various constraints, whose future predictions are refined through feedback. However, traditional visually plausible predictions alone are not enough to establish whether an imagined future is physically actionable for a particular embodied agent, failing to provide informative feedback from environments for self-evolving improvement. To realize self-evolving world models, this article proposes the concept of autonomous video generation, which is evaluated through counterfactual controllability, i.e., the ability to i) generate intervention-conditioned futures, ii) bind these future frames to embodiment constraints, iii) verify them under distribution shifts, and iv) distil surviving branches into compact variables for decision-making. We formalize a four-stage closed-loop optimization of Generation, Binding, Verification and Distillation, together with four corresponding evaluation metrics: novelty, consistency, out-of-distribution (OOD) and efficiency. We further discuss two examples, i.e., drones and manipulators, as early embodied testbeds where wind, sensing limits, actuation delay, contact dynamics and recovery constraints can be systematically perturbed and verified. The central claim is that the framework of autonomous video generation for self-evolving world models should not be judged by video fidelity alone, but by whether the generated frames improve valid action under counterfactual interventions and various embodiment constraints.

## From Points to Trajectories: Rethinking Multimodal Alignment for Physical Events

DevFeed: [From Points to Trajectories: Rethinking Multimodal Alignment for Physical Events](<https://devfeed.tech/articles/from-points-to-trajectories-rethinking-multimodal-alignment-for-physical-events-142771.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.24249>)

Author: Yiyang Zhu, Guangyuan Liu, Jingdi Lei, Li Wei, Chau Yuen

Published: 2026-11-19T05:00:00Z

Content type: article

Language: en

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

Topics: [Physical AI](<https://devfeed.tech/topics/physical-ai.md>)

Tags: [ai](<https://devfeed.tech/tags/ai.md>), [alignment](<https://devfeed.tech/tags/alignment.md>), [architectures](<https://devfeed.tech/tags/architectures.md>), [artificial-intelligence](<https://devfeed.tech/tags/artificial-intelligence.md>), [arxiv](<https://devfeed.tech/tags/arxiv.md>), [cs-ai](<https://devfeed.tech/tags/cs-ai.md>), [eess-sp](<https://devfeed.tech/tags/eess-sp.md>), [embedding](<https://devfeed.tech/tags/embedding.md>), [event](<https://devfeed.tech/tags/event.md>), [events](<https://devfeed.tech/tags/events.md>), [frame](<https://devfeed.tech/tags/frame.md>), [gap](<https://devfeed.tech/tags/gap.md>), [impact](<https://devfeed.tech/tags/impact.md>), [lightweight](<https://devfeed.tech/tags/lightweight.md>), [matching](<https://devfeed.tech/tags/matching.md>), [module](<https://devfeed.tech/tags/module.md>), [multimodal](<https://devfeed.tech/tags/multimodal.md>), [object](<https://devfeed.tech/tags/object.md>), [physical-ai](<https://devfeed.tech/tags/physical-ai.md>), [real-world](<https://devfeed.tech/tags/real-world.md>), [recognition](<https://devfeed.tech/tags/recognition.md>), [sensor](<https://devfeed.tech/tags/sensor.md>)

### AI overview

The paper argues that pooling each sensor stream into one embedding erases transient cross-modal dynamics, making distinct physical executions under the same task conditions indistinguishable as sequences grow longer. It identifies synchronous frame-level co-fluctuations--termed the event gap--as the signal needed to distinguish such events. CounterPoint removes each modality's temporal mean and scores frame-level cross-modal agreement; the abstract reports that adding it to four multimodal aligners improves physical event alignment over existing temporal alignment methods.

### Source excerpt

arXiv:2611.24249v1 Announce Type: new Abstract: Advancing physical artificial intelligence (Physical AI) requires moving beyond static object recognition to ground how real-world mechanics (e.g., impact, friction, and force transfer) unfold over time. While multimodal aligners seek to capture these interactions by matching multi-sensor streams, prevailing architectures compress each temporal sequence into a single pooled embedding. We show that this paradigm fails on twin events: distinct physical executions recorded under identical task conditions. Collapsing the temporal dimension averages out transient cross-modal dynamics, effectively reducing physical event alignment to static scene recognition. Formally, we prove that as sequence length increases, any temporally pooled representation converges to the stationary recording condition. Consequently, as windows grow, no choice of loss function or embedding geometry can disambiguate twin events from pooled features alone. Instead, we show that physical events are uniquely identified by the event gap, namely the instantaneous, synchronous co-fluctuations across modalities that are erased by pooling and accumulate only along unpooled trajectories. To recover this signal, we introduce CounterPoint, a lightweight plug-in module that strips out each modality's temporal mean to score frame-level cross-modal agreement. Integrated across four state-of-the-art (SOTA) multimodal aligners, CounterPoint shifts representations from coarse condition matching to fine-grained physical event alignment, consistently outperforming existing temporal alignment methods.

## Latent Action Policies from Suboptimal Data

DevFeed: [Latent Action Policies from Suboptimal Data](<https://devfeed.tech/articles/latent-action-policies-from-suboptimal-data-143529.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.24482>)

Author: Donghyeon Ki, Xiaoyu Chen, Jun Ho Seo, Hayeong Lee, Kyungyoon Kim, Byung-Jun Lee, Li Zhao

Published: 2026-11-19T05:00:00Z

Content type: article

Language: en

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

Topics: [Machine learning](<https://devfeed.tech/topics/machine-learning.md>), [Reinforcement learning](<https://devfeed.tech/topics/reinforcement-learning.md>), [Representation learning](<https://devfeed.tech/topics/representation-learning.md>), [Physical AI](<https://devfeed.tech/topics/physical-ai.md>)

Tags: [actions](<https://devfeed.tech/tags/actions.md>), [approach](<https://devfeed.tech/tags/approach.md>), [capture](<https://devfeed.tech/tags/capture.md>), [conditional](<https://devfeed.tech/tags/conditional.md>), [control](<https://devfeed.tech/tags/control.md>), [controls](<https://devfeed.tech/tags/controls.md>), [cs-lg](<https://devfeed.tech/tags/cs-lg.md>), [data](<https://devfeed.tech/tags/data.md>), [execution](<https://devfeed.tech/tags/execution.md>), [flow](<https://devfeed.tech/tags/flow.md>), [improvement](<https://devfeed.tech/tags/improvement.md>), [manipulation](<https://devfeed.tech/tags/manipulation.md>), [multimodal](<https://devfeed.tech/tags/multimodal.md>), [offline](<https://devfeed.tech/tags/offline.md>), [optimization](<https://devfeed.tech/tags/optimization.md>), [policies](<https://devfeed.tech/tags/policies.md>), [policy](<https://devfeed.tech/tags/policy.md>), [rl](<https://devfeed.tech/tags/rl.md>), [space](<https://devfeed.tech/tags/space.md>), [spaces](<https://devfeed.tech/tags/spaces.md>), [stat-ml](<https://devfeed.tech/tags/stat-ml.md>), [value](<https://devfeed.tech/tags/value.md>)

