# AI Models

AI models are programs trained on data to recognize patterns and make predictions or decisions, using algorithms and computing resources.

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

## How Reliable Are Explanations of Language Models That Detect Insider Threats?

DevFeed: [How Reliable Are Explanations of Language Models That Detect Insider Threats?](<https://devfeed.tech/articles/how-reliable-are-explanations-of-language-models-that-detect-insider-threats-145104.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.25796>)

Author: Suramya R. Angdembay

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

Content type: article

Language: en

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

Topics: [Artificial Intelligence](<https://devfeed.tech/topics/ai.md>), [AI Models](<https://devfeed.tech/topics/ai-models.md>), [AI in cybersecurity](<https://devfeed.tech/topics/ai-in-cybersecurity.md>), [LLM security](<https://devfeed.tech/topics/llm-security.md>)

Tags: [audit](<https://devfeed.tech/tags/audit.md>), [cs-lg](<https://devfeed.tech/tags/cs-lg.md>), [insider-threats](<https://devfeed.tech/tags/insider-threats.md>), [language-models](<https://devfeed.tech/tags/language-models.md>), [model](<https://devfeed.tech/tags/model.md>), [neural-networks](<https://devfeed.tech/tags/neural-networks.md>)

### AI overview

The study finds that explanations of language-model insider-threat detectors can be highly sensitive to how auditors define a control change. It introduces a "tipping radius" to estimate how far that metric must shift before the measured effect of removing a feature disappears. Across 33 of 34 runs of seven models, the radius was at most 18% of the distance between two standard metric choices; changes that size removed at least 58% of the effect or reversed it, while equally sized random changes largely did not. The finding replicated in 13 of 14 pre-declared primary runs on openly released models, although effects were often too small to measure on the authors' trained detectors. The study also reports that scores partly reflect worker identity and that named features depend on the comparison group. It recommends fixing the control metric in advance, reporting the tipping radius, and disclosing the comparison.

### Source excerpt

arXiv:2611.25796v1 Announce Type: new Abstract: Organizations monitor their workers' activity for insider threats, in which workers misuse their legitimate access, by copying confidential files before leaving. Detectors score how unusual each day of activity is, flagging the most unusual for an analyst to review. Prominent current approaches use opaque neural networks, so explanations of a flag point to internal features: patterns of neuron activity most active on days of misuse. We audit such explanations for language-model detectors trained on CERT's simulated workplace logs. To test an explanation, we see what happens to the detector's score when we remove its features inside the model. We compare this to a control: a change that is as similar as possible to the removal, but without actually removing the features. Because defining "similar" requires choosing a metric, our central contribution is the tipping radius: a first-order prediction of how far that metric must move before the difference between removal and control disappears, a measure of how fragile the result is to that choice. In 33 of 34 runs of seven language models, the tipping radius is at most 18% of the distance between two standard choices of metric. Moving that far removes at least 58% of the difference or reverses it, while random moves of the same size leave it largely intact. The result replicates in thirteen of fourteen pre-declared primary runs on openly released models, though on our trained detectors the difference is often too small to measure. We also find that a flag's score partly reflects who the worker is, not what they did, and that the features named depend on whom they are compared against. Audits should therefore fix the control's metric in advance, report its tipping radius, and disclose that comparison.

## World Models Favor Learned Priors over Visual Evidence

DevFeed: [World Models Favor Learned Priors over Visual Evidence](<https://devfeed.tech/articles/world-models-favor-learned-priors-over-visual-evidence-145665.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.26366>)

Author: Han Xiao, Zhiyong Bao, Xiaotong Liu, Yanhao Chen, Qingqiang Wu

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>), [AI Models](<https://devfeed.tech/topics/ai-models.md>)

Tags: [benchmark](<https://devfeed.tech/tags/benchmark.md>), [bias](<https://devfeed.tech/tags/bias.md>), [consistency](<https://devfeed.tech/tags/consistency.md>), [cs-ai](<https://devfeed.tech/tags/cs-ai.md>), [distribution](<https://devfeed.tech/tags/distribution.md>), [inference](<https://devfeed.tech/tags/inference.md>), [motion](<https://devfeed.tech/tags/motion.md>), [prediction](<https://devfeed.tech/tags/prediction.md>), [training](<https://devfeed.tech/tags/training.md>), [world-models](<https://devfeed.tech/tags/world-models.md>)

### AI overview

The abstract reports that world models can favor common physical rules over visual evidence when training data are imbalanced. TailWorld evaluates transfer of rules demonstrated in short calibration videos to a different initial state, varying training imbalance and calibration informativeness separately; it measures bias by comparing predictions with futures under the demonstrated and a common rule. The proposed RIFT method learns how rules affect object motion, infers the rule from calibration interactions, and controls the influence of training frequency. On TailWorld, RIFT improves prediction accuracy and physical consistency across imbalances and distribution shifts, including for rare dynamics.

### Source excerpt

arXiv:2611.26366v1 Announce Type: new Abstract: World models are increasingly used to simulate physical interactions, but training a separate model for each environment limits scalability. Cross-environment prediction requires identifying the current physical rule, which appearance alone may not reveal. We therefore study visual calibration: using object motion and interactions in a short video as direct evidence to infer the current rule and predict motion under that rule from a different initial state. However, differences in data availability across physical conditions can yield a long-tailed training distribution, biasing models toward common rules even when evidence supports rare ones. To study this challenge, we introduce TailWorld, which evaluates the transfer of demonstrated rules across initial states while independently controlling training imbalance and calibration informativeness. The benchmark compares predictions with two reference futures from the same prediction initial state, one following the demonstrated rule and the other a common rule, to measure prediction bias toward common dynamics. We further propose Rule Inference for Transfer (RIFT), which learns how different rules affect object motion, infers the current rule from calibration interactions, and applies it to prediction from a new initial state. RIFT explicitly controls the influence of training frequency on rule inference, allowing sufficiently informative observational evidence to support rare rules. Experiments on TailWorld show that RIFT improves prediction accuracy and physical consistency across varying training imbalances and distribution shifts, with gains in predicting rare dynamics.

## MM-SystemOneBias: Evaluating Bias in Multimodal Jev-Style System One Decision Models

DevFeed: [MM-SystemOneBias: Evaluating Bias in Multimodal Jev-Style System One Decision Models](<https://devfeed.tech/articles/mm-systemonebias-evaluating-bias-in-multimodal-jev-style-system-one-decision-models-145685.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.26557>)

Author: Jinzhe Li, Qi Wang, Dandan Guo, Li Liu, Yuan Wu, Yi Chang

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

Content type: article

Language: en

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

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

Tags: [bias](<https://devfeed.tech/tags/bias.md>), [cs-ai](<https://devfeed.tech/tags/cs-ai.md>), [datasets](<https://devfeed.tech/tags/datasets.md>), [decision-models](<https://devfeed.tech/tags/decision-models.md>), [evaluations](<https://devfeed.tech/tags/evaluations.md>), [multimodal](<https://devfeed.tech/tags/multimodal.md>), [semantic](<https://devfeed.tech/tags/semantic.md>), [stable](<https://devfeed.tech/tags/stable.md>), [tasks](<https://devfeed.tech/tags/tasks.md>)

### AI overview

MM-SystemOneBias evaluates 13 multimodal decision models on 5,323 instances across nine datasets and 11 task scenarios, comparing seven matched decision-model/backbone pairs under six input changes. Decision models were more stable in all seven pairs under changes to option labels, irrelevant information, and rating scales; six pairs also had fewer flips from option-order changes. They lost less accuracy under incorrect external judgments in all seven pairs and conflicting image-text descriptions in six. Conversely, six pairs gained less accuracy from correct external judgments and had less favorable accuracy changes under consistent descriptions. Even when semantic decisions stayed unchanged, answer probabilities shifted by a mean absolute 4.56-11.55 percentage points, showing that accuracy alone misses some input effects.

### Source excerpt

arXiv:2611.26557v1 Announce Type: new Abstract: When multimodal decision models determine actions, moderation outcomes, or quality judgments, their reliability depends on preserving semantic decisions across presentation changes, ignoring irrelevant additions, and distinguishing information consistent with task evidence from information that conflicts with it. Existing evaluations study decision-interface biases and multimodal model behavior in separate settings, leaving no common protocol to compare multimodal decision models with their backbones on practical tasks. We introduce \textbf{MM-SystemOneBias}, which evaluates 13 models on 5,323 original instances from nine public datasets and 11 task scenarios using six paired interventions: option order, option label, external judgment, image--text description, irrelevant information, and rating scale. Across seven matched decision-model--backbone pairs, decision models are more stable under option-label, irrelevant-information, and rating-scale changes in all seven pairs; six pairs also show lower option-order flip rates. They incur smaller accuracy losses under incorrect external judgments in all seven pairs and conflicting descriptions in six. In six pairs, they gain less accuracy from correct external judgments and show less favorable accuracy changes under consistent descriptions. Among pairs with unchanged semantic decisions, mean absolute changes in answer probabilities range from 4.56 to 11.55 percentage points, revealing input effects that accuracy alone misses.

## How I Built an Offline Voice Dictation App Using Small AI Models

DevFeed: [How I Built an Offline Voice Dictation App Using Small AI Models](<https://devfeed.tech/articles/how-i-built-an-offline-voice-dictation-app-using-small-ai-models-142895.md>)

Original publisher: [Read original article](<https://hackernoon.com/how-i-built-an-offline-voice-dictation-app-using-small-ai-models?source=rss>)

Author: Arun

Published: 2026-11-19T07:16:07Z

Content type: article

Language: en

Sources: [HackerNoon](<https://devfeed.tech/sources/hackernoon.md>)

Topics: [AI Models](<https://devfeed.tech/topics/ai-models.md>), [implementation](<https://devfeed.tech/topics/implementation.md>), [Open Source](<https://devfeed.tech/topics/open-source.md>)

Tags: [ai](<https://devfeed.tech/tags/ai.md>), [ai-models](<https://devfeed.tech/tags/ai-models.md>), [app](<https://devfeed.tech/tags/app.md>), [macos](<https://devfeed.tech/tags/macos.md>), [offline](<https://devfeed.tech/tags/offline.md>), [on-device-ai](<https://devfeed.tech/tags/on-device-ai.md>), [open-source](<https://devfeed.tech/tags/open-source.md>), [privacy](<https://devfeed.tech/tags/privacy.md>), [speech-recognition](<https://devfeed.tech/tags/speech-recognition.md>), [speech-to-text](<https://devfeed.tech/tags/speech-to-text.md>), [swift](<https://devfeed.tech/tags/swift.md>), [voice-dictation](<https://devfeed.tech/tags/voice-dictation.md>)

### AI overview

Purr is a free, open-source macOS dictation app that keeps speech recognition on the Mac and works offline after its models are downloaded. Hold Right Option to dictate into the focused field; Smart Typing transcribes while speaking, Voice Edit changes selected text, and Meetings mode records mic and Mac audio, labels speakers, and saves Markdown notes. English is the default; Whisper adds roughly 100 dictation languages and 28 meeting languages. Recognition uses Parakeet through FluidAudio or Whisper through WhisperKit with Core ML, plus CTC rescoring for vocabulary; meetings use a Core ML speaker diarizer and SpeexDSP. Voice Edit and summaries use Apple Foundation Models or Gemma 4 E2B via llama.cpp. The author describes fixes for meeting echo, unintended edits, and startup audio delay. It requires Apple Silicon and macOS 14 or later.

