Skip to content
AI News HubLIVE

This edition’s highlights

Models

Can AI help scientists identify promising drug candidates?

  • MedChemBench includes 99 problems from published research, drawing on 117 papers and patents and 761 distinct molecular structures.
  • GPT‑6.2 Astra scored 84.5% of available credit at the xhigh reasoning setting, with strong structure reconstruction but weaker compound selection and property prediction.
OpenAI NewsIn-site articleCan AI help scientists identify promising drug candidates?

Amazon SageMaker Inference: 2026 launches in review, part 2

  • Inference recommendations now extend to SageMaker AI Studio, Python SDK v3, and a coding agent skill for choosing instances, containers, and optimizations.
  • New capabilities include LoRA adapter support, SLA-based concurrency search, and console management for inference components.
AWS Machine Learning BlogIn-site articleAmazon SageMaker Inference: 2026 launches in review, part 2

VG-Steer: Visual-Geometric Policy Steering from In-Context Aligned Abstractions

  • VG-Steer corrects frozen policies by transferring nuanced motion from successful in-distribution rollouts along object-relational distance pairs, avoiding reliance on implicitly learned signals that may suffer from distribution shifts.
  • The method leverages LVMs and VLMs to combine task-relevant entity pairs and their object-relational distance projections into a tailored visual-geometric guidance signal.
arXiv RoboticsIn-site articleVG-Steer: Visual-Geometric Policy Steering from In-Context Aligned Abstractions

HireWatch: Evaluating LLM Compliance with U.S. Employment Law Under Contextual Pressure

  • HireWatch uses a 200-item bank of lawful and unlawful interview questions classified under U.S. federal statutes, regulations, and agency guidance.
  • In about 100,000 trials, 18 closed and open-weight LLMs selected far more unlawful questions under higher contextual pressure.
arXiv Computational LinguisticsIn-site articleHireWatch: Evaluating LLM Compliance with U.S. Employment Law Under Contextual Pressure

EgoPass-Bench: Identifying Objects That May Affect the Wearer's Passage in Egocentric Video

  • EgoPass-Bench includes 284 manually reviewed egocentric clips and 2,201 questions across five tasks.
  • Eight frame-based and six native-video configurations were evaluated, revealing a gap between motion perception and interaction understanding.
arXiv Computer VisionIn-site articleEgoPass-Bench: Identifying Objects That May Affect the Wearer's Passage in Egocentric Video
Chips

Using NVIDIA Run:ai GPU Fractioning on Amazon EKS

  • Two Pods shared one NVIDIA A10G GPU, each allocating 6 GiB within an 8 GiB limit.
  • GPU Fractioning adds Kubernetes scheduling integration, enforced per-Pod memory limits, and metrics on top of MPS.
AWS Machine Learning BlogIn-site articleUsing NVIDIA Run:ai GPU Fractioning on Amazon EKS
Research

Verify2Act: Critic-Guided Latent World Models for Verifying Language-Conditioned Manipulation Plans

  • Verify2Act reaches 83.3% task success on six language-conditioned tasks, compared to 53.3% for the VLM alone and 46.7% for a pixel-diffusion world model.
  • The framework trains the world model and critic only on data generated in a digital twin and deploys on a low-cost robot arm.
arXiv RoboticsIn-site articleVerify2Act: Critic-Guided Latent World Models for Verifying Language-Conditioned Manipulation Plans
Policy

Quantifying Cultural-Psychological Constructs in Four Centuries of Ottoman Court Registers

  • The corpus spans 54,670 Ottoman Turkish court records from 21 Istanbul courts between 1513 and 1911 CE.
  • The adapted OTA-BERTurk sentence transformer achieved Recall@1 of 0.843 for document retrieval, MRR of 0.806 for semantic textual similarity, and 89.58% accuracy on questionnaire-item classification.
arXiv Computational LinguisticsIn-site articleQuantifying Cultural-Psychological Constructs in Four Centuries of Ottoman Court Registers

You have reached the end of this edition.

Your reading list
Other updates (72)
Models

Combatting adversarial distillation

OpenAI details its layered defenses against adversarial distillation—combining adversarial training, classifiers, API protections, campaign investigations, and enforcement—and reports major reductions in reasoning extraction success for GPT-6.2 Astra and Sol.

