Skip to content
AI News HubLIVE

This edition’s highlights

Agents

Managing frontier AI risk: From training to deployment

  • Defense in depth with four layers: model alignment, system-level controls, monitoring, and sandboxing.
  • Different layers emphasized at research, internal deployment, and external deployment stages.
OpenAI NewsIn-site articleManaging frontier AI risk: From training to deployment

Managed NVIDIA Dynamo on Amazon SageMaker HyperPod Inference

  • HyperPod Inference add-on installs Dynamo CRDs and serving components, including the Inference Operator, Dynamo operator, Grove, and KAI Scheduler.
  • Two deployment paths: bring-your-own DynamoGraphDeployment (DGD) or InferenceEndpointConfig with dynamoSpec.
AWS Machine Learning BlogIn-site articleManaged NVIDIA Dynamo on Amazon SageMaker HyperPod Inference

Amazon SageMaker AI real-time endpoints now support KV cache aware routing

  • KV-aware routing uses cache events reported by the model container to route requests to the instance with the longest cached match and available capacity.
  • It differs from prefix-aware routing by seeing actual cache contents, handling evictions, and matching full multi-turn histories.
AWS Machine Learning BlogIn-site articleAmazon SageMaker AI real-time endpoints now support KV cache aware routing
Models

Introducing GPT-6.2 Astra and Sol

  • GPT-6.2 Astra scores 5.2 points above GPT-6 Astra on Agents' Last Exam at under a third of the cost
  • Astra's API input and output prices are 20% lower than GPT-6 Astra, with a 60% cached-input discount
OpenAI NewsIn-site articleIntroducing GPT-6.2 Astra and Sol

GPT‑6.2 Astra and GPT‑6.2 Sol

  • GPT‑6.2 Astra costs $8/$40 per million input/output tokens, 80% of GPT‑6 Astra; cached input drops to $0.40/million.
  • GPT‑6.2 Sol remains at $2/$10 per million tokens, half the price of GPT‑5.6 Sol.
Simon Willison's WeblogIn-site articleGPT‑6.2 Astra and GPT‑6.2 Sol

Kimi K3.1 Pro vs K3 and Claude Opus 5.5: Should You Switch?

  • K3.1 Pro outperforms K3 on 14/17 benchmarks but scores 22% lower on memory creation and 64% lower on contract review.
  • In banking customer support, K3.1 Pro resolves 13% more requests than Opus 5.5 but costs 89% more per success and has over 2x latency.
Fireworks AI BlogIn-site articleKimi K3.1 Pro vs K3 and Claude Opus 5.5: Should You Switch?

Baseten is now available on Microsoft Foundry

  • Baseten is in public preview on Microsoft Foundry, with general availability expected in early 2027
  • Launch models include DeepSeek V4.1 Flash, GLM-5.3, GLM-5.3-Flash, and Kimi K3
Baseten BlogIn-site articleBaseten is now available on Microsoft Foundry
Chips

NVIDIA RTX Spark Compact Desktops Out Now

  • RTX Spark desktops feature a 20-core Grace CPU and Blackwell RTX GPU with 6,144 CUDA cores, delivering up to 1 petaflop of AI performance.
  • Systems can run 120-billion-parameter LLMs locally, render 90GB+ 3D scenes in real time, and edit 4:2:2 video smoothly.
NVIDIA BlogIn-site articleNVIDIA RTX Spark Compact Desktops Out Now
Startups

How we raised Databricks cluster rate limits by 7.5×

  • Default cluster-creation request-rate limits raised by 7.5× and over 100 customer-specific overrides removed.
  • Adaptive concurrency limits replace fixed rate limits to better handle long-running operations.
Databricks BlogIn-site articleHow we raised Databricks cluster rate limits by 7.5×
Policy

The new Azure Databricks on Azure Government is now in Beta

  • New Azure Databricks on Azure Government enters Beta with the same architecture as commercial Azure Databricks.
  • Beta includes Unity Catalog, Databricks SQL, serverless compute, and Compliance Security Profile enabled by default.
Databricks BlogIn-site articleThe new Azure Databricks on Azure Government is now in Beta

You have reached the end of this edition.

Your reading list
Other updates (82)
Agents

Quoting Daniel Stenberg

curl creator Daniel Stenberg notes that while AI coding agents converted 185,000 lines of C into 340,000 lines of Rust over a little more than three weeks, the real curl project merged 300+ commits, shipped 200+ bugfixes and fixed several vulnerabilities that the rewritten clones lack.

Simon Willison's WeblogIn-site articleQuoting Daniel Stenberg

LangSmith for Coding Agents: Now track spend by outcome

LangSmith introduces coding-agent-specific reporting that links agent spend to GitHub pull requests or GitLab merge requests, helping teams calculate coding agent ROI. The feature is in public beta and supports Claude Code, Codex, Cursor, Copilot, and Deep Agents.

LangChain BlogIn-site articleLangSmith for Coding Agents: Now track spend by outcome

Inside the Codex sandbox

Codex sandbox uses OS-native facilities to constrain commands in the user's existing environment, balancing meaningful work with filesystem and network limits.