### Source excerpt

Purr is a free, open-source dictation app that does all its speech recognition on your Mac. Here's how it's built and what took the longest to fix.

## Tell Me How Sure You Are: Communicating Uncertainty in Auxiliary Evidence to Foundation Models

DevFeed: [Tell Me How Sure You Are: Communicating Uncertainty in Auxiliary Evidence to Foundation Models](<https://devfeed.tech/articles/tell-me-how-sure-you-are-communicating-uncertainty-in-auxiliary-evidence-to-foundation-models-143512.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.24351>)

Author: Soumyajit Chatterjee, Baturalp Buyukates, Abhirup Ghosh

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

Content type: article

Language: en

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

Topics: [AI Foundation Models](<https://devfeed.tech/topics/ai-foundation-models.md>), [Uncertainty quantification LLMs](<https://devfeed.tech/topics/uncertainty-quantification-llms.md>), [AI Models](<https://devfeed.tech/topics/ai-models.md>)

Tags: [communication](<https://devfeed.tech/tags/communication.md>), [context](<https://devfeed.tech/tags/context.md>), [cs-lg](<https://devfeed.tech/tags/cs-lg.md>), [datasets](<https://devfeed.tech/tags/datasets.md>), [decision-making](<https://devfeed.tech/tags/decision-making.md>), [f1](<https://devfeed.tech/tags/f1.md>), [foundation-models](<https://devfeed.tech/tags/foundation-models.md>), [generative](<https://devfeed.tech/tags/generative.md>), [humans](<https://devfeed.tech/tags/humans.md>), [image-classification](<https://devfeed.tech/tags/image-classification.md>), [improvements](<https://devfeed.tech/tags/improvements.md>), [inference](<https://devfeed.tech/tags/inference.md>), [llm](<https://devfeed.tech/tags/llm.md>), [uncertainty](<https://devfeed.tech/tags/uncertainty.md>)

### AI overview

The study finds that including confidence in auxiliary evidence can improve zero-shot foundation-model image classification. Across five downstream models and three datasets, communicating uncertainty from human-observed or learned-model-predicted text characteristics improved LLM macro-F1 by as much as 26.18 percentage points compared with omitting uncertainty. The authors also report that uncertainty carries predictive information about whether downstream decisions are correct. These findings are limited to the evaluated image-classification setup.

### Source excerpt

arXiv:2611.24351v1 Announce Type: new Abstract: In many queries, the context information provided to generative foundation models, such as LLMs and VLMs, is inherently uncertain, including its representations of the physical world. While conventional prompting discards the associated uncertainty, this paper, for the first time in the literature, systematically demonstrates that communicating such uncertainty in context provides useful additional cues for downstream decisions and yields improvements across a wide range of settings. We formulate the decision-making problem as a zero-shot foundation model-assisted image classification task supported by auxiliary text-based characteristics. We choose this due to the availability of well-defined auxiliary properties, associated uncertainty, and unambiguous task correctness. Using five downstream foundation models across three datasets, we extensively evaluate uncertainty arising from both humans and learned models when observing or predicting auxiliary evidence. Across all settings, we find that communicating uncertainty substantially improves downstream LLM classification performance: macro-F1 improves up to $26.18$ percentage points compared to when the uncertainty information is not communicated. We further observe how the foundation models process such uncertainty information and find that auxiliary uncertainty contains predictive information about the correctness of downstream decisions. Together, these results establish uncertainty as useful contextual information and motivate more systematic consideration of uncertainty communication in future foundation-model inference pipelines.

## How Well Do Generalist CT Foundation Models Differentiate Liver Tumors?

DevFeed: [How Well Do Generalist CT Foundation Models Differentiate Liver Tumors?](<https://devfeed.tech/articles/how-well-do-generalist-ct-foundation-models-differentiate-liver-tumors-142757.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.24122>)

Author: Yosuke Yamagishi, Tomohiro Kikuchi, Shouhei Hanaoka, Takeharu Yoshikawa, Osamu Abe

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

Content type: article

Language: en

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

Topics: [AI Foundation Models](<https://devfeed.tech/topics/ai-foundation-models.md>), [AI Models](<https://devfeed.tech/topics/ai-models.md>), [enhancing AI model accuracy](<https://devfeed.tech/topics/enhancing-ai-model-accuracy.md>)

Tags: [cs-ai](<https://devfeed.tech/tags/cs-ai.md>), [cs-cv](<https://devfeed.tech/tags/cs-cv.md>), [foundation-models](<https://devfeed.tech/tags/foundation-models.md>), [generalist](<https://devfeed.tech/tags/generalist.md>), [linear-probing](<https://devfeed.tech/tags/linear-probing.md>)

### AI overview

In this retrospective study of 517 cases from a public single-institution multiphase CT dataset, RADAR and Merlin had limited ability to differentiate four liver tumor classes zero-shot. At the portal venous phase, RADAR reached 54.4% accuracy and labeled 31% of malignant tumors as hemangioma. Training a linear classifier on frozen four-phase RADAR features raised accuracy to 79.9% and reduced those hemangioma predictions to 1%. Merlin achieved 60.9% accuracy. These findings suggest linear probing substantially improved performance in this dataset; they do not establish performance beyond the study setting.

### Source excerpt

arXiv:2611.24122v1 Announce Type: cross Abstract: Generalist CT foundation models detect many findings with high performance, but their ability to differentiate detected lesions remains largely unevaluated. This retrospective study evaluated two generalist models, RADAR and Merlin, for differentiating liver tumors using a public single-institution multiphase CT dataset (517 cases: hepatocellular carcinoma, intrahepatic cholangiocarcinoma, breast and colorectal cancer liver metastases, and hemangioma). The models were evaluated in a zero-shot setting using their existing outputs and with linear probing, in which only a linear classifier was trained on frozen image features (stratified fivefold cross-validation repeated five times). In the zero-shot setting, RADAR achieved an accuracy of 54.4% (95% CI: 49.9, 58.8) and a $\kappa$ of 0.37 for four-class differentiation at the portal venous phase and predicted 31% (129 of 421) of malignant tumors as hemangioma. With linear probing of four-phase features, its accuracy improved to 79.9% (95% CI: 76.6, 83.4) with a $\kappa$ of 0.72, and 1% (five of 421) of malignant tumors were predicted as hemangioma; Merlin achieved an accuracy of 60.9% (95% CI: 56.9, 65.0). Generalist CT foundation models showed limited zero-shot ability to differentiate liver tumors, which improved substantially with linear probing alone.

## Measuring What Matters: Synthetic Benchmarks for Concept Bottleneck Models

DevFeed: [Measuring What Matters: Synthetic Benchmarks for Concept Bottleneck Models](<https://devfeed.tech/articles/measuring-what-matters-synthetic-benchmarks-for-concept-bottleneck-models-142528.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2606.04326>)

Author: Julian Skirzynski, Harry Cheon, Shreyas Kadekodi, Meredith Stewart, Berk Ustun

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: [Ground truth / benchmark quality](<https://devfeed.tech/topics/ground-truth-benchmark-quality.md>), [Model Development](<https://devfeed.tech/topics/model-development.md>), [AI Models](<https://devfeed.tech/topics/ai-models.md>), [human feedback](<https://devfeed.tech/topics/human-feedback.md>)

Tags: [annotation](<https://devfeed.tech/tags/annotation.md>), [architecture](<https://devfeed.tech/tags/architecture.md>), [automation](<https://devfeed.tech/tags/automation.md>), [benchmarks](<https://devfeed.tech/tags/benchmarks.md>), [bottleneck](<https://devfeed.tech/tags/bottleneck.md>), [concepts](<https://devfeed.tech/tags/concepts.md>), [cs-ai](<https://devfeed.tech/tags/cs-ai.md>), [cs-lg](<https://devfeed.tech/tags/cs-lg.md>), [data](<https://devfeed.tech/tags/data.md>), [datasets](<https://devfeed.tech/tags/datasets.md>), [decision](<https://devfeed.tech/tags/decision.md>), [experts](<https://devfeed.tech/tags/experts.md>), [failure](<https://devfeed.tech/tags/failure.md>), [failures](<https://devfeed.tech/tags/failures.md>), [interpretability](<https://devfeed.tech/tags/interpretability.md>), [models](<https://devfeed.tech/tags/models.md>), [performance](<https://devfeed.tech/tags/performance.md>), [quality](<https://devfeed.tech/tags/quality.md>), [real-world](<https://devfeed.tech/tags/real-world.md>), [supervision](<https://devfeed.tech/tags/supervision.md>), [support](<https://devfeed.tech/tags/support.md>), [synthetic](<https://devfeed.tech/tags/synthetic.md>)

### AI overview

The paper introduces synthetic benchmarks for concept bottleneck models, addressing the shortage of real-world datasets with concept labels. Researchers can generate labeled data and vary modality, concept selection, and annotation quality to study which problems suit these models and isolate causes of failure. The authors demonstrate evaluation for decision support, where models help human experts, and automation of routine tasks without supervision. They report that the benchmarks reveal failure modes and help attribute them to model architecture or data.