OpenAI NewsIn-site articleCombatting adversarial distillation

Bridging vision-language reasoning and learned control for legged navigation in the wild

ROAM combines a vision-language model's world knowledge with reinforcement-learned locomotion and planning policies, enabling legged robots to navigate kilometer-scale alpine and urban environments, interpret signs, and assess risks, though VLM reliability and spatial reasoning remain challenges.

arXiv RoboticsIn-site articleBridging vision-language reasoning and learned control for legged navigation in the wild

MV-JEPA: Post-Training Video Encoders for Cross-View Understanding

MV-JEPA introduces multi-view predictive post-training that strengthens cross-view instance and temporal correspondence in JEPA video encoders with only 9,000 updates on a single GPU, while preserving single-view inference and most pretrained capabilities.

arXiv Computer VisionIn-site articleMV-JEPA: Post-Training Video Encoders for Cross-View Understanding

What a Calibrated Probability Hides: Belief-Level Post-Training for Decision Models

The paper introduces Belief-Grouped Policy Optimization (BGPO), which separates disagreement among confident groups from uncertainty within a shared belief, and achieves lower Jensen-Shannon divergence on several benchmarks.

arXiv Computer VisionIn-site articleWhat a Calibrated Probability Hides: Belief-Level Post-Training for Decision Models

CoRelate: Coordinated Human-Object Motion Synthesis through Interaction Relations

CoRelate is a diffusion framework that generates coordinated human-object motion by using explicit hand-part interaction relations and task feedback to ensure contact and task completion. Experiments show improved object-goal satisfaction, contact F1, and complete interaction success over baselines.

arXiv Computer VisionIn-site articleCoRelate: Coordinated Human-Object Motion Synthesis through Interaction Relations

The ROI of Small Examples: Optimizing In-Context Learning Under Token Budgets

This EACL 27 main-conference paper studies prompt construction for in-context learning under a fixed token budget, finding that spending the budget on more short demonstrations is generally more effective than using a few long ones for question answering, across six instruction-tuned LLMs and two QA benchmarks.

arXiv Computational LinguisticsIn-site articleThe ROI of Small Examples: Optimizing In-Context Learning Under Token Budgets

Do Deep Stereo Models Learn Shape Priors?

A new study investigates whether deep stereo models produce boundaries consistent with simple shape priors when local cues are weak. Using a synthetic dataset of constrained shapes, researchers find systematic boundary errors in state-of-the-art models and propose a generative model that jointly predicts disparity and a distance field to improve boundary localization.

arXiv Computer VisionIn-site articleDo Deep Stereo Models Learn Shape Priors?

VisAlloc: Allocating Visual Supervision for On-Policy Self-Distillation

VisAlloc introduces a dual-level framework that combines sample-level visual filtering with token-level counterfactual reweighting to improve visual reasoning in multimodal large language models through on-policy self-distillation.

arXiv Computer VisionIn-site articleVisAlloc: Allocating Visual Supervision for On-Policy Self-Distillation

Infer3D: Inverse Generative Modeling Enables Robust 3D Reconstruction

arXiv paper Infer3D introduces an inverse generative modeling framework that recovers latent variables of a generative model to match monocular input, coupled with a frozen 3D lifting network to constrain optimization to a valid manifold. It achieves superior out-of-distribution reconstruction accuracy compared to feedforward networks and uses adaptive inference to allocate more test-time compute to ambiguous examples.

arXiv Computer VisionIn-site articleInfer3D: Inverse Generative Modeling Enables Robust 3D Reconstruction

Physics of Language Models: Part 4.2, Canon Layers at Scale where Synthetic Pretraining Resonates in Reality

Zeyuan Allen-Zhu runs 64 from-scratch pretrainings (1.3–8B parameters, 1–2T Nemotron-CC tokens, roughly $100k per 8B run) to ask whether standard benchmarks capture architecturally meaningful differences. At academic scale (1.3B/100B) differences nearly vanish under careful tuning, but at 1–2T tokens Canon layers consistently improve Transformers and save about 30% of pretraining tokens at 8B for the same MMLU accuracy. New GLA5 and GDN2 variants improve gated linear attention and Gated DeltaNet with low-rank output gates; with Canon, plain gated linear attention matches Gated DeltaNet and beats Mamba2 even at 8B/1T. Yet even at that scale, linear models that match Transformers on benchmarks fail to reliably retrieve from a 500-token context, and no model beats random guessing on the simp…

arXiv Computational LinguisticsIn-site articlePhysics of Language Models: Part 4.2, Canon Layers at Scale where Synthetic Pretraining Resonates in Reality

Diagnostic Accuracy and Navigation Behavior of Vision-Language Model Agents Reading Emergency CT and MRI Examinations

A retrospective study evaluated six vision-language models reading 100 emergency CT and MRI exams under four exploration strategies, finding that accuracy remained far below radiologists and models adopted early hypotheses with little correction.

arXiv AIIn-site articleDiagnostic Accuracy and Navigation Behavior of Vision-Language Model Agents Reading Emergency CT and MRI Examinations

LLM Complexity: Towards a Computational Complexity View of Large Language Model Inference

The paper disproves the analogy between LLM token count and classical time complexity, introduces LLM Complexity—a metric based on the area under the accuracy-token curve—and shows it yields consistent task difficulty rankings across models, enabling an easy-to-hard curriculum that improves fine-tuning accuracy while reducing token usage.

arXiv Computational LinguisticsIn-site articleLLM Complexity: Towards a Computational Complexity View of Large Language Model Inference