OpenAI NewsIn-site articleInside the Codex sandbox

RoboCoEvo: Expanding Robot Capabilities through Agent-Policy Co-Evolution

RoboCoEvo introduces a failure-guided co-evolution framework that couples the reusable executable knowledge refined by an MLLM-based agent with parameter learning in robotic foundation models such as VLA models. Using a code-as-policy formulation, the agent writes and iteratively improves failure-detection and recovery programs that yield targeted corrective experience; post-training then consolidates those behaviors into the robot's end-to-end policy, and the updated policy's failure distribution guides the next round of program revisions. Experiments on BEHAVIOR-1K, RoboCasa, and RoboTwin show sustained gains in autonomous VLA task performance across successive co-evolution rounds.

arXiv RoboticsIn-site articleRoboCoEvo: Expanding Robot Capabilities through Agent-Policy Co-Evolution

FieldGuide: A Corpus Map for Token-Efficient Agentic Document Exploration

This paper introduces FieldGuide, an agent-facing corpus map that organizes a document collection into a topic hierarchy and provides tools for corpus orientation, scoped search, and bounded reading. Evaluated on five journalistic document-exploration tasks using 38 Chicago Police Department public-records documents, FieldGuide used 27% fewer tokens and 41% less elapsed time than raw-file access, and 38% fewer tokens and 21% less elapsed time than RAG. An accuracy audit found four minor issues, comparable to other conditions. In blind ratings of generated reporting leads, FieldGuide scored higher than RAG but lower than raw-file access, suggesting that selective access may limit broad cross-document discovery.

arXiv Computational LinguisticsIn-site articleFieldGuide: A Corpus Map for Token-Efficient Agentic Document Exploration

DiscoveryAgent: A Framework for Scientist-centric Agentic AI

Researchers present DiscoveryAgent, a domain-agnostic, scientist-centric agentic research system that pairs autonomous investigation with scientist-directed hypothesis steering and rigorous checkpointing that enables scalable audit-agent oversight.

arXiv AIIn-site articleDiscoveryAgent: A Framework for Scientist-centric Agentic AI

Regularized Score Centering

Regularized Score Centering (RSC) adds a reverse KL penalty to score centering to keep the learner close to the behavior policy in asynchronous RL. Theoretical analysis shows improved convergence, and experiments on math reasoning and six agentic tasks show substantial sample efficiency gains, with 14.3% improvement on AppWorld and 9.9x the strongest baseline on BabyAI.

arXiv Machine LearningIn-site articleRegularized Score Centering

DeltaHarness: Self-Evolving Harnesses for Resource-Adaptive Tool Execution

DeltaHarness is an execution-guided policy-search method that adapts agent harnesses from cold-start measurements, using a QHI loop to update runtime dispatch and revise plans, significantly reducing latency in tasks like video editing.

arXiv Machine LearningIn-site articleDeltaHarness: Self-Evolving Harnesses for Resource-Adaptive Tool Execution

Le Groupe Figaro and OpenAI partner to bring Le Figaro journalism to ChatGPT

Le Groupe Figaro and OpenAI have announced a strategic partnership to make Le Figaro's journalism discoverable in ChatGPT through summaries and excerpts with links to original stories. The collaboration covers current reporting and archive content across news, business, culture and lifestyle, while Le Figaro retains full editorial independence.

OpenAI NewsIn-site articleLe Groupe Figaro and OpenAI partner to bring Le Figaro journalism to ChatGPT

External sandboxes for coding agents

Simon Willison and Prime Radiant introduce the external sandbox pattern, where agent-executed code runs in a dedicated sandbox while the agent stays outside, releasing three server implementations supporting Pi, Claude Code, and Codex to safely test local models and defend against prompt injection.

Simon Willison's WeblogIn-site articleExternal sandboxes for coding agents

Reducing token costs and latency in Amazon Bedrock AgentCore JSON workloads

This post shares three composable strategies—offloading heavy artifacts to Amazon S3, on-demand result paging, and dictionary compaction—used to build a production agent handling 20 MB+ JSON payloads. These optimizations reduced token consumption by up to 71% and average latency by up to 80% in benchmark runs.

AWS Machine Learning BlogIn-site articleReducing token costs and latency in Amazon Bedrock AgentCore JSON workloads

Cut generative AI costs: Code for rules, models for judgment

The biggest cost lever in a generative AI system is whether you call a model at all. Learn how to decide whether each workflow step needs a model, and map each option to AWS: deterministic code, a constrained decision call, or a right-sized model on Amazon Bedrock.

AWS Machine Learning BlogIn-site articleCut generative AI costs: Code for rules, models for judgment

Govern and discover records with common metadata in AWS Agent Registry

Custom metadata schemas in AWS Agent Registry let administrators define typed, validated fields that every agent, MCP server, and skill must carry. The post follows an MCP record from schema definition through registration, enrichment, approval, and discovery.

AWS Machine Learning BlogIn-site articleGovern and discover records with common metadata in AWS Agent Registry

Secure OAuth patterns for Snowflake Cortex AI through Amazon Bedrock AgentCore

This post explores how an AI agent on Amazon Bedrock AgentCore authenticates to Snowflake Cortex AI over MCP using three OAuth patterns tested on AgentCore Identity: Snowflake OAuth, three-legged OAuth with a corporate IdP, and two-legged OAuth, helping organizations match identity, audit, and compliance needs.

AWS Machine Learning BlogIn-site articleSecure OAuth patterns for Snowflake Cortex AI through Amazon Bedrock AgentCore

Announcing AI data preparation capabilities in Amazon SageMaker: Build high-quality datasets for model development workflows

Amazon SageMaker AI now offers AI data preparation capabilities for building high-quality training and evaluation data, including synthetic data generation, dataset assessment, auto-labeling at scale, and expert review, accessible through Python SDK APIs and an agent skill from your notebook or IDE.

AWS Machine Learning BlogIn-site articleAnnouncing AI data preparation capabilities in Amazon SageMaker: Build high-quality datasets for model development workflows

Getting AI out of the demo and into businesses globally

AWS Forward Deployed Engineering is now available in 15 countries, helping enterprises move AI from prototype to production in 45-day sprints while leaving them with a foundation, workflows, and governance they own.