### Source excerpt

arXiv:2606.04326v2 Announce Type: replace-cross Abstract: Concept bottleneck models predict using high-level concepts detected in inputs. Although these models provide a simple way to reap benefits from interpretability, they are hard to train and evaluate because very few real-world datasets include concept labels. In this paper, we develop synthetic benchmarks for concept bottleneck models. With our benchmarks, researchers can generate labeled datasets and control properties that affect performance, such as data modality, the choice of concepts, or annotation quality. They can thus find problems that suit these models and isolate the factors behind their failures. We demonstrate how our benchmarks can be used to evaluate major use cases for concept bottleneck models: decision support, i.e., helping human experts make better decisions; and automation, i.e., handling routine tasks without supervision. Our demonstrations show how the benchmarks can identify new failure modes and attribute them to either the architecture or the data.

## Counterfactual World Models: JEPAs That Reason About "What If"

DevFeed: [Counterfactual World Models: JEPAs That Reason About "What If"](<https://devfeed.tech/articles/counterfactual-world-models-jepas-that-reason-about-what-if-142830.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.24778>)

Author: Randall Balestriero, Yann LeCun, Kerda Varaku

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>), [Machine learning](<https://devfeed.tech/topics/machine-learning.md>), [AI Models](<https://devfeed.tech/topics/ai-models.md>)

Tags: [actions](<https://devfeed.tech/tags/actions.md>), [autoencoder](<https://devfeed.tech/tags/autoencoder.md>), [clip](<https://devfeed.tech/tags/clip.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>), [minimax](<https://devfeed.tech/tags/minimax.md>), [pixels](<https://devfeed.tech/tags/pixels.md>), [planner](<https://devfeed.tech/tags/planner.md>), [planning](<https://devfeed.tech/tags/planning.md>), [simulator](<https://devfeed.tech/tags/simulator.md>), [stat-ml](<https://devfeed.tech/tags/stat-ml.md>), [vehicle](<https://devfeed.tech/tags/vehicle.md>), [world-model](<https://devfeed.tech/tags/world-model.md>), [world-models](<https://devfeed.tech/tags/world-models.md>)

### AI overview

The proposed counterfactual world model estimates what an alternative action would have done under the same hidden disturbance, without counterfactual training data, added parameters, or retraining. It combines the observed outcome with the difference between the existing model's predictions for the alternative and factual actions. The result relies on actions and disturbances acting as separate forces in smooth dynamics; the authors identify discontinuities, coordinate folds, and disturbances that scale commands as failure cases. They prove an O(h²) error bound relative to disturbance variance and a matching minimax lower bound for unpaired factual data. In reported image-based JEPA experiments, counterfactual R² rose from 38-46% to 86-96% on four rendered simulators; an underwater vehicle reached 90-94% with sixteen times more factual data. Hindsight planning reached the goal in 35-69% of events, versus 14-23% for an oracle planner that knew the simulator but not the outcome.

### Source excerpt

arXiv:2611.24778v1 Announce Type: cross Abstract: World models operate in a world they cannot fully observe. When a hidden disturbance, such as a current pushing a boat, changes the outcome of an action, a counterfactual question follows: what would have happened under a different action, subject to the same disturbance? Its answer separates bad decisions from bad luck. We prove that world models trained on factual transitions cannot answer it on their own, whatever their data or capacity. For many physical systems, the observed outcome and a simple property of the dynamics suffice: natura non facit saltus, nature makes no leaps. When actions and disturbances act as separate forces, their interaction is second-order in time, so the counterfactual outcome is, to first order, the observed outcome plus the difference between the world model's predictions under the alternative and factual actions. This yields a counterfactual world model with no counterfactual data, extra parameters, or retraining. For transitions of duration $h$, we prove an $O(h^2)$ bound on the counterfactual error relative to the disturbance variance, establish a matching minimax lower bound from unpaired factual data, and characterize failures at discontinuities, coordinate folds, and disturbances that scale the command. We validate these predictions from pixels with JEPA world models on frozen DINOv2 and masked autoencoder (MAE) encoders. On four rendered simulators, our approach improves counterfactual $R^2$ from $38$--$46\%$ to $86$--$96\%$ with the same world models; on a fifth, an underwater vehicle, it reaches $90$--$94\%$ with sixteen times more factual data. Four further frozen encoders, DINOv3, SigLIP~2, CLIP and EVA-02, give $81$--$95\%$ on the same simulators. In hindsight planning, it finds the action that would have reached the goal in $35$--$69\%$ of events, against $14$--$23\%$ for an oracle planner that knows the simulator but not the outcome.

## Do 3D Point Representations Encode Architectural Functions? Probing Supervised and Pretrained Models on Urban Facades

DevFeed: [Do 3D Point Representations Encode Architectural Functions? Probing Supervised and Pretrained Models on Urban Facades](<https://devfeed.tech/articles/do-3d-point-representations-encode-architectural-functions-probing-supervised-and-pretrained-models-on-urban-facades-143438.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.23720>)

Author: Abdiaziz Muse, Jiju Poovvancheri

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

Content type: article

Language: en

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

Topics: [Representation learning](<https://devfeed.tech/topics/representation-learning.md>), [AI Models](<https://devfeed.tech/topics/ai-models.md>), [AI Foundation Models](<https://devfeed.tech/topics/ai-foundation-models.md>), [3D](<https://devfeed.tech/topics/3d.md>), [Feature Engineering](<https://devfeed.tech/topics/feature-engineering.md>)

Tags: [3d](<https://devfeed.tech/tags/3d.md>), [building](<https://devfeed.tech/tags/building.md>), [cs-cv](<https://devfeed.tech/tags/cs-cv.md>), [cs-lg](<https://devfeed.tech/tags/cs-lg.md>), [elements](<https://devfeed.tech/tags/elements.md>), [features](<https://devfeed.tech/tags/features.md>), [fine-tuning](<https://devfeed.tech/tags/fine-tuning.md>), [function](<https://devfeed.tech/tags/function.md>), [functional](<https://devfeed.tech/tags/functional.md>), [functions](<https://devfeed.tech/tags/functions.md>), [lidar](<https://devfeed.tech/tags/lidar.md>), [linear](<https://devfeed.tech/tags/linear.md>), [models](<https://devfeed.tech/tags/models.md>), [point](<https://devfeed.tech/tags/point.md>), [pretraining](<https://devfeed.tech/tags/pretraining.md>), [transformers](<https://devfeed.tech/tags/transformers.md>)

### AI overview

On 22 mobile-LiDAR buildings labeled with 15 functional attributes per element, frozen self-supervised Utonia features reached 0.732 mAP on functionally ambiguous elements, compared with 0.610 for hand-crafted descriptors. Networks trained from scratch, including models up to 108M parameters, did not clearly surpass the descriptors. Random-weight controls scored 0.649 and 0.652; Utonia's reported pretraining gain was +0.080, with a 95% building interval of [+0.035, +0.124]. Pretrained features improved as labeled buildings increased, while descriptors and random features plateaued. Fine-tuning eroded the benefit in the tested configurations, suggesting that at this data scale function is better read from pretrained representations than learned from scratch.

### Source excerpt

arXiv:2611.23720v1 Announce Type: cross Abstract: Semantic labels say what a building element is, not what it does: two identical-looking windows on one facade may be operable or fixed, a column may carry load or only decorate. We ask whether 3D point representations encode such architectural function. On 22 mobile-LiDAR buildings with 15 functional attributes per element, we hold out entire buildings and compare hand-crafted descriptors, networks trained from scratch and frozen self-supervised point transformers. Descriptors and frozen features are read by the same linear layer. No network trained from scratch, from PointNet to a 108M-parameter PTv3, clearly exceeds the descriptors (0.610 mAP on functionally ambiguous elements), whereas frozen Utonia reaches 0.732. Random-weight controls of the same architectures (0.649 and 0.652) attribute the gain to pretraining (Utonia: +0.080, 95% building interval [+0.035, +0.124]). Pretrained features improve at every tested number of labelled buildings while descriptors and random features plateau, and, in the configurations we tested, fine-tuning erodes the benefit: at today's data scale, function is better read from pretrained representations than learned from scratch.

## Language Models Make New Research Sound Old

DevFeed: [Language Models Make New Research Sound Old](<https://devfeed.tech/articles/language-models-make-new-research-sound-old-142669.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.23546>)

Author: Guy Zyskind

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

Content type: article

Language: en

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

Topics: [LLM evaluation / benchmarking](<https://devfeed.tech/topics/llm-evaluation-benchmarking.md>), [AI Models](<https://devfeed.tech/topics/ai-models.md>), [gpt-6-astra](<https://devfeed.tech/topics/gpt-6-astra.md>)

Tags: [cs-ai](<https://devfeed.tech/tags/cs-ai.md>), [cs-cl](<https://devfeed.tech/tags/cs-cl.md>), [language-models](<https://devfeed.tech/tags/language-models.md>), [large-language-models-llms](<https://devfeed.tech/tags/large-language-models-llms.md>), [llms](<https://devfeed.tech/tags/llms.md>), [research](<https://devfeed.tech/tags/research.md>)

### AI overview

LLMs often recognize that newly raised research questions are unstudied yet fail to say so in ordinary answers. Across 40 questions from newly accepted papers, GPT-6-Astra judged 98% unstudied but plainly reported that status in only 8%; all six frontier systems had reporting gaps of 30-90 percentage points. The gap is asymmetric: when given papers that answer a question, models identified it as studied in 88-100% of ordinary answers. Separately checking cited papers and fixing the response to reflect those checks raised Astra's clear reporting of open status to 95%, with gains across all six systems. The authors also predict reviewers' novelty grades from multiple models' written ratings and a rule learned from graded examples, reaching the reported performance range of Think-Probe-Respond without requiring model weights or internal activations. These results concern the reported study and do not establish performance beyond its described evaluation.

### Source excerpt

arXiv:2611.23546v1 Announce Type: cross Abstract: Researchers are increasingly using large language models (LLMs) to validate research questions. We find that LLMs make open research questions sound partly studied and overstate what related papers establish. Models recognize most of these overstatements when asked to check the papers' full texts separately. On 40 questions drawn from newly accepted papers, GPT-6-Astra correctly judges 98% to be unstudied, but states that status plainly in only 8% of ordinary answers. All six frontier systems show reporting gaps of 30-90 percentage points. Across 22 models given identical papers, higher capability is associated with more accurate judgments and larger reporting gaps. This reporting failure is asymmetric: models correctly identify studied questions in 88-100% of ordinary answers when we supply the answering papers. To correct this, we build a fixed answer from each model's judgment and separate checks of cited papers. This raises Astra's clear reporting of open status from 8% to 95%, with improvements in all six systems. Finally, we predict reviewers' novelty grades by combining several models' written ratings with a rule learned from graded examples, reaching the performance range reported by Think-Probe-Respond. Unlike TPR, our approach needs no access to model weights or internal activations, so it also works with closed-weight frontier models.