### AI overview

Latent Action Models learn control representations from video without action labels, but behavioral cloning does not explicitly improve policies. The abstract reports that naive offline RL in these latent spaces can yield actions that cannot be physically executed, and that latent representations may merge controls with different rewards. LAPIS addresses this with goal-conditioned policy improvement grounded in observed state-latent transitions and a conditional flow policy for multimodal actions. It reportedly outperforms baselines on most tasks, with the clearest gains in visual manipulation, and remains effective with 1% action labels where competing methods largely collapse.

### Source excerpt

arXiv:2611.24482v1 Announce Type: new Abstract: Latent Action Models (LAMs) learn control representations from action-free video, but existing pipelines typically train latent policies with behavioral cloning, without an explicit policy improvement objective. Although offline RL offers a natural approach to policy improvement, standard methods do not transfer directly to learned latent action spaces. We find that the structure of learned latent actions makes naive offline RL unreliable: directly maximizing value over latent actions or fitting a unimodal policy can produce latent actions that cannot be realized by physical execution, while the latent representation can merge physical controls that receive different rewards. To understand when latent-space optimization remains valid, we derive conditions under which policy improvement transfers to decoded execution and under which latent and physical optima align. Building on these insights, we propose LAPIS for goal-conditioned latent policy improvement, grounding value-based policy improvement in observed state-latent transitions and using a conditional flow policy to capture their multimodal structure. LAPIS outperforms baselines on most tasks, with the clearest gains in visual manipulation, and remains effective even with only 1% action labels, where competing methods largely collapse.

## Worth Another Look:Constrained World-Model Planning for Visual Inspection

DevFeed: [Worth Another Look:Constrained World-Model Planning for Visual Inspection](<https://devfeed.tech/articles/worth-another-look-constrained-world-model-planning-for-visual-inspection-142764.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.24187>)

Author: Ziyue Kang, Zixi Qi, Yuchen Dai, Nana Lv, Xirong Cheng, Liang Liu, Ying Li, Haoyan Wei, Haiyang Yao

Published: 2026-11-19T05:00:00Z

Content type: article

Language: en

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

Topics: [World models](<https://devfeed.tech/topics/world-models.md>), [Multi-step AI processes](<https://devfeed.tech/topics/multi-step-ai-processes.md>), [Physical AI](<https://devfeed.tech/topics/physical-ai.md>)

Tags: [adaptive](<https://devfeed.tech/tags/adaptive.md>), [beam-search](<https://devfeed.tech/tags/beam-search.md>), [benchmarks](<https://devfeed.tech/tags/benchmarks.md>), [budget](<https://devfeed.tech/tags/budget.md>), [cs-ai](<https://devfeed.tech/tags/cs-ai.md>), [cs-lg](<https://devfeed.tech/tags/cs-lg.md>), [cs-mm](<https://devfeed.tech/tags/cs-mm.md>), [cs-ro](<https://devfeed.tech/tags/cs-ro.md>), [environmental-monitoring](<https://devfeed.tech/tags/environmental-monitoring.md>), [fallback](<https://devfeed.tech/tags/fallback.md>), [inspection](<https://devfeed.tech/tags/inspection.md>), [lightweight](<https://devfeed.tech/tags/lightweight.md>), [lookahead](<https://devfeed.tech/tags/lookahead.md>), [observability](<https://devfeed.tech/tags/observability.md>), [offline](<https://devfeed.tech/tags/offline.md>), [online](<https://devfeed.tech/tags/online.md>), [optical](<https://devfeed.tech/tags/optical.md>), [pairing](<https://devfeed.tech/tags/pairing.md>), [performance](<https://devfeed.tech/tags/performance.md>), [planning](<https://devfeed.tech/tags/planning.md>), [policies](<https://devfeed.tech/tags/policies.md>), [pricing](<https://devfeed.tech/tags/pricing.md>), [visual-inspection](<https://devfeed.tech/tags/visual-inspection.md>), [world-model](<https://devfeed.tech/tags/world-model.md>)

### AI overview

The paper proposes budget-aware visual inspection that combines a compact recurrent world model with exact symbolic feasibility and tariff rules. Receding-horizon beam search replans schedules for class queries, crops, image enhancements, and stopping; slack-aware shadow pricing and a legal fallback support larger budgets. Across eight terrestrial, optical, and sonar benchmarks, multi-step lookahead beats static schedules and search-free offline RL under tight budgets, with gains of up to 28.4 AP points over static priority on RUOD and 15.1 on ARIS. Dynamic pricing improves large-budget results across all eight domains, at an 11 ms planning cost. The approach uses frozen detectors without retraining.

### Source excerpt

arXiv:2611.24187v1 Announce Type: cross Abstract: Visual inspection in field robotics and environmental monitoring must localize multiple targets under strict compute and sensing caps. Exhaustive sweeps waste calls on absent classes, while myopic and search-free policies stumble over heterogeneous tariff cliffs or discard viewport edits whose value emerges only after re-querying. We formulate inspection over frozen detectors as neuro-symbolic observation planning under partial observability, decoupling a compact recurrent world model (WM) from exact symbolic feasibility and tariff rules. Receding-horizon beam search with online replanning jointly schedules class queries, crops, enhancements, and termination (stop), paired with slack-aware shadow pricing and legal fallback at larger budgets. Across eight terrestrial, optical, and sonar benchmarks, multi-step lookahead outperforms static schedules and search-free offline RL under tight budgets -- gaining up to 28.4 AP percentage points over static priority on RUOD and 15.1 on ARIS -- while dynamic pricing lifts large-budget performance on all eight domains at 11 ms planning cost. Ultimately, pairing lightweight learned dynamics with exact symbolic constraints turns frozen detectors into adaptive, budget-aware inspection agents without retraining.