AWS Machine Learning BlogIn-site articleGetting AI out of the demo and into businesses globally

Branch, sync, serve: What our Lakebase Hackathon taught us about building on Lakebase

Databricks' Lakebase hackathon showcased 11 projects across two tracks: building applications with database branching and connecting Lakebase to the lakehouse for real-time serving. The projects demonstrate how Lakebase, as fully managed Postgres under Unity Catalog governance, enables AI agent testing, per-PR preview environments, real-time pricing, and conversational shopping assistants.

Databricks BlogIn-site articleBranch, sync, serve: What our Lakebase Hackathon taught us about building on Lakebase

Build WhatsApp AI agents with Amazon Bedrock AgentCore

SMART, an open-source Python SDK and Terraform stack, deploys the full WhatsApp channel layer to your AWS account with one command. Bring any agent that runs on Amazon Bedrock AgentCore, and SMART handles webhook ingestion, media, conversation memory, scheduling, and Meta message templates.

AWS Machine Learning BlogIn-site articleBuild WhatsApp AI agents with Amazon Bedrock AgentCore

Protect sensitive data in your traces: new controls in LangSmith

LangSmith introduces three new controls to help teams manage access to sensitive trace data, enabling debugging without exposing customer information through granular permissions.

LangChain BlogIn-site articleProtect sensitive data in your traces: new controls in LangSmith

How LoanPro built an agentic AI loan servicing assistant with Amazon Bedrock AgentCore

LoanPro worked with AWS Partner AllCloud and the AWS Generative AI Innovation Center to build a multi-agent AI assistant on Amazon Bedrock and Amazon Bedrock AgentCore. It gives loan servicing agents instant borrower context and next-best-action recommendations, reducing customer call times by up to 15 percent.

AWS Machine Learning BlogIn-site articleHow LoanPro built an agentic AI loan servicing assistant with Amazon Bedrock AgentCore

Put Gemini to work for you

Tasks in Gemini is now free of charge in six countries, with new features starting in the U.S. and deeper Google Workspace integration to get things done on your behalf.

Google AI BlogIn-site articlePut Gemini to work for you

Responsible AI for agentic AI: A phased roadmap for platform and application teams

Enterprises adopting agentic AI at scale need platforms that embed responsible AI from the start. This post gives platform and application teams a three-phase roadmap (governance foundation, pilot deployment, and evaluation and hardening) that maps the AWS Responsible AI Lens pillars to each phase, with guidance on when to revisit each one.

AWS Machine Learning BlogIn-site articleResponsible AI for agentic AI: A phased roadmap for platform and application teams

What We’ve Learned Building Proactive Agents

LangChain shares lessons from building its GTM Agent and Robin: memory and review-friendly UX are what let proactive agents check in when it matters instead of becoming noise.

LangChain BlogIn-site articleWhat We’ve Learned Building Proactive Agents

ElevenLabs acquires Kotoba to advance Japanese voice AI

ElevenLabs is acquiring Kotoba, the leading Japanese voice AI lab, to accelerate AI agent and voice interface adoption in Japan. Kotoba's Koto model excels in Japanese TTS, STT, and simultaneous interpretation, and is already used by enterprises like Salesforce. The deal is expected to close in Q4 2026.

ElevenLabs BlogIn-site articleElevenLabs acquires Kotoba to advance Japanese voice AI
Models

How AUMOVIO centralized AI access with Databricks Unity Gateway

AUMOVIO adopted Databricks Unity Gateway as a central AI gateway to unify fragmented LLM access, achieving cost transparency, centralized governance, and provider flexibility. The article shares its decision process, rationale, and three core capabilities built on top.

Databricks BlogIn-site articleHow AUMOVIO centralized AI access with Databricks Unity Gateway

Certify, Then Select: Viability-Aware Runtime Steering of Frozen Diffusion Policies

The paper proposes CAST (Certified Action Selection at Test time), which samples action chunks from a frozen diffusion policy, certifies each with a barrier condition, and executes the certified chunk ranked highest by a learned value ensemble. It also introduces viability estimation to avoid dead ends that one-step certification cannot see, improving joint task-and-constraint success across multiple robotic platforms.

arXiv RoboticsIn-site articleCertify, Then Select: Viability-Aware Runtime Steering of Frozen Diffusion Policies

Robustness as a Transferable Skill: Cross-Task Adversarial Defense for Vision-Language-Action Policies

This paper asks whether adversarial robustness learned on source manipulation tasks can protect independently specialized target VLA policies without target-side adversarial training, proposing SN-TRV to transfer source defense updates to the target visual encoder. It reports improved robustness across three VLA backbones and attack settings, and lifts average physical PiPER success under printed adversarial patches from 20.3% to 45.8%.

arXiv RoboticsIn-site articleRobustness as a Transferable Skill: Cross-Task Adversarial Defense for Vision-Language-Action Policies

Consistency Is Not Correction: Separating Confidence, Abstention, and Recovery in LLM Agents

The paper argues that ranking, selective prediction, and correction utility are separate claims about a confidence signal, derives a pre-experiment bound on when gated correction can be detected, and shows empirically that many reported gains are capped by measurement resolution rather than signal quality.

arXiv Computational LinguisticsIn-site articleConsistency Is Not Correction: Separating Confidence, Abstention, and Recovery in LLM Agents

From Preference Data to Personalization: Tracing Sycophancy in Large Language Models