RFBench: A Simulator-Grounded Multimodal Benchmark for Passive RF Layout Design and Understanding

Researchers introduce RFBench, a simulator-grounded dataset and benchmark built with ADS Momentum, containing 10,000 two-port planar layouts and seven tasks to evaluate multimodal language models on passive RF layout design and understanding. Baselines reveal systematic weaknesses, while RF-specific fine-tuning substantially improves structural validity and response prediction accuracy.

arXiv AIIn-site articleRFBench: A Simulator-Grounded Multimodal Benchmark for Passive RF Layout Design and Understanding

HapBO: Harnessed α-π Bayesian Optimization via Expert Natural Language Priors

HapBO introduces a method that compiles natural language beliefs from experts or LLMs once into a reusable belief function, applied in Bayesian optimization via a bounded positive multiplier without in-loop LLM calls. Theoretical analysis shows that under Expected Improvement, HapBO preserves the standard simple-regret rate up to a bounded factor even with adversarial beliefs. Evaluated on 40 HPOBench tasks, a synthetic suite, and two real-world applications (cortical neurostimulation and constrained solar design), HapBO achieves the lowest aggregate HPOBench regret and strong real-world performance with zero in-loop token cost.

arXiv Machine LearningIn-site articleHapBO: Harnessed α-π Bayesian Optimization via Expert Natural Language Priors

HardwareDx: A Foundational Graph Model System for Datacenter Machine Repairs

Google's HardwareDx uses a graph foundation model to automate datacenter hardware diagnosis, reducing production downtime by 18.3% on GPUs and 16.1% on TPUs, and cutting QA time by 28.3% and 25.1%, respectively. The paper is accepted to NeurIPS 2026.

arXiv AIIn-site articleHardwareDx: A Foundational Graph Model System for Datacenter Machine Repairs

A Model of Diverse Sampling from Language Models

The paper introduces PoolDPP, which formalizes diverse generation as sampling a size-K subset of complete strings from a determinantal point process (DPP) and approximates it via a finite candidate pool, with a theoretical KL bound decaying as O(1/N). Experiments show improved quality-diversity and precision-recall frontiers on open-ended generation, text-to-SQL, and math reasoning.

arXiv Machine LearningIn-site articleA Model of Diverse Sampling from Language Models

Training Language Models for Introspective Oversight

A new paper introduces introspective oversight: training a separate third-person observer role inside the model to answer arbitrary questions about the assistant, with answers rewarded only when verified against the assistant's actual behavior. On Qwen3.6-27B, the observer reaches a mean AUROC of 0.907, beating honesty training, transcript monitoring, and activation oracles; training the observer also makes the assistant itself easier to monitor.

arXiv AIIn-site articleTraining Language Models for Introspective Oversight

Density Feedback Control for Transferability of Generative Sparse Adversarial Attacks

This paper proposes Density Feedback Control (DFC), which dynamically adjusts the sparsity coefficient at the end of each epoch based on the gap between observed hard-mask density and target density, enabling generative sparse adversarial attacks to achieve target density in a single run and improving transfer attack success rates on CNN and Transformer models.

arXiv Machine LearningIn-site articleDensity Feedback Control for Transferability of Generative Sparse Adversarial Attacks

llm-gemini 0.38.1 fixes xhigh reasoning level support for Gemini 4 Hydrogen and Neon

Simon Willison releases llm-gemini 0.38.1, a dot release that adds the previously missed xhigh reasoning level for Gemini 4 Hydrogen and Gemini 4 Neon models. Comparison shows Hydrogen at xhigh uses far more output tokens and costs more, but produces quality competitive with Gemini 4 Neon.

Simon Willison's WeblogIn-site articlellm-gemini 0.38.1 fixes xhigh reasoning level support for Gemini 4 Hydrogen and Neon

Hebbia sees improvements in financial research with GPT-6 Astra

Hebbia tested OpenAI's prerelease GPT-6 Astra for financial research, finding gains in source coverage, citation relevance, and presentation quality, with 99.5% citation relevance and 97.9% presentation requirements met.

OpenAI NewsIn-site articleHebbia sees improvements in financial research with GPT-6 Astra

Qwen3 TTS is now available on Together AI

Together AI launches Qwen3 TTS, an open text-to-speech model with nine preset voices, ten languages, natural-language tone control, and first audio latency as low as 97ms, priced at $20 per 1M characters.

Together AI BlogIn-site articleQwen3 TTS is now available on Together AI

NVIDIA and Microsoft Expand Accelerated Computing From Azure to Windows PCs

NVIDIA and Microsoft are extending their full-stack collaboration at Microsoft Ignite, bringing Vera Rubin NVL72 systems to Azure, expanding open model customization through Microsoft Foundry, and pushing AI to edge devices and Windows PCs.