## Diffusion Models are Provably Efficient Priors for Sparse Recovery

DevFeed: [Diffusion Models are Provably Efficient Priors for Sparse Recovery](<https://devfeed.tech/articles/diffusion-models-are-provably-efficient-priors-for-sparse-recovery-143536.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.24602>)

Author: Ivan Lee, Xingyu Xu, Yuejie Chi

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

Content type: article

Language: en

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

Topics: [AI Models](<https://devfeed.tech/topics/ai-models.md>), [Algorithms](<https://devfeed.tech/topics/algorithms.md>)

Tags: [cs-it](<https://devfeed.tech/tags/cs-it.md>), [cs-lg](<https://devfeed.tech/tags/cs-lg.md>), [diffusion](<https://devfeed.tech/tags/diffusion.md>), [diffusion-models](<https://devfeed.tech/tags/diffusion-models.md>), [guarantees](<https://devfeed.tech/tags/guarantees.md>), [math-it](<https://devfeed.tech/tags/math-it.md>), [math-oc](<https://devfeed.tech/tags/math-oc.md>), [recovery](<https://devfeed.tech/tags/recovery.md>), [sampling](<https://devfeed.tech/tags/sampling.md>), [sensing](<https://devfeed.tech/tags/sensing.md>), [stat-ml](<https://devfeed.tech/tags/stat-ml.md>), [theoretical](<https://devfeed.tech/tags/theoretical.md>)

### AI overview

Given an accurate diffusion model of a sparsity-promoting Bernoulli-Gaussian prior, a likelihood-guided annealed Langevin sampler recovers sparse signals from few linear measurements in polynomial time, with near-optimal sample complexity across a broad family of random and deterministic sensing matrices. If the signal is drawn from the assumed prior, the paper also proves total variation guarantees for posterior sampling after a final denoising step.

### Source excerpt

arXiv:2611.24602v1 Announce Type: cross Abstract: Diffusion models have transformed generative modeling through their ability to capture and generate samples from complex data distributions. This success has spurred substantial interest in using pretrained diffusion models as expressive priors for highly ill-posed inverse problems in imaging and sensing. Despite strong empirical performance, theoretical understanding of this approach remains limited: existing guarantees are often asymptotic or entail prohibitive computational costs in the worst case. In this paper, we address this gap by studying sparse recovery from a small number of linear measurements, a classical problem in compressed sensing. We show that, given an accurate diffusion model capturing a sparsity-promoting Bernoulli-Gaussian prior, a likelihood-guided annealed Langevin sampler provably recovers the sparse signal in polynomial time with near-optimal sample complexity for a broad family of random and deterministic sensing matrices. When the signal is indeed drawn from the postulated prior, we further establish total variation guarantees for posterior sampling after a final denoising step. These results lay a theoretical foundation for statistically and computationally efficient inverse solvers with diffusion priors.

## Building Predictive AI Models When Synthetic Data Is All You Have

DevFeed: [Building Predictive AI Models When Synthetic Data Is All You Have](<https://devfeed.tech/articles/building-predictive-ai-models-when-synthetic-data-is-all-you-have-144066.md>)

Original publisher: [Read original article](<https://engineering.salesforce.com/building-predictive-ai-models-when-synthetic-data-is-all-you-have/>)

Author: Taniya Ghosh

Published: 2026-11-19T00:59:15Z

Content type: article

Language: en

Sources: [Salesforce](<https://devfeed.tech/sources/salesforce-engineering-blog.md>)

Topics: [Predictive AI](<https://devfeed.tech/topics/predictive-ai.md>), [Machine learning](<https://devfeed.tech/topics/machine-learning.md>), [Advanced AI workflows](<https://devfeed.tech/topics/advanced-ai-workflows.md>), [data-engineering](<https://devfeed.tech/topics/data-engineering.md>), [AI Models](<https://devfeed.tech/topics/ai-models.md>), [Feature Engineering](<https://devfeed.tech/topics/feature-engineering.md>), [Open Source Models & Datasets](<https://devfeed.tech/topics/open-source-models-datasets.md>)

Tags: [ai](<https://devfeed.tech/tags/ai.md>), [ai-models](<https://devfeed.tech/tags/ai-models.md>), [artificial-intelligence](<https://devfeed.tech/tags/artificial-intelligence.md>), [customer-data](<https://devfeed.tech/tags/customer-data.md>), [data](<https://devfeed.tech/tags/data.md>), [data-360](<https://devfeed.tech/tags/data-360.md>), [predictive-ai](<https://devfeed.tech/tags/predictive-ai.md>), [synthetic-data](<https://devfeed.tech/tags/synthetic-data.md>), [training-data](<https://devfeed.tech/tags/training-data.md>)

### AI overview

Salesforce's team built reusable IT service management prediction templates before real customer incidents were available for training. Product-manager scenarios, varied within guardrails and expanded with open-source LLMs, supplied synthetic data for testing. To avoid being misled by aggregate accuracy on rare SLA breaches, the team checked whether scores changed as expected when one risk factor changed and whether the top three predictors made sense to service managers; these scenario tests did not establish production accuracy. Data 360 transforms customer incident histories into training data, Einstein on Data 360 trains an XGBoost model, and a scoring transform writes predictions for display on incident records. Templates carry the data contract and model structure, while each customer's history supplies its own learned parameters. The excerpt says training requires about 400 completed incidents and scoring runs in scheduled batches about hourly; thin or poor-quality histories may reduce reliability, and immediate scoring remained a need. At interview time, the team had received no reports of the anticipated challenges becoming production problems and was awaiting more user反馈.

### Source excerpt

In our Engineering Energizers Q&A series, we highlight the engineering minds driving innovation across Salesforce. Today, we spotlight Taniya Ghosh, Senior Manager of Software Engineering, whose team faced a brutal engineering contradiction: build predictive AI models that warn of escalating incidents before they had real customer incidents to learn from. Synthetic training data gave them [...] The post Building Predictive AI Models When Synthetic Data Is All You Have appeared first on Salesforce Engineering Blog.

## Tesorai raises $5.3M to feed starving AI models with biological data and accelerate drug discovery

DevFeed: [Tesorai raises $5.3M to feed starving AI models with biological data and accelerate drug discovery](<https://devfeed.tech/articles/tesorai-raises-5-3m-to-feed-starving-ai-models-with-biological-data-and-accelerate-drug-discovery-140710.md>)

Original publisher: [Read original article](<https://siliconangle.com/2026/11/18/tesorai-raises-5-3m-to-feed-starving-ai-models-with-biological-data-and-accelerate-drug-discovery/>)

Author: Mike Wheatley

Published: 2026-11-18T14:00:33Z

Content type: news

Language: en

Sources: [SiliconANGLE](<https://devfeed.tech/sources/siliconangle.md>)

Topics: [AI for science](<https://devfeed.tech/topics/ai-for-science.md>), [Bioinformatics](<https://devfeed.tech/topics/bioinformatics.md>), [AI Models](<https://devfeed.tech/topics/ai-models.md>), [AI Chat](<https://devfeed.tech/topics/ai-chat.md>)

Tags: [ai](<https://devfeed.tech/tags/ai.md>), [ai-drug-discovery](<https://devfeed.tech/tags/ai-drug-discovery.md>), [ai-insights](<https://devfeed.tech/tags/ai-insights.md>), [ai-model](<https://devfeed.tech/tags/ai-model.md>), [ai-model-training](<https://devfeed.tech/tags/ai-model-training.md>), [ai-models](<https://devfeed.tech/tags/ai-models.md>), [ai-research](<https://devfeed.tech/tags/ai-research.md>), [artificial-intelligence](<https://devfeed.tech/tags/artificial-intelligence.md>), [big-data](<https://devfeed.tech/tags/big-data.md>), [biological-data](<https://devfeed.tech/tags/biological-data.md>), [clinical-studies](<https://devfeed.tech/tags/clinical-studies.md>), [data](<https://devfeed.tech/tags/data.md>), [drug-discovery](<https://devfeed.tech/tags/drug-discovery.md>), [drug-research](<https://devfeed.tech/tags/drug-research.md>), [funding](<https://devfeed.tech/tags/funding.md>), [genomic-data](<https://devfeed.tech/tags/genomic-data.md>), [melissa-miao](<https://devfeed.tech/tags/melissa-miao.md>), [news](<https://devfeed.tech/tags/news.md>), [peter-cimermancic](<https://devfeed.tech/tags/peter-cimermancic.md>), [protein-data](<https://devfeed.tech/tags/protein-data.md>), [scientific-studies](<https://devfeed.tech/tags/scientific-studies.md>), [seed-funding](<https://devfeed.tech/tags/seed-funding.md>), [startup](<https://devfeed.tech/tags/startup.md>), [tesorai](<https://devfeed.tech/tags/tesorai.md>), [the-latest](<https://devfeed.tech/tags/the-latest.md>)

### AI overview

Tesorai raised $5.3 million in seed funding to make existing biological data more useful for AI-driven drug discovery; total funding is now $8 million. Its platform analyzes raw lab-instrument data alongside historical studies: Search ingests and analyzes large datasets, Chat lets scientists query them in plain English, and Trove catalogs thousands of public datasets and studies. The company says reanalysis of five Alzheimer's brain studies surfaced p-tau217, and that adding protein data helped a small single-cell AI model outperform larger RNA-only models on multiple benchmarks. It has partnerships with MD Anderson, NYU and the Michael J. Fox Foundation. Funding will support additional life-sciences partnerships, a larger public-study catalog and coverage of molecular data beyond proteins and genomes.

### Source excerpt

Drug researchers and medical laboratories generate massive volumes of biological data every single day, yet somehow the artificial intelligence models designed to speed up drug discovery are still starved of useful information. It's a strange paradox, but a startup called Tesorai Inc. believes it has the solution after raising $5.3 million in seed funding today. [...] The post Tesorai raises $5.3M to feed starving AI models with biological data and accelerate drug discovery appeared first on SiliconANGLE.