This paper reframes sycophancy as "inappropriate agreement" and introduces SycoDrift, an evaluation set of 3,298 validity-annotated probes. It analyzes three pressure sources: human preference data, automated judge specification, and in-context personalization. Findings show weak and mostly appropriate preference for validation in human datasets, modest sycophancy increases from emphasized judge rubrics, and notable increases in inappropriate agreement from inference-time personalization, especially with matched user profiles placed before the query.

arXiv Computational LinguisticsIn-site articleFrom Preference Data to Personalization: Tracing Sycophancy in Large Language Models

Literary Language Mashup: Curating Fictions with Large Language Models

The paper proposes the GrAImes framework to use LLMs as auxiliary evaluators for literary fiction. Comparing LLM and human assessments of Spanish microfiction and short stories, it finds that LLMs cannot replace human judges (Krippendorff's alpha < 0.66) but can offer an initial perspective on editorial quality.

arXiv Computational LinguisticsIn-site articleLiterary Language Mashup: Curating Fictions with Large Language Models

Sketch, Don't Paint: Allocating Test-Time Compute in Generative Visual Reasoning

This paper studies how multimodal models should allocate test-time compute in visual reasoning: when to render visual thoughts and how many denoising steps each thought needs. It finds that visual thoughts fall into causal, decorative, and harmful regimes, that 8-step sketches often match 50-step renders, and proposes sketch-and-vote, which beats full rendering with fewer denoising steps across several reasoners.

arXiv Computer VisionIn-site articleSketch, Don't Paint: Allocating Test-Time Compute in Generative Visual Reasoning

Can an LLM dream? Testing the overfitted brain hypothesis on an agent with text memory

A study tests the overfitted brain hypothesis on a text-memory LLM agent: across 10 simulated days and 30 runs, corrupting conversation transcripts did not improve transfer, 40% detail corruption lowered recall from 76% to 60%, and random-text corruption had no consistent effect because the model recognized and excluded it.

arXiv Computational LinguisticsIn-site articleCan an LLM dream? Testing the overfitted brain hypothesis on an agent with text memory

Limite 1B - Violetto: Technical Report

Researchers present Limite 1B - Violetto, a one-billion-parameter model built for high-throughput mathematical reasoning. It was trained from scratch on 237 billion tokens from a curated mathematics-centric corpus and supports contexts up to 131k tokens. Post-training combines knowledge distillation, targeted supervised fine-tuning, and RLVR on a curated curriculum. The paper reports 94.01% on AIME 2026 and 74.25% on BeyondAIME, placing it among much larger models on the team's evaluation suite. The model is only lightly instruction-tuned and is positioned as a single-turn math solver and a fast parallel component of larger systems rather than a general assistant. The model, base and value checkpoints, and an optimized serving implementation are released under Apache 2.0.

arXiv Computational LinguisticsIn-site articleLimite 1B - Violetto: Technical Report

Every Camera Grows Up: Camera-Height Regression in Interactive Video World Models

Six recent open-weight interactive video world models fail to maintain the camera height set by the starting image, drifting toward human eye level as motion replaces the initial view. The EyeHeight benchmark reveals this camera-height regression, traces it to human-centric video priors, and shows that viewpoint drift alone can shift visual navigation policy predictions.

arXiv Computer VisionIn-site articleEvery Camera Grows Up: Camera-Height Regression in Interactive Video World Models

EgoLink: Egocentric Language-Vision Interactive Network Knowledge Challenge Summary

The EgoLink Challenge at ACM Multimedia 2026 benchmarks egocentric multimodal intelligence linking first-person perception, social reasoning, and interactive action. It features two tracks: social reasoning in egocentric video and interactive agents completing service tasks via dialogue, tool use, and closed-loop execution. The best Track 1 system achieved 0.9263 accuracy, while the best Track 2 system achieved 0.844 joint success and 0.916 micro accuracy. Participant reports reveal a shared principle: strong systems make intermediate evidence explicit through candidate verification, option-robust reasoning, scenario-aware planning, tool canonicalization, and state-aware correction before answering or acting.

arXiv Computer VisionIn-site articleEgoLink: Egocentric Language-Vision Interactive Network Knowledge Challenge Summary

RAW: Refusal Aversion in the Wild, A Causal Measurement Method for Deployed LLMs

Researchers introduce RAW, a method that uses LLM stochasticity as a natural experiment to causally measure how refusals affect user re-engagement from existing logs; across 15 deployed models, about half show a 3–17 pp re-engagement cost per refusal, at roughly $0.30 per 1,000 conversations.

arXiv Computational LinguisticsIn-site articleRAW: Refusal Aversion in the Wild, A Causal Measurement Method for Deployed LLMs

Depth-Aware Activation Steering via Attention Residuals

This paper combines Attention Residuals' learnable cross-layer pathways with Jacobian-lens concept readout into a lightweight depth-aware steering pipeline, showing on Qwen3-0.6B and Qwen3.5-0.8B that distributing a persona vector by depth weights yields stronger internal alignment and behavioral shifts than single-layer injection, alongside findings on geometric decomposition, an intensity window, and metric discipline.

arXiv Computer VisionIn-site articleDepth-Aware Activation Steering via Attention Residuals

Learning Calibrated and Samplable Predictive Distributions of Surface-Ocean ΔpCO₂ Using Tabular Diffusion Models