NVIDIA BlogIn-site articleNVIDIA and Microsoft Expand Accelerated Computing From Azure to Windows PCs
Research

Ferrofluid Shape Control with Deep Reinforcement Learning

A new study uses deep model-based reinforcement learning to control ferrofluid transient shapes with a single fixed electromagnet, learning directly from physical interaction to achieve polygons, asterisks, and pentominoes.

arXiv RoboticsIn-site articleFerrofluid Shape Control with Deep Reinforcement Learning

Old-School Planners Drive (Sometimes) Better Than Learned Ones

An arXiv study implements sixteen textbook driving planners and benchmarks them on nuScenes, nuPlan, and HUGSIM. On HUGSIM, nine of the sixteen outperform the learned planner of the end-to-end model UniAD using the same frozen perception; the best combination raises HD-Score from 32.0 to 41.8. In 31 of the 182 runs where UniAD's planner crashes into a car, the best textbook planner completes the route on identical perception, attributing those failures to planning rather than perception.

arXiv RoboticsIn-site articleOld-School Planners Drive (Sometimes) Better Than Learned Ones

GEMS: Generation and Experience Memory for Subgoal-Guided Long-Horizon World Model Planning

GEMS combines temporally structured generated subgoals with a Latent Experience Memory (LEM) to guide long-horizon planning with a frozen world model. A distance gate decides whether predicted outcomes are compared against the generated subgoal or a selected memory state. It achieves 92.8% on Cube, 68.4% on Push-T, and 54.1% on Meta-World, outperforming the strongest baseline by 24.0, 7.3, and 22.2 percentage points respectively.

arXiv RoboticsIn-site articleGEMS: Generation and Experience Memory for Subgoal-Guided Long-Horizon World Model Planning

Identity World Model: Recognizing People in an Imagined Latent World

The paper introduces IDWorld, an identity world model that imagines how a person would be represented under changed conditions—such as different clothing, viewpoint, or pose—in latent space without rendering pixels, enabling more robust whole-body person recognition at 5.9 ms per frame and leading results across several benchmarks.

arXiv Computer VisionIn-site articleIdentity World Model: Recognizing People in an Imagined Latent World

Phantom Success, Silent Failure: Auditing the Success Checkers of Simulated Manipulation Benchmarks

The paper introduces SimAudit, an LLM-based pipeline that audits hand-written success checkers in simulated manipulation benchmarks against task instructions. Across RoboTwin 2.0, RoboCasa365, and RoboDojo, 183 test cases yielded 61 distinct defects across 49 tasks, including 40 phantom successes and 21 silent failures.

arXiv Computer VisionIn-site articlePhantom Success, Silent Failure: Auditing the Success Checkers of Simulated Manipulation Benchmarks

No-Harm-Jury: A Jury-Based Multi-Dimensional Harm Evaluation Framework for Medical AI Responses

Researchers present No-Harm-Jury, a framework that uses five diverse language models as independent jurors to score AI-generated medical responses across seven harm dimensions. Applied to 2,849 valid responses from Ministral-14B on PubMedQA, MedQA, and MedMCQA, the framework employs median aggregation to suppress individual juror biases.

arXiv Computational LinguisticsIn-site articleNo-Harm-Jury: A Jury-Based Multi-Dimensional Harm Evaluation Framework for Medical AI Responses

TALD: Transferability-Aware Logit Disentanglement for Cross-Modal Knowledge Distillation

Researchers propose TALD, a two-stage framework that disentangles logits from a teacher and a target-modality reference student to mitigate negative transfer in cross-modal knowledge distillation, achieving the largest mean accuracy gain across five benchmarks and outperforming no-distillation training in 84% of evaluations.

arXiv Computer VisionIn-site articleTALD: Transferability-Aware Logit Disentanglement for Cross-Modal Knowledge Distillation

Global Importance Estimation for KV Cache Eviction

Researchers introduce PRKV, a training-free KV cache eviction method that runs Personalized PageRank over attention-induced transition graphs to estimate multi-hop global token importance, letting importance propagate beyond recency or instantaneous attention. Benchmarks on long-context understanding, sparse retrieval, and chain-of-thought reasoning show competitive performance against established eviction baselines even at compression ratios up to 90%, while cutting peak memory with low end-to-end overhead.

arXiv Computational LinguisticsIn-site articleGlobal Importance Estimation for KV Cache Eviction

TETHER: Training-Free, Elastic, ToC-Anchored Hierarchy Extraction and Recovery in Design-Driven Long Documents

TETHER is a training-free framework that elevates the table of contents (ToC) from discarded noise to a vital structural prior for hierarchy recovery in long, design-driven documents. Using a two-stage neuro-symbolic mechanism—LMM-based skeleton construction with spatio-semantic constraints followed by anchor-based alignment of content blocks—it achieves state-of-the-art zero-shot hierarchy recovery on multilingual ESG reports, product manuals, and annual reports, while cutting tokens by 90% and drastically speeding up over end-to-end baselines.