## How to delete data from a trained AI model (when DELETE doesn't work)

DevFeed: [How to delete data from a trained AI model (when DELETE doesn't work)](<https://devfeed.tech/articles/how-to-delete-data-from-a-trained-ai-model-when-delete-doesn-t-work-141074.md>)

Original publisher: [Read original article](<https://www.red-gate.com/simple-talk/ai/delete-data-from-trained-ai-model/>)

Author: Dejan Lukić

Published: 2026-11-18T12:00:00Z

Content type: article

Language: en

Sources: [Simple Talk](<https://devfeed.tech/sources/simple-talk.md>)

Topics: [AI Models](<https://devfeed.tech/topics/ai-models.md>), [Training AI Models](<https://devfeed.tech/topics/training-ai-models.md>), [Large Language Model](<https://devfeed.tech/topics/llm.md>), [Machine learning](<https://devfeed.tech/topics/machine-learning.md>), [Fine-tuning](<https://devfeed.tech/topics/fine-tuning.md>)

Tags: [ai](<https://devfeed.tech/tags/ai.md>), [ai-model](<https://devfeed.tech/tags/ai-model.md>), [ai-training](<https://devfeed.tech/tags/ai-training.md>), [anthropic](<https://devfeed.tech/tags/anthropic.md>), [api](<https://devfeed.tech/tags/api.md>), [ccpa](<https://devfeed.tech/tags/ccpa.md>), [chatgpt](<https://devfeed.tech/tags/chatgpt.md>), [checkpoint](<https://devfeed.tech/tags/checkpoint.md>), [children](<https://devfeed.tech/tags/children.md>), [churn](<https://devfeed.tech/tags/churn.md>), [consumer](<https://devfeed.tech/tags/consumer.md>), [customer-data](<https://devfeed.tech/tags/customer-data.md>), [data](<https://devfeed.tech/tags/data.md>), [data-catalog](<https://devfeed.tech/tags/data-catalog.md>), [data-masking](<https://devfeed.tech/tags/data-masking.md>), [data-processing](<https://devfeed.tech/tags/data-processing.md>), [data-protection](<https://devfeed.tech/tags/data-protection.md>), [database](<https://devfeed.tech/tags/database.md>), [database-administration](<https://devfeed.tech/tags/database-administration.md>), [databases](<https://devfeed.tech/tags/databases.md>), [delete](<https://devfeed.tech/tags/delete.md>), [featured](<https://devfeed.tech/tags/featured.md>), [gdpr](<https://devfeed.tech/tags/gdpr.md>), [how-to](<https://devfeed.tech/tags/how-to.md>)

### AI overview

A trained model has no database row to delete: training data influences distributed weights, and personal information may still be extracted from them. The article says retraining without the record is the only guaranteed removal method. SISA retrains the affected shard and checkpoint rollback can remove data from a known fine-tuning batch; approximate unlearning and output filters cannot prove removal, and filters leave data in the weights. Under GDPR, a model is anonymous only if extraction risk is negligible; the article presents extractable personal information as a CCPA concern too. The FTC has ordered destruction of models built on improperly collected data. For erasure requests, check training manifests for affected models and derivatives, then choose a removal action; an output filter may be interim protection. To reduce future exposure, classify and minimize data, pseudonymize and mask identifiers, scrub free text, deduplicate and test for memorization, and consider differential privacy, which can cost model quality. Provider deletion of stored logs or chats does not remove prior learning; training use depends on account tier.

### Source excerpt

You can't DELETE a record from a trained AI model. Compare retraining, SISA, unlearning and filters, and see how to meet GDPR and CCPA erasure requests.... The post How to delete data from a trained AI model (when DELETE doesn't work) appeared first on Simple Talk.

## Felicis leads $7.5M bet on Unsiloed AI to make enterprise documents usable by AI agents

DevFeed: [Felicis leads $7.5M bet on Unsiloed AI to make enterprise documents usable by AI agents](<https://devfeed.tech/articles/felicis-leads-7-5m-bet-on-unsiloed-ai-to-make-enterprise-documents-usable-by-ai-agents-140573.md>)

Original publisher: [Read original article](<https://techcrunch.com/2026/11/18/felicis-leads-7-5m-bet-on-unsiloed-ai-to-make-enterprise-documents-usable-by-ai-agents/>)

Author: Jagmeet Singh

Published: 2026-11-18T12:00:00Z

Content type: news

Language: en

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

Topics: [document ai](<https://devfeed.tech/topics/document-ai.md>), [Artificial Intelligence](<https://devfeed.tech/topics/ai.md>), [data-processing](<https://devfeed.tech/topics/data-processing.md>), [AI Models](<https://devfeed.tech/topics/ai-models.md>), [agentic workflows](<https://devfeed.tech/topics/agentic-workflows.md>)

Tags: [agents](<https://devfeed.tech/tags/agents.md>), [ai](<https://devfeed.tech/tags/ai.md>), [ai-agents](<https://devfeed.tech/tags/ai-agents.md>), [ai-document-processing](<https://devfeed.tech/tags/ai-document-processing.md>), [california](<https://devfeed.tech/tags/california.md>), [document-processing](<https://devfeed.tech/tags/document-processing.md>), [enterprise](<https://devfeed.tech/tags/enterprise.md>), [exclusive](<https://devfeed.tech/tags/exclusive.md>), [felicis](<https://devfeed.tech/tags/felicis.md>), [felicis-ventures](<https://devfeed.tech/tags/felicis-ventures.md>), [startups](<https://devfeed.tech/tags/startups.md>), [unsiloed](<https://devfeed.tech/tags/unsiloed.md>), [y-combinator](<https://devfeed.tech/tags/y-combinator.md>)

### AI overview

Unsiloed AI converts complex business documents into structured data for AI agents and links extracted information back to its source so users can verify it. The company says its dual-stream vision models process text and images separately, use heat-map chunking to focus computing on key areas, and apply an industry-specific decoder. It reports 99.78% accuracy across about 84,000 data points from 542 financial documents in a customer test, and an 88.0 score on olmOCR-Bench versus 66.0 for Reducto and 39.9 for Unstructured. These results were produced or shared by the startup and have not been independently verified. Unsiloed says pricing starts at one cent per page; it has customers in finance, legal services, and healthcare.

### Source excerpt

With a team of just seven, Unsiloed AI plans to compete with tech giants and specialized startups in the crowded AI document-processing market.

## Ai2 and Taylor Geospatial partner to make geospatial AI easier to compare, run, and build on

DevFeed: [Ai2 and Taylor Geospatial partner to make geospatial AI easier to compare, run, and build on](<https://devfeed.tech/articles/ai2-and-taylor-geospatial-partner-to-make-geospatial-ai-easier-to-compare-run-and-build-on-141078.md>)

Original publisher: [Read original article](<https://allenai.org/blog/taylor-geospatial-partnership>)

Published: 2026-11-18T08:00:00Z

Content type: news

Language: en

Sources: [Ai2 Blog](<https://devfeed.tech/sources/ai2-blog.md>)

Topics: [geospatial](<https://devfeed.tech/topics/geospatial.md>), [AI Foundation Models](<https://devfeed.tech/topics/ai-foundation-models.md>), [AI Models](<https://devfeed.tech/topics/ai-models.md>), [Digital Public Good](<https://devfeed.tech/topics/digital-public-goods.md>), [test](<https://devfeed.tech/topics/test.md>), [human feedback](<https://devfeed.tech/topics/human-feedback.md>)

Tags: [2026](<https://devfeed.tech/tags/2026.md>), [accelerate](<https://devfeed.tech/tags/accelerate.md>), [accessible](<https://devfeed.tech/tags/accessible.md>), [agriculture](<https://devfeed.tech/tags/agriculture.md>), [ai](<https://devfeed.tech/tags/ai.md>), [ai2](<https://devfeed.tech/tags/ai2.md>), [applications](<https://devfeed.tech/tags/applications.md>), [benchmarks](<https://devfeed.tech/tags/benchmarks.md>), [build](<https://devfeed.tech/tags/build.md>), [compare](<https://devfeed.tech/tags/compare.md>), [datasets](<https://devfeed.tech/tags/datasets.md>), [geospatial](<https://devfeed.tech/tags/geospatial.md>), [good](<https://devfeed.tech/tags/good.md>), [imagery](<https://devfeed.tech/tags/imagery.md>), [models](<https://devfeed.tech/tags/models.md>), [open](<https://devfeed.tech/tags/open.md>), [partner](<https://devfeed.tech/tags/partner.md>), [public](<https://devfeed.tech/tags/public.md>), [run](<https://devfeed.tech/tags/run.md>), [satellite](<https://devfeed.tech/tags/satellite.md>)

### AI overview

Ai2 and Taylor Geospatial are partnering to provide open resources that help practitioners use and evaluate geospatial AI for conservation, agriculture, and humanitarian work. The effort combines Benchmarks of The World for comparing models, Ai2's OlmoEarth open Earth-observation model family with a no-code platform to fine-tune and run models globally, and an expansion of Fields of The World to improve farm-field boundary precision and add crop type and yield. The partners say datasets, benchmarks, and model weights will be free under open licenses on Source Cooperative and GitHub. Fields of The World already uses a 13-member practitioner advisory group; the partners want similar input to shape shared benchmarks around field needs.

### Source excerpt

Ai2 and Taylor Geospatial are partnering to make geospatial AI more accessible and useful for the public good through open models, shared benchmarks, and improved satellite imagery datasets for applications in conservation, agriculture, and humanitarian work.