The ocean absorbs nearly 30% of anthropogenic CO₂ emissions, but sparse observations make estimating surface-ocean CO₂ challenging. Most existing ML estimates are deterministic, with uncertainty approximated post hoc. This paper uses Treeffuser, a conditional diffusion model with gradient-boosted-tree score functions, to learn the full predictive distribution of ΔpCO₂ from 13 environmental, spatial, and temporal predictors. On a held-out SOCAT test set, Treeffuser achieves the lowest continuous ranked probability score (8.229 μatm) and coverage closest to nominal 68%/95% levels (0.686/0.930), while retaining point-prediction skill comparable to a deterministic XGBoost baseline (RMSE 18.4 vs. 17.7 μatm; R² 0.77 vs. 0.79). Results show conditional diffusion can provide well-calibrated, obse…

arXiv Machine LearningIn-site articleLearning Calibrated and Samplable Predictive Distributions of Surface-Ocean ΔpCO₂ Using Tabular Diffusion Models

Conserved Verdicts, Drifting Commitments: Identifying Harmful Overthinking

An arXiv paper introduces a commitment-versus-verdict framework that separates the correctness of a model's current answer from the correctness of its final answer, shows that final-correctness probes alone cannot decide when to stop reasoning, and finds that continuing to think costs an average of 3.6 points at points where DeepSeek-R1-Distill-Qwen-1.5B already holds a correct answer on MATH-500.

arXiv AIIn-site articleConserved Verdicts, Drifting Commitments: Identifying Harmful Overthinking

Margin Dynamics for Large Language Model Alignment

This paper provides a theoretical foundation for LLM alignment methods by modeling pairwise preference methods as vector fields on the probability simplex. Closed-form margin ODE solutions reveal unbounded DPO margins, finite IPO margins, a label-noise phase transition in DPO, and gradient concentration effects on trainability. The theory guides method design, yielding TPO-S, a label-smoothed SimPO variant validated across three 8B-9B models.

arXiv Machine LearningIn-site articleMargin Dynamics for Large Language Model Alignment

Hybrid Quantum Selection of Learning Rates in Language-Model Training: Termination Criteria and an Interaction-Guided Schedule Oracle

This paper studies quantum minimum finding for layer-group learning-rate selection in a hybrid classical-quantum training framework, comparing fixed-budget and confidence-based stopping policies and introducing an Interaction-Guided Schedule Oracle that scores joint schedules in superposition without a precomputed score table. In closed-loop fine-tuning of a 1.5-billion-parameter language model, each selected schedule drives 200 classical steps and benchmark accuracies fall within the range seen across five uniform-rate seeds.

arXiv Machine LearningIn-site articleHybrid Quantum Selection of Learning Rates in Language-Model Training: Termination Criteria and an Interaction-Guided Schedule Oracle

FACE: Flow-Guided Angular Channel Estimation Framework for mmWave MIMO Systems

The paper proposes FACE, a flow-matching-based channel estimation framework for mmWave MIMO. FACE exploits angular-domain sparsity and uses a learned flow model as a generative channel prior, outperforming conventional and diffusion-based estimators while requiring fewer pilots, fewer sampling steps, and lower inference cost.

arXiv Machine LearningIn-site articleFACE: Flow-Guided Angular Channel Estimation Framework for mmWave MIMO Systems

Language Models Can Rewrite Their Own Weights

A new arXiv paper introduces "self-rewriting": a 27B model diagnoses its own failures and directly edits its weights without gradients or training, gaining 6.6 points on 490 unseen problems and repairing a damaged copy of itself. The ability emerges with scale—only the 27B model succeeds, while 2B makes no edit and 4B/9B break themselves.

arXiv AIIn-site articleLanguage Models Can Rewrite Their Own Weights

Chain-of-Thought Controllability Can Reduce Chain-of-Thought Monitorability

A new study shows that persona-control training can teach models to control their chain-of-thought, reducing monitorability of hidden tasks. Experiments with gpt-oss-20b and Kimi-K2.6 model organisms reveal significant drops in monitorability metrics, though targeted transparency prompts can partially recover it.

arXiv AIIn-site articleChain-of-Thought Controllability Can Reduce Chain-of-Thought Monitorability

AdaFlowLite: Scalable and Non-blocking Inference on Asynchronous Mobile Data

AdaFlowLite introduces opportunistic inference for asynchronous multi-modal mobile data, using a hierarchical normalized matrix to model cross-modality affinity and a multi-modal lightweight Swin Transformer for real-time data imputation, achieving significant reductions in latency and energy consumption while improving accuracy.

arXiv Machine LearningIn-site articleAdaFlowLite: Scalable and Non-blocking Inference on Asynchronous Mobile Data

NEBench R1: a Reasoning Benchmark for Evaluating Nuclear Domain Competence in Large Language Models

NEBench R1 is a 7,546-question reasoning benchmark drawn from NRC reactor operator licensing exams across 18 nuclear facilities. Six LLMs ranged from 34.0% to 73.9% accuracy, and correct answers often came with weak factual and procedural justification; longer reasoning traces correlated with lower accuracy in frontier reasoning models.

arXiv AIIn-site articleNEBench R1: a Reasoning Benchmark for Evaluating Nuclear Domain Competence in Large Language Models

Azure Databricks sessions at Microsoft Ignite 2026

Azure Databricks is a Visionary-level sponsor at Microsoft Ignite 2026, November 17–20 at Moscone Center in San Francisco. Eleven sessions will cover governed AI, AI agents, model fine-tuning, and Genie in Microsoft Teams and Microsoft Copilot Cowork, alongside booth demos and executive meetings.

Databricks BlogIn-site articleAzure Databricks sessions at Microsoft Ignite 2026

Reason over time-series data with OpenTSLM on Amazon SageMaker JumpStart

OpenTSLM, an open source family of time-series language models, is now available in Amazon SageMaker JumpStart. Nexthink uses it for proactive IT support with its Spark agent, while Aionic Labs provides the TimeNet data infrastructure. Learn how to deploy OpenTSLM and fine-tune it on your own data.