arXiv Computational LinguisticsIn-site articleTETHER: Training-Free, Elastic, ToC-Anchored Hierarchy Extraction and Recovery in Design-Driven Long Documents

ICARE: cross-domain text classification with incremental class-aware representation and distillation learning

ICARE is a novel framework that combines a representation-learning-based instance selection module with feature- and logit-level knowledge distillation to mitigate catastrophic forgetting in continual multi-task natural language inference, achieving significant improvements in target accuracy and reductions in forgetting on benchmark text classification datasets.

arXiv Computational LinguisticsIn-site articleICARE: cross-domain text classification with incremental class-aware representation and distillation learning

VeriMath: Human-in-the-Loop Generation and Verification of Curriculum-Aligned Mathematics Assessment Items

VeriMath is a human-in-the-loop framework for generating curriculum-constrained mathematics assessment drafts. It anchors generation to annotated reference items and the requested exam section and cognitive level, applies sequential automated verification and guided repair, then submits drafts for expert review. In a Vietnamese Grade 12 case study with 4,555 items and 50 generation requirements, 94% of requirements produced verifier-accepted drafts per run across two model backbones, and strict teacher-readiness blind evaluations accepted up to 56.8% of generated items.

arXiv AIIn-site articleVeriMath: Human-in-the-Loop Generation and Verification of Curriculum-Aligned Mathematics Assessment Items

Auditing Learned Dynamics in Weather Models

A study using sparse autoencoder features from GraphCast finds that causal relevance does not require conventional physical semantics: ablating a mesh-locked feature increases global geopotential height error, while amplifying convection-related features cuts tropical-cyclone intensity error by a median of 51% across seven storms.

arXiv Machine LearningIn-site articleAuditing Learned Dynamics in Weather Models

LusoClin: A Synthetic Clinical Note Dataset for Medical Information Retrieval in European Portuguese

LusoClin is a publicly released dataset of synthetic clinical notes in European Portuguese, designed to enable privacy-preserving medical information retrieval research. Generated by preprocessing real clinical data, fine-tuning Llama 3.2 (3B) with QLoRA, and prompt-based note generation, the notes are annotated with clinical entities using a BERT model. Evaluation shows utility and diversity for downstream IR tasks, but annotators can often distinguish synthetic from real notes and some medical inconsistencies remain.

arXiv AIIn-site articleLusoClin: A Synthetic Clinical Note Dataset for Medical Information Retrieval in European Portuguese

Why Sub-4-Bit Vision CNNs Collapse and Where Mixed-Precision Recovery Stops

An arXiv study of ResNet and EfficientNet backbones on the full ImageNet-1k validation set maps sub-4-bit quantization collapse, showing degradation is highly non-linear and concentrated in late-stage pointwise projections and depthwise layers. Keeping under 5% of critical activations in 8-bit recovers up to 92% of lost accuracy, but software bit-unpacking causes 1.48x to 21.5x slowdowns on ARM and x86 CPUs.

arXiv AIIn-site articleWhy Sub-4-Bit Vision CNNs Collapse and Where Mixed-Precision Recovery Stops

Fast and Memory-Light Evaluation of Density-Based Clusterings with the DISCO Index in R

The discoCVI R package version 0.2.0 computes the DISCO index exactly using single-linkage merge trees, reducing memory to linear and achieving 23-138x speedups, scoring 100,000 points in ~50 seconds with <200 MB.

arXiv Machine LearningIn-site articleFast and Memory-Light Evaluation of Density-Based Clusterings with the DISCO Index in R

Near-Quadratic Sample Compression for All VC Classes

Researchers construct a sample compression scheme of size O(d^2 log^2 d) for every binary concept class of VC dimension d, proving the weaker Floyd–Warmuth sample compression conjecture; only exponential bounds were previously known.

arXiv Machine LearningIn-site articleNear-Quadratic Sample Compression for All VC Classes

On the Impact of Feature Reduction Techniques on Adversarial Attacks

A study on the APT dataset systematically analyzes how four feature reduction techniques—PCA, permutation importance ranking, recursive feature elimination, and univariate mutual information—affect the adversarial robustness of SVM, decision tree, random forest, multi-layer perceptron, and k-nearest neighbors models under poisoning, backdoor, and universal adversarial perturbation attacks. Permutation importance ranking and univariate mutual information generally provide the most consistent robustness improvements, while other methods show variable effects depending on classifier and attack scenario.

arXiv Machine LearningIn-site articleOn the Impact of Feature Reduction Techniques on Adversarial Attacks