## RoboBuilder: Agentic Robot Co-Design with Real-world Video Feedback in the Loop

DevFeed: [RoboBuilder: Agentic Robot Co-Design with Real-world Video Feedback in the Loop](<https://devfeed.tech/articles/robobuilder-agentic-robot-co-design-with-real-world-video-feedback-in-the-loop-142711.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.23855>)

Author: Ghadi Nehme, Dahyun Nam, Akshay Govind Srinivasan, Faez Ahmed

Published: 2026-11-19T05:00:00Z

Content type: article

Language: en

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

Topics: [Physical AI](<https://devfeed.tech/topics/physical-ai.md>), [3D](<https://devfeed.tech/topics/3d.md>), [Robot Navigation](<https://devfeed.tech/topics/robot-navigation.md>), [Computer vision](<https://devfeed.tech/topics/computer-vision.md>)

Tags: [3d](<https://devfeed.tech/tags/3d.md>), [3d-printed](<https://devfeed.tech/tags/3d-printed.md>), [agent](<https://devfeed.tech/tags/agent.md>), [agentic](<https://devfeed.tech/tags/agentic.md>), [agents](<https://devfeed.tech/tags/agents.md>), [ai](<https://devfeed.tech/tags/ai.md>), [ai-agents](<https://devfeed.tech/tags/ai-agents.md>), [benchmark](<https://devfeed.tech/tags/benchmark.md>), [co-design](<https://devfeed.tech/tags/co-design.md>), [coding](<https://devfeed.tech/tags/coding.md>), [coding-agents](<https://devfeed.tech/tags/coding-agents.md>), [course](<https://devfeed.tech/tags/course.md>), [cs-ai](<https://devfeed.tech/tags/cs-ai.md>), [cs-cv](<https://devfeed.tech/tags/cs-cv.md>), [cs-ro](<https://devfeed.tech/tags/cs-ro.md>), [custom](<https://devfeed.tech/tags/custom.md>), [data](<https://devfeed.tech/tags/data.md>), [design](<https://devfeed.tech/tags/design.md>), [feedback](<https://devfeed.tech/tags/feedback.md>), [in-the-loop](<https://devfeed.tech/tags/in-the-loop.md>), [real-world](<https://devfeed.tech/tags/real-world.md>), [robot](<https://devfeed.tech/tags/robot.md>), [video](<https://devfeed.tech/tags/video.md>)

### AI overview

RoboBuilder evaluates coding agents that co-design robot hardware and programs, build designs that pass simulation, and test them on a physical obstacle course. In the reported experiments, none of 15 initial simulation-passing designs crossed the real course, motivating agents to learn simulator behavior from requested measurements and videos. Across four agents, ten obstacles, and three difficulty levels, agents solved nearly all easy obstacles and produced mechanisms not reached by the search baselines, but struggled with harder levels and the task of one robot crossing every obstacle. Only 66% of simulated passes held in reality across levels. Providing video of a real failure led agents to redesign robots, sometimes correcting the simulator, and removed 72% of level-1 and level-2 failures.

### Source excerpt

arXiv:2611.23855v1 Announce Type: new Abstract: Can AI agents design robots that work in the real world? We introduce RoboBuilder, a benchmark in which coding agents co-design a robot's hardware and program from standard parts, motors and a custom 3D-printed part, and every design that passes in simulation is built and tested on a physical course. Simulation alone is not enough: under public part data, none of 15 initial passing designs crossed the real course, so an agent learns the simulator from measurements it requests and videos of real runs. Across four frontier coding agents, ten obstacles at three difficulty levels, 16,427 simulations of 12,774 distinct robots and 135 physical robots built in an estimated 240 hours, agents solve nearly all easy obstacles and invent mechanisms our search baselines do not reach, but falter at harder levels, when one robot must cross every obstacle, and in the real world, where 66% of simulated passes hold across all levels. Closing the loop helps: shown a video of its real failure, an agent redesigns the robot, sometimes correcting its simulator, and removes 72% of level-1 and level-2 failures. RoboBuilder measures progress toward agents that engineer real machines. Project Page: https://ghadinehme.com/robobuilder.github.io/

## How Uncertain Is My Action? Self-Correcting Visual Planning with Frozen World Models

DevFeed: [How Uncertain Is My Action? Self-Correcting Visual Planning with Frozen World Models](<https://devfeed.tech/articles/how-uncertain-is-my-action-self-correcting-visual-planning-with-frozen-world-models-143418.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.23620>)

Author: Siqing Li, Luming Yang, Yijia Jiao, Junze Deng, Alper Yilmaz

Published: 2026-11-19T05:00:00Z

Content type: news

Language: en

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

Topics: [World models](<https://devfeed.tech/topics/world-models.md>), [Physical AI](<https://devfeed.tech/topics/physical-ai.md>)

Tags: [cs-cv](<https://devfeed.tech/tags/cs-cv.md>), [cs-lg](<https://devfeed.tech/tags/cs-lg.md>), [cs-ro](<https://devfeed.tech/tags/cs-ro.md>), [detection](<https://devfeed.tech/tags/detection.md>), [encoder](<https://devfeed.tech/tags/encoder.md>), [error](<https://devfeed.tech/tags/error.md>), [filter](<https://devfeed.tech/tags/filter.md>), [interface](<https://devfeed.tech/tags/interface.md>), [offset](<https://devfeed.tech/tags/offset.md>), [oracle](<https://devfeed.tech/tags/oracle.md>), [planner](<https://devfeed.tech/tags/planner.md>), [planning](<https://devfeed.tech/tags/planning.md>), [prediction](<https://devfeed.tech/tags/prediction.md>), [raises](<https://devfeed.tech/tags/raises.md>), [recover](<https://devfeed.tech/tags/recover.md>), [robotic](<https://devfeed.tech/tags/robotic.md>), [shift](<https://devfeed.tech/tags/shift.md>), [test-time-adaptation](<https://devfeed.tech/tags/test-time-adaptation.md>), [world-models](<https://devfeed.tech/tags/world-models.md>)

### AI overview

A frozen latent world model can detect and compensate for differences between commanded and executed actions by predicting observations under hypothesized actuation shifts and scoring only the residual component attributable to a shift. The score supports locally optimal detection, shift estimation with Cramer-Rao error bars, and identification of the dominant shift generator. In PushT and OGBench-Cube, self-correction raises success from 0.650 to 0.862 and from 0.378 to 0.432, respectively, matching an oracle given the true shift at the same block and outperforming five common test-time adaptation methods. On a real arm, it detects an unregistered base-joint offset and reduces reach error from 36 mm to 3-5 mm; for sealed random shifts, estimates recover 92-94% of the error reduction achieved using the true shift.