AWS Machine Learning BlogIn-site articleReason over time-series data with OpenTSLM on Amazon SageMaker JumpStart
Research

How Sumitomo Life’s leaders helped teams adopt ChatGPT Enterprise

Sumitomo Life uses ChatGPT Enterprise to research insurers, prepare proposals, and draft code for investment analysis. Early use by executives helped drive adoption.

OpenAI NewsIn-site articleHow Sumitomo Life’s leaders helped teams adopt ChatGPT Enterprise

Research acceleration: The jagged frontier

OpenAI measures how coding agents are accelerating AI research, where their performance remains uneven, and what this signals about recursive self-improvement.

OpenAI NewsIn-site articleResearch acceleration: The jagged frontier

Correcting Persistent Failures in Point-Cloud Grasping Policies using Human Demonstrations in Virtual Reality

The paper introduces a human-in-the-loop failure loop that automatically mines situations where a reinforcement learning grasping policy persistently fails, sends them to a VR application for human demonstration, and returns the grasp points to training. In simulation, RL raises success on unseen objects from 55% to 83%, and adding 58 human demonstrations nearly doubles success on hard cases from 24% to 45% without reducing generalization.

arXiv RoboticsIn-site articleCorrecting Persistent Failures in Point-Cloud Grasping Policies using Human Demonstrations in Virtual Reality

Can a World Model Predict a Deepfake Video? A Consistency Margin Prior for Deepfake Detection

This paper leverages the predictive video model V-JEPA 2.1 to compute prediction residuals as a consistency-margin prior for cross-dataset deepfake detection. With only 8.25M trainable parameters, the model achieves 90.5 AUC on Celeb-DFv2 when trained solely on FF++.

arXiv Computer VisionIn-site articleCan a World Model Predict a Deepfake Video? A Consistency Margin Prior for Deepfake Detection

Rethinking Perceptual Similarity for Remote Sensing: Multi-Layer Attention Scoring and Human Preference Benchmarking

This paper introduces Remote Sensing Perceptual Distance (RSPD), a human-aligned perceptual metric for generated remote sensing imagery based on multi-layer attention scores from RemoteCLIP, along with a blind 2AFC benchmark and an ordered misregistration benchmark. RSPD substantially outperforms simple cosine distance on realistic RS generation tasks and is competitive with the strongest generic baselines.

arXiv Computer VisionIn-site articleRethinking Perceptual Similarity for Remote Sensing: Multi-Layer Attention Scoring and Human Preference Benchmarking

Data-Driven Variable-Length Compression of 3D Representations

The paper proposes a learned variable-length compression method that exploits structure across a training dataset while adapting the compressed representation size to each shape's complexity through fast feed-forward inference, improving rate-distortion for both meshes and signed distance fields.

arXiv Computer VisionIn-site articleData-Driven Variable-Length Compression of 3D Representations

Video recordings and play behaviour annotations of group-housed dairy calves on two Dutch farms

A public dataset of 1,173 video files spanning 651 hours from four camera deployments on two Dutch farms, with 104 hours continuously coded by trained observers using an 18-item ethogram, plus 1.1 million decoded frames, per-frame timestamps, per-calf segmentation tracks, and frame-level play labels, supporting calf behaviour and welfare studies and automated behaviour recognition.

arXiv Computer VisionIn-site articleVideo recordings and play behaviour annotations of group-housed dairy calves on two Dutch farms

MInfoVisQA: A Multilingual Benchmark for Information-Rich Visual Question Answering

Researchers introduce MInfoVisQA, a parallel benchmark with 8,960 QA configurations across 24 languages, to test whether multimodal LLMs keep a consistent understanding of information-rich visuals when the surface language changes. Evaluating 25 models shows that average accuracy hides large language-robustness gaps, and that every model performs worse with non-English visuals under fixed English QA.

arXiv Computational LinguisticsIn-site articleMInfoVisQA: A Multilingual Benchmark for Information-Rich Visual Question Answering

Sleep2Word: Peripheral-to-EEG Translation via Duration-Aware Wordization

Sleep2Word translates respiration, ECG, or PPG into EEG spectrograms using duration-aware physiological words, trained on over 20K PSG cases across 9 datasets, outperforming baselines in spectrogram reconstruction, sleep staging, arousal detection, and survival prediction while compressing full-night sequences by up to 69.7%.

arXiv Computational LinguisticsIn-site articleSleep2Word: Peripheral-to-EEG Translation via Duration-Aware Wordization

Edge-Adaptive Hiera-Seg Model for Building-Integrated Photovoltaic Power Forecasting

This paper proposes an edge-adaptive HieraSeg model for short-term BIPV power forecasting using multi-orientation visual-irradiance fusion, validated on a simulated dataset to handle occlusion, orientation, and edge device issues.

arXiv Computer VisionIn-site articleEdge-Adaptive Hiera-Seg Model for Building-Integrated Photovoltaic Power Forecasting

CrisisBench: an evaluation suite for crisis counseling AI

CrisisBench is a privacy-preserving evaluation suite for crisis counseling AI, featuring single- and multi-turn tasks and simulated patient LLMs. Evaluations show all systems underperform human counselors on excessive empathy and questioning, though they excel at core structural competencies.

arXiv Computational LinguisticsIn-site articleCrisisBench: an evaluation suite for crisis counseling AI