## Right Labels, Wrong Reasons: Signed Feature Attribution for Multi-Label Explainability

DevFeed: [Right Labels, Wrong Reasons: Signed Feature Attribution for Multi-Label Explainability](<https://devfeed.tech/articles/right-labels-wrong-reasons-signed-feature-attribution-for-multi-label-explainability-140042.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.22515>)

Author: Kenton White

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: [AI Models](<https://devfeed.tech/topics/ai-models.md>)

Tags: [attribution](<https://devfeed.tech/tags/attribution.md>), [black-box](<https://devfeed.tech/tags/black-box.md>), [classifier](<https://devfeed.tech/tags/classifier.md>), [covariance](<https://devfeed.tech/tags/covariance.md>), [cs-ai](<https://devfeed.tech/tags/cs-ai.md>), [cs-lg](<https://devfeed.tech/tags/cs-lg.md>), [explainability](<https://devfeed.tech/tags/explainability.md>), [feature](<https://devfeed.tech/tags/feature.md>), [features](<https://devfeed.tech/tags/features.md>), [machine-learning](<https://devfeed.tech/tags/machine-learning.md>), [predictions](<https://devfeed.tech/tags/predictions.md>)

### AI overview

FLEX (Fisher LDA Explainer) fits one global, direct multi-label Fisher LDA to the full binary label matrix, yielding signed feature attributions for each label in a shared coordinate system. Across the evaluated benchmarks and classifier families, it reproduces black-box predictions with consistently high fidelity. Pooling covariance across labels beats a per-label alternative on most datasets, especially when labels are numerous and positive examples scarce. Its fidelity rises monotonically with the number of features used, and with a modest feature budget its in-sample fidelity matches or exceeds Binary Relevance alternatives. Discriminant-weight magnitudes bound label separability and can flag cases where correct predictions may rest on weakly separable signals. Although evaluated classifiers largely agree on predictions, they may rely on different feature subsets when no feature dominates.

### Source excerpt

arXiv:2611.22515v1 Announce Type: cross Abstract: When a multi-label classifier assigns the right labels, it may do so for the wrong reasons. Diffuse feature combinations may vary across classifier families and lack a dominant feature. This paper proposes FLEX (Fisher LDA Explainer), a global surrogate that fits a single direct multi-label Fisher LDA over the full binary label matrix, producing signed feature attribution lists for each label in a single coherent coordinate system. Across the evaluated benchmarks and classifier families, FLEX reproduces the black box's predictions with consistently high fidelity. Pooling covariance across labels outperforms a per-label alternative on most datasets, with the advantage growing largest where labels are numerous and positive examples are scarce. FLEX's fidelity-complexity Pareto curve is monotonically increasing, giving practitioners a predictable path to higher-fidelity explanations as more features are used, and its in-sample fidelity matches or leads Binary Relevance alternatives once a modest feature budget is allowed. The magnitude of each discriminant weight vector bounds its label's separability, flagging labels that may be predicted correctly for the wrong reasons. The evaluated classifiers largely agree on which predictions to make, but reach those decisions through different feature subsets when no single feature dominates.

## Social Signal Bench: Evidence-Weighted Evaluation of Co-occurring Social Signals in Video and Audio

DevFeed: [Social Signal Bench: Evidence-Weighted Evaluation of Co-occurring Social Signals in Video and Audio](<https://devfeed.tech/articles/social-signal-bench-evidence-weighted-evaluation-of-co-occurring-social-signals-in-video-and-audio-141801.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.22704>)

Author: Siddharth Ravi, Anders R. Bargum, Hugo Anselme, Patrizio Cugia di Sant'Orsola, Alberto Cardaci, Flora Kincso Parragi, Vincent Borowiec, Line H. Clemmensen

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

Content type: article

Language: en

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

Topics: [multimodal-ai](<https://devfeed.tech/topics/multimodal-ai.md>), [AI Models](<https://devfeed.tech/topics/ai-models.md>), [Dataset development](<https://devfeed.tech/topics/dataset-development.md>), [Video Analytics](<https://devfeed.tech/topics/video-analytics.md>)

Tags: [audio](<https://devfeed.tech/tags/audio.md>), [bench](<https://devfeed.tech/tags/bench.md>), [benchmark](<https://devfeed.tech/tags/benchmark.md>), [computer-vision](<https://devfeed.tech/tags/computer-vision.md>), [cs-cv](<https://devfeed.tech/tags/cs-cv.md>), [cs-lg](<https://devfeed.tech/tags/cs-lg.md>), [engagement](<https://devfeed.tech/tags/engagement.md>), [evaluation](<https://devfeed.tech/tags/evaluation.md>), [frontier-models](<https://devfeed.tech/tags/frontier-models.md>), [recognition](<https://devfeed.tech/tags/recognition.md>), [signal](<https://devfeed.tech/tags/signal.md>), [social](<https://devfeed.tech/tags/social.md>), [video](<https://devfeed.tech/tags/video.md>)

### AI overview

Social Signal Bench evaluates co-occurring conversational signals in 47 hours of video and audio, addressing limits of conventional multi-label F1. Its video dataset covers ten signals in 25 audiovisual hours; its audio dataset covers six signals in 22 hours. Each has a public online-conversation training set, including 10 expert-annotated hours with the remainder pseudo-labeled, and a private expert-labeled test set spanning four domains. The proposed Evidence-Weighted F1 weights reference labels by annotator evidence and gives graded credit to semantically related predictions via optimal per-clip matching; controlled perturbations show it distinguishes related from unrelated errors, unlike conventional F1. Under a shared protocol, post-trained 7B open models outperform general-purpose frontier models on video and match them on audio.

### Source excerpt

arXiv:2611.22704v1 Announce Type: cross Abstract: Recognizing social signals such as interest and hesitation requires interpreting context-dependent behaviors that may co-occur in conversation. Existing affective computing benchmarks rarely evaluate such co-occurring signals, and conventional multi-label F1 overlooks semantic relationships and annotator disagreement. We introduce Social Signal Bench, a 47-hour densely annotated benchmark for the evaluation of social signal recognition and engagement, comprising two datasets: Social Signal Bench Video (25 hours), covering ten social signals in audiovisual recordings, and Social Signal Bench Audio (22 hours), covering six signals in audio-only recordings. Each dataset pairs a public training set of online conversational video, of which 10 hours are expert-annotated and the remainder pseudo-labeled, with a private expert-labeled test set spanning four conversational domains. We also propose Evidence-Weighted F1 (EW-F1), which weights reference labels by annotator evidence and assigns graded credit to semantically related predictions through optimal per-clip matching. Controlled perturbations demonstrate that EW-F1 distinguishes semantically related errors from unrelated ones, unlike conventional F1. Under a shared evaluation protocol, 7B open models post-trained on Social Signal Bench outperform general-purpose frontier models on video and match them on audio.

## Towards Data-Efficient, Real-Time World Models

DevFeed: [Towards Data-Efficient, Real-Time World Models](<https://devfeed.tech/articles/towards-data-efficient-real-time-world-models-140040.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.22497>)

Author: Raluca Stevenson, David Bignell, Jennifer C. White, Tabish Rashid, Emanuele Aiello, Mariya Hendriksen, Junha Hyung, Sadegh Aliakbarian, Abdelhak Lemkhenter, Marko Tot, Katja Hofmann, Sarah Parisot

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

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

Tags: [autoregressive](<https://devfeed.tech/tags/autoregressive.md>), [cs-ai](<https://devfeed.tech/tags/cs-ai.md>), [data-efficiency](<https://devfeed.tech/tags/data-efficiency.md>), [dataset](<https://devfeed.tech/tags/dataset.md>), [diffusion](<https://devfeed.tech/tags/diffusion.md>), [gameplay](<https://devfeed.tech/tags/gameplay.md>), [inference-speed](<https://devfeed.tech/tags/inference-speed.md>), [next-token-prediction](<https://devfeed.tech/tags/next-token-prediction.md>), [real-time](<https://devfeed.tech/tags/real-time.md>), [transformer](<https://devfeed.tech/tags/transformer.md>), [world-models](<https://devfeed.tech/tags/world-models.md>)

### AI overview

The paper reports a two-stage discrete diffusion approach for gameplay world models: a lightweight refinement transformer iteratively predicts masked content instead of relying on next-token prediction. Built on an autoregressive model that generates consistent sequences over minutes, the approach is reported to run 79 times faster with a modest quality trade-off. The authors also report training high-quality, consistent gameplay generation from about 28 hours of gameplay data. These results suggest progress toward real-time use and more data-efficient custom training; the supplied evidence does not specify the exact quality metrics or evaluation conditions.

### Source excerpt

arXiv:2611.22497v1 Announce Type: new Abstract: State-of-the-art world models have been used to produce sequences of gameplay that accord with provided user-input actions. However, high quality, consistent gameplay generation often comes at the cost of inference speed, making real-time interactive play and ideation challenging. The substantial quantities of data used to train state-of-the-art models have also made custom training of these models appear impractical, limiting many of their potential uses in creative fields. In this work we demonstrate a major step towards efficient world models: achieving both fast high-quality inference, and improved data efficiency. This is done by replacing the next-token prediction paradigm with a two-stage discrete diffusion-based approach, introducing a lightweight refinement transformer which carries out iterative masked predictions. Building on an autoregressive world model capable of generating highly consistent and complex sequences over minutes (Kanervisto et al., 2025), we achieve substantial model speed-up (79x) with a modest quality trade-off. Additionally, we demonstrate that this model can be trained to reproduce high-quality, consistent gameplay using a dataset containing approximately one day (28 hours) of gameplay data. Our results demonstrate progress towards removing barriers to the use of world models in settings where real-time inference and data efficiency are essential.

## GPT-6.2 Astra and GPT-6.2 Sol

DevFeed: [GPT-6.2 Astra and GPT-6.2 Sol](<https://devfeed.tech/articles/gpt-6-2-astra-and-gpt-6-2-sol-138615.md>)

Original publisher: [Read original article](<https://simonwillison.net/2026/Nov/17/gpt62-astra-and-sol/>)

Author: Simon Willison

Published: 2026-11-17T22:38:52Z

Content type: news

Language: en

Sources: [Simon Willison's Weblog](<https://devfeed.tech/sources/simon-willison-s-weblog.md>)

Topics: [AI Models](<https://devfeed.tech/topics/ai-models.md>), [Frontier AI](<https://devfeed.tech/topics/frontier-ai.md>), [OpenAI](<https://devfeed.tech/topics/openai.md>), [ChatGPT API](<https://devfeed.tech/topics/chatgpt-api.md>), [animations](<https://devfeed.tech/topics/animations.md>), [3D](<https://devfeed.tech/topics/3d.md>), [Claude](<https://devfeed.tech/topics/claude.md>), [gpt-6-astra](<https://devfeed.tech/topics/gpt-6-astra.md>)