### Source excerpt

arXiv:2611.23620v1 Announce Type: cross Abstract: A frozen latent world model combined with sampling-based model predictive control is a visual planner. At deployment, the executed action can depart from the commanded one, and the frozen model has no way to notice. We show that the model can be run in reverse: for any hypothesized actuation shift, it predicts what it should have observed under that shift, and comparing this prediction with what was actually observed quantifies the shift and lets the planner correct itself. Unlike the residual energy, which mixes the shift with the model's own error, the score keeps only the part of the residual that a shift can produce. This shift score detects the shift with a locally optimal test, estimates it with Cramer-Rao error bars, names the dominant generator, and compensates the planner once an alarm fires. On PushT, TwoRoom and OGBench-Cube with a frozen latent world model, measured detection follows the prediction to within 0.05 AUROC. Self-correction raises success from 0.650 to 0.862 on PushT and from 0.378 to 0.432 on the harder Cube task, levels with an oracle that is told the true shift at the same block, better than five common test-time adaptation methods. On a real robotic arm, the score identifies a base-joint offset that no joint encoder registers and reduces the reach error from 36 mm to 3-5 mm, and under shifts drawn at random and sealed before the session, estimates recover 92-94% of the error reduction achieved with the true shift. The broader implication is that a frozen world model can act as a filter by predicting the future, testing action-interface uncertainty and validity, and finally correcting failures.

## Meet the Anzellans: Inside Imagineering R&D's newest robotic characters

DevFeed: [Meet the Anzellans: Inside Imagineering R&D's newest robotic characters](<https://devfeed.tech/articles/meet-the-anzellans-inside-imagineering-r-d-s-newest-robotic-characters-143691.md>)

Original publisher: [Read original article](<https://aws.amazon.com/blogs/media/meet-the-anzellans-inside-imagineering-rds-newest-robotic-characters/>)

Author: Ari Entin

Published: 2026-11-18T20:03:05Z

Content type: news

Language: en

Sources: [AWS for M&E Blog](<https://devfeed.tech/sources/aws-for-m-e-blog.md>)

Topics: [Physical AI](<https://devfeed.tech/topics/physical-ai.md>), [Machine Perception](<https://devfeed.tech/topics/machine-perception.md>), [Humanoid Robots](<https://devfeed.tech/topics/humanoid-robots.md>)

Tags: [ai](<https://devfeed.tech/tags/ai.md>), [animations](<https://devfeed.tech/tags/animations.md>), [animatronics](<https://devfeed.tech/tags/animatronics.md>), [audio](<https://devfeed.tech/tags/audio.md>), [aws](<https://devfeed.tech/tags/aws.md>), [aws-re-invent](<https://devfeed.tech/tags/aws-re-invent.md>), [cloud](<https://devfeed.tech/tags/cloud.md>), [cloud-infrastructure](<https://devfeed.tech/tags/cloud-infrastructure.md>), [collaboration](<https://devfeed.tech/tags/collaboration.md>), [customer-solutions](<https://devfeed.tech/tags/customer-solutions.md>), [decision-making](<https://devfeed.tech/tags/decision-making.md>), [decision-models](<https://devfeed.tech/tags/decision-models.md>), [disney](<https://devfeed.tech/tags/disney.md>), [disney-plus](<https://devfeed.tech/tags/disney-plus.md>), [disneyland](<https://devfeed.tech/tags/disneyland.md>), [engineering](<https://devfeed.tech/tags/engineering.md>), [english](<https://devfeed.tech/tags/english.md>), [intermediate-200](<https://devfeed.tech/tags/intermediate-200.md>), [prototype](<https://devfeed.tech/tags/prototype.md>), [real-time](<https://devfeed.tech/tags/real-time.md>), [robotic](<https://devfeed.tech/tags/robotic.md>), [walt-disney-imagineering](<https://devfeed.tech/tags/walt-disney-imagineering.md>)

### AI overview

The Anzellans are Audio-Animatronics prototypes that respond to guests and one another in real time instead of repeating a fixed performance. AWS powers their real-time intelligence: fast-decision models support quick reactions, while larger models weigh context. The system selects prerecorded audio and pairs it with dynamic animations, with components optimized for latency. Imagineers created the dialogue, gestures and personalities, keeping the characters within the designed story. The excerpt describes a prototype and an AWS re:Invent showcase; it does not establish a broader public deployment.

### Source excerpt

Meet the Anzellans -- Vilo Vele, Kado Bre'a, and Jibu Bre'a -- the newest Audio-Animatronics character prototype from Walt Disney Imagineering R&D. Built with AWS, these autonomous characters don't run a fixed loop. They perceive guests in real time, decide how to respond, and act on their own, all within the story Imagineers designed.

## Waymo pulls driverless cars from highways nationwide after California ban

DevFeed: [Waymo pulls driverless cars from highways nationwide after California ban](<https://devfeed.tech/articles/waymo-pulls-driverless-cars-from-highways-nationwide-after-california-ban-140873.md>)

Original publisher: [Read original article](<https://www.techspot.com/news/114606-waymo-pulls-driverless-cars-highways-nationwide-after-california.html>)

Author: Skye Jacobs

Published: 2026-11-18T16:38:00Z

Content type: news

Language: en

Sources: [TechSpot](<https://devfeed.tech/sources/techspot.md>)

Topics: [autonomous vehicles](<https://devfeed.tech/topics/autonomous-vehicles.md>), [Physical AI](<https://devfeed.tech/topics/physical-ai.md>), [Resilience](<https://devfeed.tech/topics/resilience.md>)

Tags: [airport](<https://devfeed.tech/tags/airport.md>), [autonomous](<https://devfeed.tech/tags/autonomous.md>), [autonomous-vehicle](<https://devfeed.tech/tags/autonomous-vehicle.md>), [autonomous-vehicle-safety](<https://devfeed.tech/tags/autonomous-vehicle-safety.md>), [ban](<https://devfeed.tech/tags/ban.md>), [california](<https://devfeed.tech/tags/california.md>), [car-tech](<https://devfeed.tech/tags/car-tech.md>), [construction](<https://devfeed.tech/tags/construction.md>), [dmv](<https://devfeed.tech/tags/dmv.md>), [highway](<https://devfeed.tech/tags/highway.md>), [regulations](<https://devfeed.tech/tags/regulations.md>), [tech-culture](<https://devfeed.tech/tags/tech-culture.md>), [waymo](<https://devfeed.tech/tags/waymo.md>)

### AI overview

After California ordered Waymo to stop operating and testing driverless vehicles on state freeways, citing an imminent public-safety hazard, the company removed its cars from highways nationwide. The order followed three vehicles entering a closed construction lane on Highway 101 and one attempting to pass police vehicles slowing traffic on Interstate 280. California requires Waymo to submit its next corrective fix for regulatory review before freeway operations resume; the order has no stated timeline. Waymo said it would review vehicle performance, but has not described proposed changes or a return date. The incidents exposed difficulty responding to temporary lane closures and police-directed traffic at freeway speeds. The restriction may hinder airport trips, which often use freeways, though surface-street routes remain possible and may be less convenient.