SLED: Sensor SLicing from Edge to Datacenter

The paper introduces Sensor Slicing (SL), a MAC-level isolation mechanism for individual sensors within a single UE in O-RAN, plus SLED, an end-to-end framework spanning edge, campus, and remote HPC tiers that keeps latency-critical telemetry stable where conventional network slicing fails.

arXiv AIIn-site articleSLED: Sensor SLicing from Edge to Datacenter

DNeo: Your Parallel Draft Model Needs Better Target Feature Encoding

Researchers propose DNeo, an architecture that uses Mixture-of-Features Attention (MoFA) and rejection memory to address the lossy, memory-less, and uniform-sampling limitations of standard feature encoding, achieving 1.3x faster generation than the state-of-the-art DFlash-2 without increasing drafter footprint or affecting quality.

arXiv AIIn-site articleDNeo: Your Parallel Draft Model Needs Better Target Feature Encoding

Layerwise Learning-Rate Scaling in Vision Networks: Effective-Rate Identification, Power-Mean Aggregation, and Budget Dependence

An arXiv paper shows that the rate-related effect of layerwise learning-rate strategies can be summarized by an effective learning rate computed directly from the multiplier profile. Test-accuracy changes closely track training-loss changes (r = -0.996), and a parameter-weighted power mean with exponent around 0.35 explains more within-architecture variance than the arithmetic mean. Ranking candidates by this scalar without training improves the selected candidate's mean percentile rank from 26th to 10th and cuts its accuracy shortfall from 3.70 to 1.12 points. The effect decays as training budgets grow.

arXiv Machine LearningIn-site articleLayerwise Learning-Rate Scaling in Vision Networks: Effective-Rate Identification, Power-Mean Aggregation, and Budget Dependence

TIBECG: Temporal Information Bottleneck for Lead-Specific Demographic Privacy in ECG Waveform Release

Researchers propose TIBECG, an architectural temporal information bottleneck that reduces leakage of demographic information such as age and sex when releasing individual ECG leads. Across 6,386 recordings, 12 leads, and 21 latent lengths, the method retained at least 95% of clean macro-F1 while reducing age/sex interval gain by a median 22.08 percentage points versus the best eligible conventional transformation. Direct downsampling usually suppresses demographic inference more strongly but at a larger diagnostic utility loss; TIBECG favors utility preservation. The findings support lead-specific evaluation rather than validated consumer deployment or expert-readable reconstruction.

arXiv Machine LearningIn-site articleTIBECG: Temporal Information Bottleneck for Lead-Specific Demographic Privacy in ECG Waveform Release

Building new research tools for Maya epigraphy on Modal

Miguel Novelo, who grew up surrounded by Maya glyphs in Mexico's Yucatán peninsula, long thought he could never understand them. Now he is building new research tools on Modal to help decode this ancient writing system.

Modal BlogIn-site articleBuilding new research tools for Maya epigraphy on Modal

Running the full server-side Datasette Python app in the browser with service workers and Pyodide

Simon Willison releases a complete rewrite of Datasette Lite using service workers and Pyodide, making the browser-based Datasette almost indistinguishable from the real server-side application, with new features like selecting SQLite versions and installing plugins.

Simon Willison's WeblogIn-site articleRunning the full server-side Datasette Python app in the browser with service workers and Pyodide

What MiniMax M3.1 Flash reveals about choosing and combining models

MiniMax M3.1 Flash excels at coding at low cost but underperforms in banking, clinical, and tax tasks. Benchmarks and simulations show that combining models by workload or escalation can improve completion and reduce costs, provided failures are reliably verified.

Fireworks AI BlogIn-site articleWhat MiniMax M3.1 Flash reveals about choosing and combining models
Robotics

Refinement vs. Replanning: Inference Compute Allocation in Robot Policies

This paper studies how a frozen chunked robot policy can allocate deployment-time inference compute between refining action chunks within each policy call and replanning more frequently from fresh observations. Using a SmolVLA checkpoint in LIBERO simulation, the authors find that with h=2, d=2 outperforms d=5 (139/160 vs 122/160) while using less arithmetic, and that h=5 vs h=2 shows similar success but 60.7% lower arithmetic per attempt. The work emphasizes joint evaluation of generation effort and feedback frequency without defining a universal setting.

arXiv RoboticsIn-site articleRefinement vs. Replanning: Inference Compute Allocation in Robot Policies

Robot Demonstration Videos Should Include Transparency Disclosures

A new paper argues that the robotics community should adopt disclosure standards for demonstration videos. A pre-registered six-country experiment (N=9,341) found that the public struggles to tell real footage from CGI and autonomy from teleoperation, leading them to overestimate robot capabilities; clear disclosures significantly improved discernment and understanding.

arXiv RoboticsIn-site articleRobot Demonstration Videos Should Include Transparency Disclosures

TMA: Sim-to-Real Forceful Humanoid Manipulation via Tactile-Driven Motor Adaptation

TMA introduces a tactile-driven motor adaptation approach for zero-shot sim-to-real forceful humanoid manipulation, benchmarked on round-doorknob opening. Trained entirely in simulation, it transfers to a Unitree G1 humanoid and substantially outperforms RMA and reactive tactile policy baselines.

arXiv RoboticsIn-site articleTMA: Sim-to-Real Forceful Humanoid Manipulation via Tactile-Driven Motor Adaptation

Optimal Operating-Point Selection for Single-Linearization Trajectory Planning of Nonlinear Robot Manipulators

The paper proposes an offline performance-driven method that treats the location of a single linearization as an optimization variable, selecting the best operating point via grid search and bounded refinement and reducing the nonlinear trajectory objective by 7% to 38% across four UR5 point-to-point motions.