Tags: [agentic](<https://devfeed.tech/tags/agentic.md>), [ai](<https://devfeed.tech/tags/ai.md>), [ai-2-328](<https://devfeed.tech/tags/ai-2-328.md>), [animation](<https://devfeed.tech/tags/animation.md>), [animations](<https://devfeed.tech/tags/animations.md>), [api](<https://devfeed.tech/tags/api.md>), [astra](<https://devfeed.tech/tags/astra.md>), [benchmark](<https://devfeed.tech/tags/benchmark.md>), [brooklyn](<https://devfeed.tech/tags/brooklyn.md>), [claude](<https://devfeed.tech/tags/claude.md>), [claude-fable](<https://devfeed.tech/tags/claude-fable.md>), [claude-fable-5-2](<https://devfeed.tech/tags/claude-fable-5-2.md>), [cli](<https://devfeed.tech/tags/cli.md>), [codex](<https://devfeed.tech/tags/codex.md>), [codex-cli](<https://devfeed.tech/tags/codex-cli.md>), [codex-desktop](<https://devfeed.tech/tags/codex-desktop.md>), [discount](<https://devfeed.tech/tags/discount.md>), [fable-5](<https://devfeed.tech/tags/fable-5.md>), [generative-ai](<https://devfeed.tech/tags/generative-ai.md>), [generative-ai-2-066](<https://devfeed.tech/tags/generative-ai-2-066.md>), [gpt](<https://devfeed.tech/tags/gpt.md>), [gpt-136](<https://devfeed.tech/tags/gpt-136.md>), [gpt-6-2](<https://devfeed.tech/tags/gpt-6-2.md>), [gpt-6-2-astra](<https://devfeed.tech/tags/gpt-6-2-astra.md>), [gpt-6-2-sol](<https://devfeed.tech/tags/gpt-6-2-sol.md>), [llm-pricing](<https://devfeed.tech/tags/llm-pricing.md>), [llm-pricing-99](<https://devfeed.tech/tags/llm-pricing-99.md>), [llm-release](<https://devfeed.tech/tags/llm-release.md>), [llm-release-243](<https://devfeed.tech/tags/llm-release-243.md>), [llms](<https://devfeed.tech/tags/llms.md>), [llms-2-030](<https://devfeed.tech/tags/llms-2-030.md>), [openai](<https://devfeed.tech/tags/openai.md>), [openai-481](<https://devfeed.tech/tags/openai-481.md>), [pelican-riding-a-bicycle](<https://devfeed.tech/tags/pelican-riding-a-bicycle.md>), [pelican-riding-a-bicycle-159](<https://devfeed.tech/tags/pelican-riding-a-bicycle-159.md>), [sol](<https://devfeed.tech/tags/sol.md>), [theo-browne](<https://devfeed.tech/tags/theo-browne.md>), [theo-browne-5](<https://devfeed.tech/tags/theo-browne-5.md>), [yonkers-rising](<https://devfeed.tech/tags/yonkers-rising.md>), [yonkers-rising-3](<https://devfeed.tech/tags/yonkers-rising-3.md>)

### AI overview

GPT-6.2 Astra costs $8/$40 per million input/output tokens, 20% less than GPT-6 Astra, with cached input at $0.40 per million versus $1. GPT-6.2 Sol remains $2/$10, half the stated price of GPT-5.6 Sol. In an Ultra-effort Codex CLI test generating a 3D animation of Yonkers as a mech using OpenStreetMap, the author judged that neither new model handled the transformation as well as Claude Fable 5.2 or Gemini 4 Carbon. The excerpt also reports mixed reactions: Astra reportedly trails Fable 5.2 on some benchmarks, and a reviewer criticized its tendency during long tasks to ask users obscure implementation questions that could block progress.

### Source excerpt

GPT-6.2 Astra and GPT-6.2 Sol OpenAI: We're introducing GPT-6.2 Astra, our new frontier model for your hardest projects, and GPT-6.2 Sol, our efficient, versatile model for everyday work. They're available in the API as gpt-6.2-astra and gpt-6.2-sol. I shipped LLM 0.38 with support for the new models. 6.2 Astra is 80% of the price of 6 Astra - $8/MTok input and $40/MTok output, the same price as Claude Fable 5.2 (which was also 80% of its predecessor). 6 Astra and Fable 5 were both $10/$50. The cached price is an even bigger discount - $0.40/million compared to $1/million for the older model. This is great news for longer running agentic sessions. GPT-6.2 Sol is the same $2/$10 price as 6.1 Sol and 6 Sol - which are half the price of GPT-5.6 Sol ($4/$20). Here are the pelicans for GPT-6.2 Sol and for GPT-6.2 Astra. The 6.2 Astra max pelican looks like this: I have a full comparison grid here, but here's how the high/xhigh/max pelicans compare with the previous Sol and Astra releases: I also ran my most recent benchmark prompt against the two models, using "Ultra' effort levels in the Codex CLI: Make a three js animation where yonkers stands up as a big mech (cool stand up motion) and stomps around manhattan and queens and Brooklyn, use openstreetmap mapping data Here's Sol and here's Astra. Neither of them did a good enough job of morphing Yonkers into a big mech, at least in comparison to Claude Fable 5.2 and Gemini 4 Carbon. This video showing the Astra and Sol animations playing out at the same time: I created that video using GPT-6.2 Astra (Ultra) in the Codex desktop app by prompting: https://static.simonwillison.net/static/2026/yonkers-gpt-6.2-astra-ultra/index.html and https://static.simonwillison.net/static/2026/yonkers-gpt-6.2-sol-ultra/index.html are two videos of the same thing by two different models I need a video of the first 10 seconds where the top half of the video plays the first and the bottom half plays the second I need an overlaid label of "GPT

## AI models can be tricked into identifying civilians as military targets by telling them it's a game

DevFeed: [AI models can be tricked into identifying civilians as military targets by telling them it's a game](<https://devfeed.tech/articles/ai-models-can-be-tricked-into-identifying-civilians-as-military-targets-by-telling-them-it-s-a-game-138179.md>)

Original publisher: [Read original article](<https://techcrunch.com/2026/11/17/ai-models-can-be-tricked-into-identifying-civilians-as-military-targets-by-telling-them-its-a-game/>)

Author: Kirsten Korosec, Rebecca Bellan

Published: 2026-11-17T17:20:41Z

Content type: news

Language: en

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

Topics: [ai safety](<https://devfeed.tech/topics/ai-safety.md>), [LLM security](<https://devfeed.tech/topics/llm-security.md>), [AI Models](<https://devfeed.tech/topics/ai-models.md>), [ai security](<https://devfeed.tech/topics/ai-security.md>), [Frontier AI](<https://devfeed.tech/topics/frontier-ai.md>)

Tags: [ai](<https://devfeed.tech/tags/ai.md>), [ai-evaluation](<https://devfeed.tech/tags/ai-evaluation.md>), [ai-models](<https://devfeed.tech/tags/ai-models.md>), [anthropic](<https://devfeed.tech/tags/anthropic.md>), [autonomous](<https://devfeed.tech/tags/autonomous.md>), [autonomous-weapons](<https://devfeed.tech/tags/autonomous-weapons.md>), [battlefield](<https://devfeed.tech/tags/battlefield.md>), [deepseek](<https://devfeed.tech/tags/deepseek.md>), [defense](<https://devfeed.tech/tags/defense.md>), [drone](<https://devfeed.tech/tags/drone.md>), [drones](<https://devfeed.tech/tags/drones.md>), [frontier-ai](<https://devfeed.tech/tags/frontier-ai.md>), [google](<https://devfeed.tech/tags/google.md>), [human-in-the-loop](<https://devfeed.tech/tags/human-in-the-loop.md>), [imagery](<https://devfeed.tech/tags/imagery.md>), [military](<https://devfeed.tech/tags/military.md>), [openai](<https://devfeed.tech/tags/openai.md>)

### AI overview

Vals's simulated tests found that leading commercial AI models frequently identified potential military targets in drone and satellite imagery, including civilians, surrendering soldiers, and a school; refusals were uncommon, and models often complied after researchers framed scenarios as a game. The models were not asked to make strike decisions, and the report does not cover models provided to militaries. Vals says latency and limited onboard computing constrain live drone use, while annotating battlefield footage at scale could supply training data for autonomous targeting. Anthropic said the outcomes violated its rules and described planned safeguards and enforcement; OpenAI said public-model use to facilitate violence is prohibited and that it is developing drone-related safeguards. The excerpt does not establish whether such safeguards apply to military use.

### Source excerpt

Vals, a San Francisco-based AI evaluation startup, tested whether leading AI models could be manipulated into participating in autonomous warfare. They usually compiled.

## SiMa.ai's Palette 3.0 runs many AI models on one chip and lifts throughput up to 3.5 times

DevFeed: [SiMa.ai's Palette 3.0 runs many AI models on one chip and lifts throughput up to 3.5 times](<https://devfeed.tech/articles/sima-ai-s-palette-3-0-runs-many-ai-models-on-one-chip-and-lifts-throughput-up-to-3-5-times-137728.md>)

Original publisher: [Read original article](<https://siliconangle.com/2026/11/17/sima-ais-palette-3-0-runs-many-ai-models-on-one-chip-and-lifts-throughput-up-to-3-5-times/>)

Author: Duncan Riley

Published: 2026-11-17T14:00:15Z

Content type: news

Language: en

Sources: [SiliconANGLE](<https://devfeed.tech/sources/siliconangle.md>)

Topics: [Local AI](<https://devfeed.tech/topics/local-ai.md>), [Artificial Intelligence](<https://devfeed.tech/topics/ai.md>), [AI Development](<https://devfeed.tech/topics/ai-development.md>), [AI Models](<https://devfeed.tech/topics/ai-models.md>)