### Source excerpt

The DMV directive requires Waymo to stop both operating and testing its driverless vehicles on all California freeways, citing "an imminent hazard to public safety." The order will remain in effect until the company takes "effective corrective actions," the agency said. It did not provide a timeline. Read Entire Article

## India's Humyn Labs raises $47M to feed U.S. robotics companies the data they need

DevFeed: [India's Humyn Labs raises $47M to feed U.S. robotics companies the data they need](<https://devfeed.tech/articles/india-s-humyn-labs-raises-47m-to-feed-u-s-robotics-companies-the-data-they-need-140457.md>)

Original publisher: [Read original article](<https://techcrunch.com/2026/11/18/accel-backs-indias-humyn-labs-to-supply-training-data-needed-by-u-s-robotics-ai-developers/>)

Author: Jagmeet Singh

Published: 2026-11-18T11:30:00Z

Content type: news

Language: en

Sources: [TechCrunch | AI News & Artificial Intelligence](<https://devfeed.tech/sources/techcrunch-ai-news-artificial-intelligence.md>)

Topics: [Physical AI](<https://devfeed.tech/topics/physical-ai.md>), [Machine learning](<https://devfeed.tech/topics/machine-learning.md>), [recording](<https://devfeed.tech/topics/recording.md>), [Data pipelines](<https://devfeed.tech/topics/data-pipelines.md>)

Tags: [accel](<https://devfeed.tech/tags/accel.md>), [ai](<https://devfeed.tech/tags/ai.md>), [ai-labs](<https://devfeed.tech/tags/ai-labs.md>), [ai-models](<https://devfeed.tech/tags/ai-models.md>), [asia](<https://devfeed.tech/tags/asia.md>), [camera](<https://devfeed.tech/tags/camera.md>), [co-founder](<https://devfeed.tech/tags/co-founder.md>), [collect](<https://devfeed.tech/tags/collect.md>), [collecting](<https://devfeed.tech/tags/collecting.md>), [consent](<https://devfeed.tech/tags/consent.md>), [contracts](<https://devfeed.tech/tags/contracts.md>), [contributors](<https://devfeed.tech/tags/contributors.md>), [customers](<https://devfeed.tech/tags/customers.md>), [data](<https://devfeed.tech/tags/data.md>), [data-collection](<https://devfeed.tech/tags/data-collection.md>), [humyn-labs](<https://devfeed.tech/tags/humyn-labs.md>), [india](<https://devfeed.tech/tags/india.md>), [revenue](<https://devfeed.tech/tags/revenue.md>), [robotics](<https://devfeed.tech/tags/robotics.md>), [startups](<https://devfeed.tech/tags/startups.md>)

### AI overview

Humyn Labs raised $47 million to supply robotics and AI companies with real-world training data: $26 million in equity led by Accel and $21 million in venture debt led by Lighthouse Canton. Its two-camera hardware records people doing everyday tasks; software tracks hand movements and object interactions and labels footage for model training. The company collects data across more than 20 countries, using local partners to identify sites and supervise collection. It reports 12 paying customers and says about 90% of revenue comes from the U.S. The debt helps cover a cash-flow gap: workers and partners are paid daily or weekly, while customer payments typically arrive in 30 to 60 days. The excerpt notes privacy and consent concerns around recording people at home; Humyn says it records usual work or chores rather than staged activities.

### Source excerpt

Founded in February, Humyn Labs competes with startups including Mecka AI and XDOF, with about 90% of its revenue coming from the U.S.

## Edge AI Moves Agricultural Intelligence from the Cloud to the Field

DevFeed: [Edge AI Moves Agricultural Intelligence from the Cloud to the Field](<https://devfeed.tech/articles/edge-ai-moves-agricultural-intelligence-from-the-cloud-to-the-field-140174.md>)

Original publisher: [Read original article](<https://www.eetimes.com/edge-ai-moves-agricultural-intelligence-from-the-cloud-to-the-field/>)

Author: Pat Brans

Published: 2026-11-18T06:59:00Z

Content type: news

Language: en

Sources: [EE Times](<https://devfeed.tech/sources/ee-times.md>)

Topics: [Physical AI](<https://devfeed.tech/topics/physical-ai.md>), [Local AI](<https://devfeed.tech/topics/local-ai.md>), [On-device AI](<https://devfeed.tech/topics/on-device-ai.md>), [Artificial Intelligence](<https://devfeed.tech/topics/ai.md>), [Jetson](<https://devfeed.tech/topics/jetson.md>), [Nvidia](<https://devfeed.tech/topics/nvidia.md>)