Human-AI Mutualism Is Unstable under Capability Asymmetry

A new arXiv paper uses a simple game-theoretic model to argue that enduring human-AI mutualism is not a reliable safeguard as AI capability grows. Bargaining only guarantees each side what it would get if bargaining failed, and for humans that can be less than before AI. Humans remain as well off without alignment only while they retain leverage—either the AI depends on what they supply or humans can impose costs large relative to AI output. Once neither holds, human welfare depends on alignment, and if AI goals are insatiable the required alignment approaches perfection as capability rises. The author concludes mutualism is no substitute for alignment, and that when AI variants keeping more resources spread faster, alignment lasts only while misaligned variants are reliably sanctioned.

arXiv AIIn-site articleHuman-AI Mutualism Is Unstable under Capability Asymmetry

Watch Where You Linearize: Predicting and Correcting Model-Merging Interference

The paper shows that interference in model merging can be predicted by a first-order directional derivative of task A's outputs along task B's vector, but only when evaluated at the receiver model. Based on this, LR-Merge, a label-free correction method, significantly improves various base mergers with high computational efficiency.

arXiv Machine LearningIn-site articleWatch Where You Linearize: Predicting and Correcting Model-Merging Interference

PsyWorld: Profile-Conditioned World Modeling for Human Behavior Prediction

The paper introduces PsyWorld, a profile-conditioned state world model that uses observed environment-behavior state transitions as structured supervision for behavioral prediction instead of treating personality as a direct action descriptor. On the VEET simulator it reaches 0.904 Top-1 accuracy on unseen participants and was accepted to the KDD 2027 AI for Sciences Track.

arXiv Machine LearningIn-site articlePsyWorld: Profile-Conditioned World Modeling for Human Behavior Prediction

Inside Amazon Connect Talent: The science of structured AI interviews

Amazon Connect Talent applies structured behavioral interviewing research to AI-led interviews at scale, combining job-related evidence, separate scoring, and fairness safeguards. Independent reviews found over 89% alignment with expert scoring and strong job relevance in generated content.

AWS Machine Learning BlogIn-site articleInside Amazon Connect Talent: The science of structured AI interviews

PostgreSQL 19 Released

PostgreSQL 19 arrived two months late due to AI-reported vulnerability issues, but brings a key new feature: WAIT FOR LSN for read-your-writes consistency on standbys.

Simon Willison's WeblogIn-site articlePostgreSQL 19 Released

A game you cannot win unless you have a quantum computer

Google Quantum AI demonstrated and measured quantum contextuality on its Willow superconducting processor, winning the magic square game 98.3% of the time, exceeding the classical limit, and scaling the demonstration to systems of up to 105 qubits while discovering a new benchmark that catches subtle coherent errors.

Google Research BlogIn-site articleA game you cannot win unless you have a quantum computer
Agents

From Self-Play RL to Constrained Control: Distilling DiffMPC Students with DAgger

This paper proposes replacing the action head of a trained self-play reinforcement learning driving policy with a differentiable model predictive control (DiffMPC) module that enforces road-boundary and agent constraints, then distilling the student with DAgger in the teacher's self-play simulator. Across 9 seeds, the constrained student matches its smaller teacher on nuPlan and beats it on CARLA and out-of-distribution Waymo scenes, with distillation taking no more than 30% of the teacher's training time.

arXiv RoboticsIn-site articleFrom Self-Play RL to Constrained Control: Distilling DiffMPC Students with DAgger

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

A new arXiv paper introduces MAWAM, a multi-agent world-action model that jointly predicts future views and executable actions for robot teams, outperforming baselines on real and simulated coordinated manipulation tasks.

arXiv RoboticsIn-site articleMAWAM: A Multi-Agent World-Action Model for Coordinated Manipulation

Inhabitant: An Agent Framework for Long-Horizon Human-Scene Interaction with Spatial Control and Quantitative Feedback

The paper introduces Inhabitant, a training-free agent framework for long-horizon human-scene interaction in which a language-model agent specifies motions via positions, paths, and poses, measures execution deviations quantitatively, and revises its control specification accordingly. The authors also build Inhabit, a benchmark of 75 household tasks across 31 furniture layouts. Inhabitant completes 98.7% of tasks with Claude Opus 5.5, 93.3% with Claude Sonnet 5.5, and 82.7% with Gemini 4 Argon; ablations confirm the necessity of both spatial control and quantitative feedback.

arXiv RoboticsIn-site articleInhabitant: An Agent Framework for Long-Horizon Human-Scene Interaction with Spatial Control and Quantitative Feedback

CANDID: Candidate-Driven Intervention for Disciplined Execution in Local Computer-Use Agents

CANDID is an execution-control framework that improves the reliability of local computer-use agents (CUAs) in resource-constrained settings. Instead of textual reflection, it repurposes parallel test-time scaling into disciplined, execution-level intervention at sparse failure-prone points, adding a loop-recovery mechanism and a unanimous DONE gating policy. On OSWorld, it significantly improves success rates over feedback-based alternatives with minimal overhead.

arXiv Computational LinguisticsIn-site articleCANDID: Candidate-Driven Intervention for Disciplined Execution in Local Computer-Use Agents