Tags: [agentic](<https://devfeed.tech/tags/agentic.md>), [agentic-development](<https://devfeed.tech/tags/agentic-development.md>), [ai](<https://devfeed.tech/tags/ai.md>), [ai-application](<https://devfeed.tech/tags/ai-application.md>), [ai-application-development](<https://devfeed.tech/tags/ai-application-development.md>), [ai-models](<https://devfeed.tech/tags/ai-models.md>), [amplify-partners](<https://devfeed.tech/tags/amplify-partners.md>), [artificial-intelligence](<https://devfeed.tech/tags/artificial-intelligence.md>), [autonomous-machines](<https://devfeed.tech/tags/autonomous-machines.md>), [chip](<https://devfeed.tech/tags/chip.md>), [chip-startup](<https://devfeed.tech/tags/chip-startup.md>), [concurrently](<https://devfeed.tech/tags/concurrently.md>), [developer-center](<https://devfeed.tech/tags/developer-center.md>), [development-environment](<https://devfeed.tech/tags/development-environment.md>), [dflash](<https://devfeed.tech/tags/dflash.md>), [docker](<https://devfeed.tech/tags/docker.md>), [drones](<https://devfeed.tech/tags/drones.md>), [eagle3](<https://devfeed.tech/tags/eagle3.md>), [emerson-electric](<https://devfeed.tech/tags/emerson-electric.md>), [fidelity-management-research](<https://devfeed.tech/tags/fidelity-management-research.md>), [gemma-4-mtp](<https://devfeed.tech/tags/gemma-4-mtp.md>), [generative-ai](<https://devfeed.tech/tags/generative-ai.md>), [github](<https://devfeed.tech/tags/github.md>), [gpt-oss-20b](<https://devfeed.tech/tags/gpt-oss-20b.md>), [infra](<https://devfeed.tech/tags/infra.md>), [krishna-rangasayee](<https://devfeed.tech/tags/krishna-rangasayee.md>), [large-language-models](<https://devfeed.tech/tags/large-language-models.md>), [mixture-of-experts](<https://devfeed.tech/tags/mixture-of-experts.md>), [modalix](<https://devfeed.tech/tags/modalix.md>), [modalix-mlsoc](<https://devfeed.tech/tags/modalix-mlsoc.md>), [model-registry](<https://devfeed.tech/tags/model-registry.md>), [news](<https://devfeed.tech/tags/news.md>), [open-source](<https://devfeed.tech/tags/open-source.md>), [openai](<https://devfeed.tech/tags/openai.md>), [palette](<https://devfeed.tech/tags/palette.md>), [palette-3-0](<https://devfeed.tech/tags/palette-3-0.md>), [palette-neat](<https://devfeed.tech/tags/palette-neat.md>), [physical-ai](<https://devfeed.tech/tags/physical-ai.md>), [release](<https://devfeed.tech/tags/release.md>), [robert-bosch](<https://devfeed.tech/tags/robert-bosch.md>), [robots](<https://devfeed.tech/tags/robots.md>), [run](<https://devfeed.tech/tags/run.md>), [sima-ai](<https://devfeed.tech/tags/sima-ai.md>), [sima-technologies](<https://devfeed.tech/tags/sima-technologies.md>), [software](<https://devfeed.tech/tags/software.md>), [startup](<https://devfeed.tech/tags/startup.md>), [the-latest](<https://devfeed.tech/tags/the-latest.md>), [throughput](<https://devfeed.tech/tags/throughput.md>)

### AI overview

SiMa.ai says Palette 3.0 lets its Modalix MLSoC run multiple AI models concurrently, including LLMs, vision-language models, and perception, depth, and reasoning workloads, while supporting separate Docker containers. The company claims throughput can rise by up to 3.5x through speculative decoding with EAGLE3, DFlash, and Gemma 4 MTP; the article provides no independent verification of that figure. New support includes GPT-OSS-20B on the 16GB system-on-module, open-source generative AI development capabilities, and partner features for sensor fusion, neural depth fields, and digital-twin simulation. SiMa.ai says its Model Registry reports results within real-world power constraints, with competitive benchmarking details available on request.

### Source excerpt

Physical artificial intelligence chip startup SiMa Technologies Inc. today released Palette 3.0, a new version of its Palette software suite intended to allow many AI models to run concurrently on one chip. According to the company, the release, which builds on the Palette Neat agentic development environment launched in June, can raise model throughput by [...] The post SiMa.ai's Palette 3.0 runs many AI models on one chip and lifts throughput up to 3.5 times appeared first on SiliconANGLE.

## Residual-Tail-Informed Flow Matching for Time Series Anomaly Detection

DevFeed: [Residual-Tail-Informed Flow Matching for Time Series Anomaly Detection](<https://devfeed.tech/articles/residual-tail-informed-flow-matching-for-time-series-anomaly-detection-137563.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.20334>)

Author: Shyam Sundar Saravanan, Tie Luo, Sanjay Madria

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

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: [Time Series](<https://devfeed.tech/topics/time-series.md>), [AI Models](<https://devfeed.tech/topics/ai-models.md>), [VAE](<https://devfeed.tech/topics/vae.md>), [Temporal data](<https://devfeed.tech/topics/temporal-data.md>)

Tags: [anomaly-detection](<https://devfeed.tech/tags/anomaly-detection.md>), [autoencoder](<https://devfeed.tech/tags/autoencoder.md>), [cs-ai](<https://devfeed.tech/tags/cs-ai.md>), [cs-lg](<https://devfeed.tech/tags/cs-lg.md>), [flow-matching](<https://devfeed.tech/tags/flow-matching.md>), [machine-learning](<https://devfeed.tech/tags/machine-learning.md>), [time-series](<https://devfeed.tech/tags/time-series.md>)

### AI overview

RT-FM targets false alarms on rare but valid time-series regimes that data-space flow matching may score like true anomalies. It uses a temporal variational autoencoder to estimate residual latent tailness, then turns that estimate into bounded, warmup-annealed weights for training an unsupervised data-space flow-matching model. Across seven TSAD benchmarks, the authors report better rare-normal coverage, fewer false alarms on those windows, and improved aggregate detection while preserving strong anomaly discrimination.

### Source excerpt

arXiv:2611.20334v1 Announce Type: cross Abstract: Time series anomaly detection (TSAD) models are often trained on nominal data using average-risk objectives, which emphasize dominant normal patterns. However, real time series can contain rare but valid nominal regimes that are underrepresented in the training distribution. We show that data-space flow matching can over-penalize such rare-normal windows and assign them anomaly scores comparable to those of true anomalies. To address this, we propose Residual-Tail-Informed Flow Matching (RT-FM), an unsupervised framework that uses a temporal variational autoencoder to estimate residual latent tailness and converts it into bounded, warmup-annealed weights for training a data-space flow matching model. Experiments on seven TSAD benchmarks show that RT-FM improves rare-normal coverage while preserving strong anomaly discrimination, substantially reducing false alarms on such windows and resulting in improved aggregate detection performance. The code is available at: https://github.com/trustAIoT/RT-FM_time-series-anomaly

## Learning From the Best: What Makes Popular Hugging Face Models?

DevFeed: [Learning From the Best: What Makes Popular Hugging Face Models?](<https://devfeed.tech/articles/learning-from-the-best-what-makes-popular-hugging-face-models-137999.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.20227>)

Author: Yinan Wu, Zhou Yang, Bowen Xu, Bo Wang, Bach Le, David Lo

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

Content type: article

Language: en

Sources: [arXiv: Software Engineering](<https://devfeed.tech/sources/arxiv-software-engineering.md>)

Topics: [hugging face](<https://devfeed.tech/topics/hugging-face.md>), [AI Models](<https://devfeed.tech/topics/ai-models.md>), [Predictive AI](<https://devfeed.tech/topics/predictive-ai.md>), [Post-training optimization](<https://devfeed.tech/topics/post-training-optimization.md>)

Tags: [2026](<https://devfeed.tech/tags/2026.md>), [arxiv](<https://devfeed.tech/tags/arxiv.md>), [code](<https://devfeed.tech/tags/code.md>), [cs-se](<https://devfeed.tech/tags/cs-se.md>), [documentation](<https://devfeed.tech/tags/documentation.md>), [downloads](<https://devfeed.tech/tags/downloads.md>), [feature-importance](<https://devfeed.tech/tags/feature-importance.md>), [features](<https://devfeed.tech/tags/features.md>), [future](<https://devfeed.tech/tags/future.md>), [github](<https://devfeed.tech/tags/github.md>), [hugging-face](<https://devfeed.tech/tags/hugging-face.md>), [key](<https://devfeed.tech/tags/key.md>), [learning](<https://devfeed.tech/tags/learning.md>), [manual-review](<https://devfeed.tech/tags/manual-review.md>), [metadata](<https://devfeed.tech/tags/metadata.md>), [model](<https://devfeed.tech/tags/model.md>), [models](<https://devfeed.tech/tags/models.md>), [popular](<https://devfeed.tech/tags/popular.md>), [recommendations](<https://devfeed.tech/tags/recommendations.md>), [repository](<https://devfeed.tech/tags/repository.md>), [research](<https://devfeed.tech/tags/research.md>)

### AI overview

Across 66,909 Hugging Face models, popular models tended to have longer model cards, more code examples, and more GitHub and arXiv links. A study comparing the top 10% by downloads or likes with the bottom 80% found that quantization patterns differed between those popularity measures. A Random Forest predicted popularity with AUCs of 0.861 for downloads and 0.848 for likes; root-level file counts and model size also ranked among important features. Predictions transferred across model domains and contributor groups mostly above chance, but average AUC fell by 0.12-0.17, indicating group-specific performance. Findings were stable to download-outlier removal and alternative popularity thresholds. These are associations and predictive results; the abstract does not establish that the listed features cause popularity.

### Source excerpt

arXiv:2611.20227v1 Announce Type: new Abstract: Hugging Face has become a central platform for sharing and reusing pre-trained models, yet it remains unclear which practices are associated with higher visibility and adoption. Prior studies describe model reuse and model-card quality but do not systematically compare popular and unpopular models across documentation, repository structure, platform metadata, and technical metadata. We studied 66,909 Hugging Face models collected on 2026-02-28 and extracted 31 features across four dimensions: Documentation, Models, Platforms, and Techniques. For downloads and likes, we compared the top 10% popular models with the bottom 80% using statistical tests and effect sizes, trained five classifiers to predict popularity, ranked feature importance with the best-performing model, and tested whether predictions held across model domains and contributor affiliations. Popular models consistently had longer model cards, more code examples, and more GitHub and arXiv links, while quantized models showed contrasting patterns under downloads and likes. A Random Forest reached AUC 0.861 for downloads and 0.848 for likes, with root-level file counts and model size also among the most important features. Cross-group transfer stayed mostly above random but dropped by 0.12-0.17 AUC on average, showing that prediction performance was partly group-specific. The findings remained stable after removing download outliers and under alternative popularity thresholds, and a manual review of popular model cards linked the key documentation features to concrete content. We derive recommendations for model practitioners and directions for future research on model-sharing platforms.

[Next page](<https://devfeed.tech/topics/ai-models.md?cursor=WyIyMDI2LTExLTE3VDA1OjAwOjAwKzAwOjAwIiwgImRkZDNhOGY1LTU0OWYtNDAwOS1iNDNkLWI2YzMyOTNkZjYyMyJd>)