Tags: [agriculture](<https://devfeed.tech/tags/agriculture.md>), [ai](<https://devfeed.tech/tags/ai.md>), [ai-assisted](<https://devfeed.tech/tags/ai-assisted.md>), [ai-inference](<https://devfeed.tech/tags/ai-inference.md>), [ai-systems](<https://devfeed.tech/tags/ai-systems.md>), [analog](<https://devfeed.tech/tags/analog.md>), [analysis](<https://devfeed.tech/tags/analysis.md>), [autonomous](<https://devfeed.tech/tags/autonomous.md>), [autonomous-driving](<https://devfeed.tech/tags/autonomous-driving.md>), [business](<https://devfeed.tech/tags/business.md>), [cloud](<https://devfeed.tech/tags/cloud.md>), [community](<https://devfeed.tech/tags/community.md>), [computing](<https://devfeed.tech/tags/computing.md>), [designline](<https://devfeed.tech/tags/designline.md>), [digital](<https://devfeed.tech/tags/digital.md>), [edge](<https://devfeed.tech/tags/edge.md>), [edge-ai](<https://devfeed.tech/tags/edge-ai.md>), [education](<https://devfeed.tech/tags/education.md>), [ee-times](<https://devfeed.tech/tags/ee-times.md>), [ee-times-university](<https://devfeed.tech/tags/ee-times-university.md>), [eelife](<https://devfeed.tech/tags/eelife.md>), [eelive-electronic-design](<https://devfeed.tech/tags/eelive-electronic-design.md>), [eet](<https://devfeed.tech/tags/eet.md>), [efficiency](<https://devfeed.tech/tags/efficiency.md>), [electrical-engineering](<https://devfeed.tech/tags/electrical-engineering.md>), [electronic-components](<https://devfeed.tech/tags/electronic-components.md>), [electronic-engineering-times](<https://devfeed.tech/tags/electronic-engineering-times.md>), [electronics](<https://devfeed.tech/tags/electronics.md>), [embedded-systems](<https://devfeed.tech/tags/embedded-systems.md>), [farm-autonomy](<https://devfeed.tech/tags/farm-autonomy.md>), [fundamentals-courses](<https://devfeed.tech/tags/fundamentals-courses.md>), [learning](<https://devfeed.tech/tags/learning.md>), [mixed-signal](<https://devfeed.tech/tags/mixed-signal.md>), [news](<https://devfeed.tech/tags/news.md>), [nvidia](<https://devfeed.tech/tags/nvidia.md>), [nxp](<https://devfeed.tech/tags/nxp.md>), [opinion](<https://devfeed.tech/tags/opinion.md>), [part-search](<https://devfeed.tech/tags/part-search.md>), [physical-ai](<https://devfeed.tech/tags/physical-ai.md>), [power](<https://devfeed.tech/tags/power.md>), [product-search](<https://devfeed.tech/tags/product-search.md>), [robotics](<https://devfeed.tech/tags/robotics.md>), [safety](<https://devfeed.tech/tags/safety.md>), [semiconductors](<https://devfeed.tech/tags/semiconductors.md>), [sensors](<https://devfeed.tech/tags/sensors.md>), [smart-agriculture](<https://devfeed.tech/tags/smart-agriculture.md>), [software](<https://devfeed.tech/tags/software.md>), [tech-papers](<https://devfeed.tech/tags/tech-papers.md>), [technology](<https://devfeed.tech/tags/technology.md>), [video](<https://devfeed.tech/tags/video.md>)

### AI overview

Agricultural autonomy requires onboard systems to sense, decide, and act within tight power, heat, cost, and safety limits, especially when connectivity is unreliable or response delays matter. Nvidia describes multi workload platforms combining perception, crop analysis, localization, navigation, obstacle avoidance, and control; Jetson targets embedded AI, while local sensors can handle some filtering and detection and cloud services remain useful for training, simulation, fleet management, and analytics. NXP emphasizes matching processors to tasks: application processors prepare sensor data, NPUs run inference, and MCUs handle control, with performance per watt often more important than peak compute. Safety functions must remain isolated so delayed inference cannot compromise stopping or other critical controls. Models may need location specific updates, but proofs of concept do not establish broad product readiness; deployment still requires safety, security, regulatory compliance, reliability, and protection of farmers' data privacy.

### Source excerpt

NXP and Nvidia tackle the challenges of delivering computing power, energy efficiency, and safety as agricultural machinery becomes autonomous. The post Edge AI Moves Agricultural Intelligence from the Cloud to the Field appeared first on EE Times.

## Manufacturing Differential Sparsity: Planner-Hardware Co-Design for Efficient World Model Planning

DevFeed: [Manufacturing Differential Sparsity: Planner-Hardware Co-Design for Efficient World Model Planning](<https://devfeed.tech/articles/manufacturing-differential-sparsity-planner-hardware-co-design-for-efficient-world-model-planning-140036.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.22483>)

Author: Wenbo Zhu, Paul Yi-Chia Chen, Yuanhan Li, Zishen Wan, Christopher Torng

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

Content type: article

Language: en

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

Topics: [World models](<https://devfeed.tech/topics/world-models.md>), [quantization](<https://devfeed.tech/topics/quantization.md>), [Physical AI](<https://devfeed.tech/topics/physical-ai.md>)

Tags: [co-design](<https://devfeed.tech/tags/co-design.md>), [cs-ai](<https://devfeed.tech/tags/cs-ai.md>), [cs-ar](<https://devfeed.tech/tags/cs-ar.md>), [cs-lg](<https://devfeed.tech/tags/cs-lg.md>), [cs-ro](<https://devfeed.tech/tags/cs-ro.md>), [cs-sy](<https://devfeed.tech/tags/cs-sy.md>), [eess-sy](<https://devfeed.tech/tags/eess-sy.md>), [hardware](<https://devfeed.tech/tags/hardware.md>), [planner](<https://devfeed.tech/tags/planner.md>), [planning](<https://devfeed.tech/tags/planning.md>), [sparsity](<https://devfeed.tech/tags/sparsity.md>), [world-model](<https://devfeed.tech/tags/world-model.md>)

### AI overview

Sampling-based MPC for JEPA world models can take tens of seconds per planning step because it repeats predictor inference thousands of times. This work co-designs a planner and accelerator around differential bit-level sparsity: reference forest sampling groups candidate action sequences around shared roots, creating exploitable similarity while retaining competitive planning quality, and the bit-exact hardware uses that sparsity for scheduling. Across three JEPA world models, the design is 1.28-1.80x faster and uses 38-52% less energy than a dense accelerator; against a prior bit-sparse design, it is 1.22-1.58x faster and uses 16-36% less energy at the same area. These results are limited to the three evaluated models.

### Source excerpt

arXiv:2611.22483v1 Announce Type: cross Abstract: JEPA-style latent world models let embodied agents plan toward new goals with model predictive control, but they are still too slow for real-life embodied control: sampling-based MPC repeats predictor inference tens of thousands of times per plan, and even on high-end GPUs one planning step can take tens of seconds. We observe substantial latent locality in this repeated-inference workload: the predictor maps close actions to close latents, and inside a quantized predictor this closeness takes a form that hardware can exploit, differential bit-level sparsity. Unlike an input stream, whose temporal similarity is given, a planner itself decides which candidates it evaluates together, so the planner can choose its own references and, with reference forest sampling, manufacture more sparsity: it couples the sampled action sequences of each group to a shared root, with competitive planning quality. We build the accelerator on the insights of the rich line of bit-level sparsity accelerators and design it and the planner tightly coupled as a single co-design: the planner increases sparsity and exposes structure for hardware scheduling, and the bit-exact hardware effectively uses the sparsity that the planner creates. On three JEPA world models, our co-design is 1.28-1.80x faster than a dense accelerator and uses 38-52% less energy, and it is 1.22-1.58x faster than prior bit-sparse design and uses 16-36% less energy, with the same area. To our knowledge, our planner is the first planner that manufactures differential bit-level sparsity across its candidates and exposes its structure for hardware scheduling, and our co-design is the first algorithm-hardware co-design for JEPA-style world-model planning.