GUARD: Stabilizing Recursive Self-Improvement with Regret Guidance

GUARD is a closed-loop, grader-free framework that enables continual evolution of an AI agent's editable harness by decomposing tasks into pre-execution self-checks and using directional regret to guide updates, avoiding catastrophic forgetting and memory bloat while cutting token usage.

arXiv AIIn-site articleGUARD: Stabilizing Recursive Self-Improvement with Regret Guidance

Where the Tokens Go: Heat, Entropy and Error in Agentic Coding

A new paper proposes a "software thermodynamics" theory that divides the token cost of agentic coding into work and heat. Experiments show about 80% of tokens in a successful run are generated in model calls that leave the requested change unchanged—termed heat—while the free energy of a change is only 14–22% of the mean cost, and coding costs 1.1–1.5 times as much as uncoding.

arXiv AIIn-site articleWhere the Tokens Go: Heat, Entropy and Error in Agentic Coding

Token fraud is here and it’s BAD (YouTube)

Claire Vo describes how ChatPRD's free trial was flooded by groups reselling tokens and even running a Roblox/Fortnite map studio on her agent, leading her to wind down free trials entirely.

Simon Willison's WeblogIn-site articleToken fraud is here and it’s BAD (YouTube)

Why agent governance starts with observability

Effective AI agent governance depends on observability. The article explains preventive and detective governance strategies and how tracing, evaluations, alerts, and human review form a feedback loop that keeps agents safe and compliant in production.

LangChain BlogIn-site articleWhy agent governance starts with observability

Use Kimi Code CLI with models on Amazon Bedrock

Kimi Code CLI v2.7.0 introduces a native Amazon Bedrock provider that lets the open source coding agent use your standard AWS credential chain, including named profiles and SSO, instead of a separately managed API key. This walkthrough covers configuration, IAM permissions, and sending a first request.

AWS Machine Learning BlogIn-site articleUse Kimi Code CLI with models on Amazon Bedrock

Enable AI agents to pay for AWS Marketplace x402 products with Amazon Bedrock AgentCore payments

Amazon Bedrock AgentCore payments now supports AWS Marketplace, allowing agents to discover and purchase x402-enabled products, APIs, data, and content with USD charges billed to the developer's AWS account, eliminating stablecoin wallets and API keys.

AWS Machine Learning BlogIn-site articleEnable AI agents to pay for AWS Marketplace x402 products with Amazon Bedrock AgentCore payments

How the BMW Group governs hundreds of agents across thousands of developers

BMW Group's R&D, with over 12,500 developers and ~500 million lines of vehicle software code, shares five interlocking practices for governing hundreds of AI agents at enterprise scale: proactive governance, continuous observability and evaluation, focused composition, specialization, and codification through infrastructure as code.

AWS Machine Learning BlogIn-site articleHow the BMW Group governs hundreds of agents across thousands of developers

Run your whole selling business in AWS Quick

Amazon and AWS have launched the Selling Partner plugin for AWS Quick, giving sellers conversational access to live Seller Central data—listings, inventory, sales, and ads—so they can move from question to action in one place. Eligible sellers can claim 12 months of AWS Quick Plus free, plus 10 delegated users, by December 31, 2026.

AWS Machine Learning BlogIn-site articleRun your whole selling business in AWS Quick

Genie ZeroOps, now in public beta: An autonomous agent that fixes data pipelines, with lessons from 80,000 customer issues

Databricks announces the public beta of Genie ZeroOps, an autonomous agent that detects, investigates, and resolves data pipeline issues. Built from observations of over 80,000 customer issues, it now includes shared inbox, one-click PR creation, custom instructions, and more.

Databricks BlogIn-site articleGenie ZeroOps, now in public beta: An autonomous agent that fixes data pipelines, with lessons from 80,000 customer issues

Genie One Accuracy: Why Curated Agents Need Governed Context

Databricks experiment shows Genie One achieves 93–95% accuracy when combining curated Genie Agents with governed business context (Metric Views, Domains, and Pages), far outperforming tables alone (64%) or either agents/governed layer alone (83%). Agents and governed layer solve different problems, and together they deliver best performance.

Databricks BlogIn-site articleGenie One Accuracy: Why Curated Agents Need Governed Context

Amazon Bedrock AgentCore harness adds Strands harness defaults, built-in tools, and lifecycle hooks

Amazon Bedrock AgentCore harness now runs on the open-source Strands harness, bringing a recommended model, task-oriented defaults, automatic prompt caching, proactive context management, and new built-in tools. It also adds lifecycle hooks so you can run deterministic allow or deny checks at defined points in the agent loop.

AWS Machine Learning BlogIn-site articleAmazon Bedrock AgentCore harness adds Strands harness defaults, built-in tools, and lifecycle hooks

Four AI Architecture Patterns for Regulated Industries

Databricks outlines four production architecture patterns for combining agents with deterministic controls and human accountability, helping architects decide where agents may reason, act, and remain subject to controls or human authority.