arXiv RoboticsIn-site articleOptimal Operating-Point Selection for Single-Linearization Trajectory Planning of Nonlinear Robot Manipulators

Passage-Focused Soft Coordination for Hierarchical Multi-Robot Navigation in Bottleneck Environments

The paper proposes a passage-focused framework that combines soft global coordination with flexible local execution for multi-robot navigation in bottleneck environments. Across 1,200 simulated instances in 4 environments with 5-30 robots, it achieves a near-perfect success rate and outperforms baselines in makespan and flowtime; hardware trials with 8 robots completed without collisions.

arXiv RoboticsIn-site articlePassage-Focused Soft Coordination for Hierarchical Multi-Robot Navigation in Bottleneck Environments

Technical Report of Auriga's 1:10 Scale Autonomous Vehicle Robot

A technical report from Shahid Beheshti University's Robotics and Intelligent Automation Laboratory details Auriga Robotics' 1:10 scale autonomous vehicle, covering its custom chassis, ESP32-based distributed embedded system, Jetson Orin NX perception and control module, ROS software stack, and integration of Stanley, PID, and Pure Pursuit control algorithms.

arXiv RoboticsIn-site articleTechnical Report of Auriga's 1:10 Scale Autonomous Vehicle Robot

AR-WAM: efficient streaming world-action modeling for autonomous driving

AR-WAM is a streaming world-action model that uses a causal video backbone to incrementally update visual context and obtain future features in a single forward pass, avoiding repeated history encoding and iterative video generation. It achieves 91.5 PDMS after supervised training and 94.0 after reinforcement learning on NAVSIM-v1 with a single front camera.

arXiv Computer VisionIn-site articleAR-WAM: efficient streaming world-action modeling for autonomous driving
Policy

Urban Representation Learning for Identifying Potential Disparities in Crime Reporting

This study proposes a self-supervised urban representation learning framework that uses over 1.8 million georeferenced crime records from Mexico City to construct hourly, weekly, and static reporting profiles, learn scale-specific representations, and integrate them into a unified latent space. Neighbor Profile Divergence (NPD) then quantifies disagreement between each region's reporting profile and those of its retrieved peers to identify unusual reporting patterns. Experiments show a mean Silhouette score of 0.4052 versus 0.2863 for Hex2Vec, with NPD rankings highly stable across neighborhood sizes (Spearman correlations 0.9720–0.9951). High-divergence regions exhibit spatially organized patterns and serve as an internal auditing and prioritization mechanism rather than direct evidence…

arXiv Machine LearningIn-site articleUrban Representation Learning for Identifying Potential Disparities in Crime Reporting

Honest Brokers in AI Regulation: How Public Comments Introduce Evidence to Regulators

A study analyzing 2,664 public comments on 40 potential AI regulations in the USA finds that business commenters supply less evidence overall, while government, university, and civil society commenters provide more research and legal sources, complicating the narrative that public comments are merely a means of information capture.

arXiv AIIn-site articleHonest Brokers in AI Regulation: How Public Comments Introduce Evidence to Regulators

Auditors Struggle to Catch Sabotage In End-to-End Transcripts of AI Safety Research

A new red-team/blue-team study finds that human expert auditors, despite having full research transcripts and the ability to run experiments, barely distinguish sabotaged from benign AI safety research (AUROC 0.59, only 17% full identification). Frontier models like Fable 5.1 outperform humans even without experiment access (AUROC 0.73, 32% full identification). Identification drops further when the saboteur strategically chooses which experiments to attack. The results suggest current auditing methods are ill-equipped to detect research sabotage by expert-level AI agents.

arXiv AIIn-site articleAuditors Struggle to Catch Sabotage In End-to-End Transcripts of AI Safety Research

The Silent Consensus

In a Harvard lecture, OpenAI's Dean Ball introduces the 'silent consensus'—the implicit belief that humans will inevitably cede primacy over world affairs to superintelligent machines—and argues that this vision must be openly rejected in favor of a human-led future.

OpenAI NewsIn-site articleThe Silent Consensus

Your complete guide to AWS Quick at AWS re:Invent 2026

AWS Quick is an AI workspace built only for work, with enterprise security and governance from the start. At AWS re:Invent 2026, customers including Allianz, Cox Automotive, Mondelez, DXC, and Southwest Airlines share how they govern, extend, and scale Quick. Use this guide to find the sessions that fit your goals.

AWS Machine Learning BlogIn-site articleYour complete guide to AWS Quick at AWS re:Invent 2026

Advancing Catalog Interoperability with Unity Catalog and Google Cloud’s Borderless Lakehouse

Databricks and Google Cloud announce a milestone in catalog interoperability: customers can access a single copy of data from both Unity Catalog and Google Cloud's borderless Lakehouse without duplication, enabling bi-directional federation and open metadata governance through Iceberg catalog labels.

Databricks BlogIn-site articleAdvancing Catalog Interoperability with Unity Catalog and Google Cloud’s Borderless Lakehouse
Chips

Advancing chip design with OpenAI models

OpenAI's models achieved PPA improvements in Jalapeño inference chip and AMD GPU/CPU designs, and RISC-V core evaluations show progress in AI-assisted chip design.

OpenAI NewsIn-site articleAdvancing chip design with OpenAI models

Cerebras Launches Disaggregated Inference

Cerebras has launched disaggregated inference, pairing GPUs for prefill with wafer-scale engines for decode, delivering 5x more tokens per megawatt while preserving ultrafast generation speeds.

Cerebras BlogIn-site articleCerebras Launches Disaggregated Inference