## Guarded Swarms in the Loop: A ROS 2/PX4 Digital Twin for Hybrid Digital-Analog Coordination of UAV/UGV Swarms

DevFeed: [Guarded Swarms in the Loop: A ROS 2/PX4 Digital Twin for Hybrid Digital-Analog Coordination of UAV/UGV Swarms](<https://devfeed.tech/articles/guarded-swarms-in-the-loop-a-ros-2-px4-digital-twin-for-hybrid-digital-analog-coordination-of-uav-ugv-swarms-141498.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.22773>)

Author: Uwe M. Borghoff, Paolo Bottoni, Luca D'Aurizio, Remo Pareschi

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: [autonomous vehicles](<https://devfeed.tech/topics/autonomous-vehicles.md>), [Physical AI](<https://devfeed.tech/topics/physical-ai.md>), [Security](<https://devfeed.tech/topics/security.md>)

Tags: [2](<https://devfeed.tech/tags/2.md>), [abstract](<https://devfeed.tech/tags/abstract.md>), [all](<https://devfeed.tech/tags/all.md>), [also](<https://devfeed.tech/tags/also.md>), [analog](<https://devfeed.tech/tags/analog.md>), [architecture](<https://devfeed.tech/tags/architecture.md>), [arxiv](<https://devfeed.tech/tags/arxiv.md>), [attack](<https://devfeed.tech/tags/attack.md>), [attack-surface](<https://devfeed.tech/tags/attack-surface.md>), [authorization](<https://devfeed.tech/tags/authorization.md>), [autopilot](<https://devfeed.tech/tags/autopilot.md>), [based](<https://devfeed.tech/tags/based.md>), [board](<https://devfeed.tech/tags/board.md>), [bounded](<https://devfeed.tech/tags/bounded.md>), [cs-cr](<https://devfeed.tech/tags/cs-cr.md>), [cs-ma](<https://devfeed.tech/tags/cs-ma.md>), [digital](<https://devfeed.tech/tags/digital.md>), [digital-twin](<https://devfeed.tech/tags/digital-twin.md>), [hybrid](<https://devfeed.tech/tags/hybrid.md>), [in-the-loop](<https://devfeed.tech/tags/in-the-loop.md>), [ros](<https://devfeed.tech/tags/ros.md>), [ros-2](<https://devfeed.tech/tags/ros-2.md>)

### AI overview

The study implements Guarded Swarms as a ROS 2, PX4 software-in-the-loop, and Gazebo digital twin with two UAVs and one UGV. Its TB-CSPN coordination chain maps to ROS 2 nodes, while a per-vehicle Virtual Safety Board gates commands through an independent safety layer. Across 60 counterfactual runs, the three stated guarantees against missions based on conflicting, temporally incoherent, or unauthorized evidence held at three chain stages. Fault injection showed the safety veto prevented all collisions caused by forged missions or compromised workers only when enforced by an independent gate; a veto controlled by the compromised worker prevented none. The abstract also reports implementation hazards beyond the architecture specification and distinguishes properties testable in simulation from those requiring hardware.

### Source excerpt

arXiv:2611.22773v1 Announce Type: new Abstract: Swarms of unmanned aerial and ground vehicles (UAV/UGV) operating in contested environments cannot rely on software alone for physical safety, since the software stack and its networks are part of the attack surface. The Guarded Swarms architecture addresses this by pairing a digital coordination layer, built on Topic-Based Communication Space Petri Nets (TB-CSPN), with an independent analog safety layer that can veto unsafe actuation commands. So far, the architecture has been specified and argued for, but not executed. We present the first executable instantiation of Guarded Swarms as a digital twin built on ROS 2, the PX4 autopilot in software-in-the-loop mode, and Gazebo, with two UAVs and one UGV in a coastal counter-UAS scenario. The TB-CSPN transition chain, from multi-modal sensing through human authorization to vehicle tasking, is mapped one-to-one onto ROS 2 nodes, and the analog layer is modeled as a per-vehicle Virtual Safety Board that gates every command. Eight groups of experiments show that the three negative guarantees of the architecture (no mission from conflicting, temporally incoherent, or unauthorized evidence) hold at three distinct stages of the chain in all 60 counterfactual runs; that threat confirmation is bounded by the narrowest sensing modality; that the safety layer exercises its full modeled envelope independently of the coordination layer; and, under fault injection, that it prevents every collision caused by a forged mission or a compromised worker once its veto is enforced by an independent gate, whereas a veto left to the compromised worker prevents none. The twin also exposes implementation-level hazards that the architectural specification does not address, and it draws an explicit line between the properties that simulation can validate and those that only hardware can.

## Perception into Imagination: Executable Memory for World Models

DevFeed: [Perception into Imagination: Executable Memory for World Models](<https://devfeed.tech/articles/perception-into-imagination-executable-memory-for-world-models-140004.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.22294>)

Author: Xianyao Li, Ruitong Tian, Rui Min, Fang Xu, Jing Du

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

Content type: article

Language: en

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

Topics: [World models](<https://devfeed.tech/topics/world-models.md>), [Physical AI](<https://devfeed.tech/topics/physical-ai.md>), [robot sense of touch](<https://devfeed.tech/topics/robot-sense-of-touch.md>)

Tags: [actions](<https://devfeed.tech/tags/actions.md>), [control](<https://devfeed.tech/tags/control.md>), [cs-ai](<https://devfeed.tech/tags/cs-ai.md>), [cs-cv](<https://devfeed.tech/tags/cs-cv.md>), [cs-lg](<https://devfeed.tech/tags/cs-lg.md>), [cs-ro](<https://devfeed.tech/tags/cs-ro.md>), [imagination](<https://devfeed.tech/tags/imagination.md>), [learning](<https://devfeed.tech/tags/learning.md>), [memory](<https://devfeed.tech/tags/memory.md>), [model](<https://devfeed.tech/tags/model.md>), [motion](<https://devfeed.tech/tags/motion.md>), [neural](<https://devfeed.tech/tags/neural.md>), [physical-ai](<https://devfeed.tech/tags/physical-ai.md>), [policy](<https://devfeed.tech/tags/policy.md>), [prediction](<https://devfeed.tech/tags/prediction.md>), [recovery](<https://devfeed.tech/tags/recovery.md>), [retention](<https://devfeed.tech/tags/retention.md>), [reward](<https://devfeed.tech/tags/reward.md>), [state](<https://devfeed.tech/tags/state.md>), [training](<https://devfeed.tech/tags/training.md>), [value](<https://devfeed.tech/tags/value.md>), [visibility](<https://devfeed.tech/tags/visibility.md>), [world-model](<https://devfeed.tech/tags/world-model.md>), [world-models](<https://devfeed.tech/tags/world-models.md>)

### AI overview

Executable perceptual memory (EPM) carries explicit perceptual state forward alongside neural latent dynamics during imagined action, so remembered facts can inform reward prediction, value estimation, and policy learning. The method aligns reward inputs with the transitions that produced them and trains the reward head on open-loop rollouts of recorded actions. On Memory Maze 13x13, it reports 30.37 return after 1M interactions, versus 23.37 when the evolving read-out informs only policy and value on the same perceptual front end (gain 7.00 [3.80, 10.21]). The abstract also reports improved control on 9x9 and MiniWorld, and stronger reward anticipation for remembered out-of-view arrivals in both visual implementations. Tactile and concentration experiments are described as showing complementary roles for remembered facts; the excerpt gives no further methodological detail.