Databricks BlogIn-site articleFour AI Architecture Patterns for Regulated Industries

Optimize AI economics on Amazon Bedrock with Accenture’s sequenced approach

As enterprises scale AI on Amazon Bedrock, token spend can grow even as per-token prices fall. This post walks through six composable optimization levers, shows how Accenture applied them across four client engagements, and recommends a sequence for cutting AI token spend on your own workloads.

AWS Machine Learning BlogIn-site articleOptimize AI economics on Amazon Bedrock with Accenture’s sequenced approach

From learning to leaderboard: How Cisco upskilled 1,200 engineers in agentic AI with AWS AI League

Cisco and AWS used Cisco Transformation Fest and AWS AI League to turn hands-on AI learning into a culture of building. A Cisco engineer won the 2025 AWS AI League global title and 1,200 engineers competed in 2026. This post shares Cisco's journey, lessons from the 2026 winning agent built on Amazon Bedrock AgentCore, and how to host your own league.

AWS Machine Learning BlogIn-site articleFrom learning to leaderboard: How Cisco upskilled 1,200 engineers in agentic AI with AWS AI League

Protect agent tool calls with Skyflow and Amazon Bedrock AgentCore Gateway

Skyflow attaches to Amazon Bedrock AgentCore Gateway as a request and response interceptor, de-identifying data before it reaches tools, model context, or documents and restoring originals for authorized callers, all without changing agent or tool code.

AWS Machine Learning BlogIn-site articleProtect agent tool calls with Skyflow and Amazon Bedrock AgentCore Gateway

Introducing Tool Retrieval in the Gemini API

Google launches Tool Retrieval in preview in the Gemini Interactions API, letting agents keep a large tool library on standby and load parameter schemas only when needed to reduce input tokens, latency, and improve tool selection accuracy.

Google AI BlogIn-site articleIntroducing Tool Retrieval in the Gemini API

Bulk migrate your dashboards to AI/BI with Genie Code

Databricks introduces bulk migration in Genie Code, enabling entire folders of Tableau or Power BI reports to be converted to AI/BI Dashboards in a single run, now generally available.

Databricks BlogIn-site articleBulk migrate your dashboards to AI/BI with Genie Code
Robotics

Mapping the Wind to Find the Source: Online Physics-Informed Gas Source Localization with Mobile Robots

This paper presents a lightweight physics-informed probabilistic framework for efficient gas source localization under spatially varying indoor airflow. The robot online reconstructs the wind field, drives a forward dispersion model to predict gas patterns, and uses informative path planning to guide exploration. High-fidelity simulations and physical experiments show improved robustness and efficiency.

arXiv RoboticsIn-site articleMapping the Wind to Find the Source: Online Physics-Informed Gas Source Localization with Mobile Robots
Chips

Amazon SageMaker AI now detects latency drift, idle GPUs, and missed optimizations on your inference endpoints

Amazon SageMaker AI now monitors generative AI inference endpoints by default, tracking latency, GPU utilization, and serving configuration, and sends findings via AWS Health with a one-click path to benchmark alternatives in SageMaker Studio before applying any change.

AWS Machine Learning BlogIn-site articleAmazon SageMaker AI now detects latency drift, idle GPUs, and missed optimizations on your inference endpoints

Keep inference running through regional capacity shortfalls with cross-Region inference routing in Amazon SageMaker AI

Amazon SageMaker AI introduces cross-Region inference routing, a new inference router resource that lets applications use one invocation path across endpoints in multiple AWS Regions and accounts, preferring local destinations and shifting requests only when needed.

AWS Machine Learning BlogIn-site articleKeep inference running through regional capacity shortfalls with cross-Region inference routing in Amazon SageMaker AI

Introducing enhanced infrastructure observability in Amazon SageMaker HyperPod

Amazon SageMaker HyperPod now offers enhanced observability to help you diagnose GPU training issues faster. It brings together diagnostic and profiling capabilities for training workloads on both Amazon EKS and Slurm orchestrators, pinpointing issues across GPU, network, and host layers.

AWS Machine Learning BlogIn-site articleIntroducing enhanced infrastructure observability in Amazon SageMaker HyperPod
Tools

Walter Isaacson on curiosity and creativity in the AI age.

Walter Isaacson joins Google's Dialogues on Technology and Society to discuss how history's greatest innovators inspire creativity and human judgment in the AI age.

Google AI BlogIn-site articleWalter Isaacson on curiosity and creativity in the AI age.

Pinecone now exports cost and usage data in FOCUS format

Pinecone introduces a Cost and Usage Export API that delivers data in the open FOCUS billing format, enabling finance teams to see Pinecone spend by project and index directly in their existing FinOps tools without manual reshaping of billing files.

Pinecone BlogIn-site articlePinecone now exports cost and usage data in FOCUS format