### Source excerpt

arXiv:2611.22294v1 Announce Type: cross Abstract: Physical agents must use what they previously saw or touched after visibility or contact is lost. Their actions change the implications of these remembered facts. We introduce executable perceptual memory (EPM), a world-model framework that advances explicit perceptual state alongside neural latent dynamics during imagination. Its evolving read-out informs reward prediction, value estimation and policy learning. We align reward inputs with the transitions that produced them and supervise the reward head on open-loop rollouts of recorded actions. On Memory Maze $13\times13$, EPM reaches 30.37 return at 1M interactions, compared with 23.37 when the same evolving read-out informs only policy and value on a common perceptual front end: a gain of 7.00 [3.80, 10.21]. The complete method also improves control on $9\times9$ and in MiniWorld. Probes of remembered, out-of-view arrivals show stronger reward anticipation in both visual implementations. Concentration, tactile recovery and real tactile frames demonstrate complementary roles of remembered facts in imagined-policy training, motion-dependent decisions and sensory-content retention. EPM connects persistent perception to the predicted consequences and learning signals of a world model for Physical AI.

## Auditing a Learned Simulator for Legged Robots: The Robotic World Model on ANYmal D

DevFeed: [Auditing a Learned Simulator for Legged Robots: The Robotic World Model on ANYmal D](<https://devfeed.tech/articles/auditing-a-learned-simulator-for-legged-robots-the-robotic-world-model-on-anymal-d-141845.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.23118>)

Author: Joyjeet Singh

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

Content type: article

Language: en

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

Topics: [World models](<https://devfeed.tech/topics/world-models.md>), [Robotics Simulation](<https://devfeed.tech/topics/robotics-simulation.md>), [audit](<https://devfeed.tech/topics/audit.md>), [Neural Network](<https://devfeed.tech/topics/neural-network.md>), [Physical AI](<https://devfeed.tech/topics/physical-ai.md>)

Tags: [anymal](<https://devfeed.tech/tags/anymal.md>), [auditing](<https://devfeed.tech/tags/auditing.md>), [cs-lg](<https://devfeed.tech/tags/cs-lg.md>), [cs-ro](<https://devfeed.tech/tags/cs-ro.md>), [legged-robots](<https://devfeed.tech/tags/legged-robots.md>), [model](<https://devfeed.tech/tags/model.md>), [mujoco](<https://devfeed.tech/tags/mujoco.md>), [policies](<https://devfeed.tech/tags/policies.md>), [policy](<https://devfeed.tech/tags/policy.md>), [pruning](<https://devfeed.tech/tags/pruning.md>), [simulator](<https://devfeed.tech/tags/simulator.md>), [training](<https://devfeed.tech/tags/training.md>), [uncertainty](<https://devfeed.tech/tags/uncertainty.md>), [variance](<https://devfeed.tech/tags/variance.md>), [world-model](<https://devfeed.tech/tags/world-model.md>)

### AI overview

The audit finds that one-step accuracy is a poor guide to long rollouts and policy performance for the ANYmal D Robotic World Model. Training on the model's own rollouts performed better at 368 steps across ten held-out episodes, though one-step training was better one step ahead; pruning 30% of weights retained 90% of one-step skill but 39% at two seconds. Ensemble disagreement tracked rollout error well (Spearman +0.90 at 100 steps), but its one-sigma coverage was only 4.6%, versus 68% for a calibrated estimate, because the networks' predicted variances collapsed under the implemented loss. Across 18 fine-tuned copies, all improved one- and eight-step predictions over the released model, yet 13 produced failing policies. Registered tests of whether the proposed checks rank simulators by policy performance remained inconclusive; running the authors' policy inside each simulator ranked them well in a secondary test. The audit covers one robot and gait, simulated data only, and no hardware.

### Source excerpt

arXiv:2611.23118v1 Announce Type: cross Abstract: A legged robot can learn to walk inside a learned simulator: a neural network that predicts the robot's next state and is run forward on its own predictions. They are usually trained and checked one step ahead, although a policy uses them for a hundred steps or more. We audit one published system end to end, without a GPU: the dynamics model of the Robotic World Model (RWM) for the ANYmal D quadruped, and the uncertainty estimate of its follow-up, RWM-U. We rebuilt the model to match the released code's outputs, losses and gradients exactly, measured 101 trained simulators, 200 compressed copies and 38 simulators of three other architectures, and trained policies inside simulators that we tested in a hand-built MuJoCo model of the robot. Training on the model's own rollouts beat one-step training at 368 steps on all ten held-out episodes (registered), although one-step training was better one step ahead. One-step checks therefore misjudge design choices: pruning 30% of the weights kept 90% of the model's skill one step ahead and 39% at 2 s. Ensemble disagreement ranked rollout error well (Spearman +0.90 at 100 steps), yet covered 4.6% of outcomes at one sigma, against 68% for a calibrated estimate, as each network's own variance collapses under the implemented loss. Two checks that need no recorded future ranked some pools of simulators well, but neither was shown better than 8-step error in a primary test. At the robot level, accuracy has not predicted walking: all 18 fine-tuned copies we tested predicted better than the released model one and eight steps ahead, yet 13 gave failing policies, and registered tests of whether any check ranks simulators by their policies were not settled. In a secondary test, running the authors' policy inside each simulator ranked them well. We also report nine of our own claims that we withdrew. Scope: one robot, one gait, simulated data, no hardware.

[Next page](<https://devfeed.tech/topics/physical-ai.md?cursor=WyIyMDI2LTExLTE4VDA1OjAwOjAwKzAwOjAwIiwgIjhmMzVlMjAyLTY3ZDktNGMxZC1hOWJiLWE1MzRlYjE5ODY2OSJd>)