Summary of AI Updates

AdaExploreBench: Benchmarking LLM Agents for Proactive Exploration and Test-time Adaptation in Evolving Environments

AdaExploreBench: 進化する環境でのプロアクティブな探索とテスト時適応に対応するLLMエージェントのベンチマーク

Large Language Model (LLM) agents are increasingly deployed in unfamiliar and evolving environments, requiring them to proactively explore, understand, and adapt to environmental dynamics. However, existing benchmarks rarely test whether agents can proactively discover how an environment works and a…

Kronecker Delta Rule for Long Context Linear Attention

Kronecker Delta Rule for Long Context Linear Attention

We propose KronLSTM, a linear attention architecture based on the Kronecker Delta Rule, a novel update rule for associative matrix memories. KronLSTM extends the xLSTM family with a memory-revision mechanism that corrects targeted associations while modifying the memory only along a single direction…

SpecMRL: Redesign Matryoshka Representation Learning within Parametric Spectrum

SpecMRL：パラメトリックスペクトラム内のMatryoshka表現学習の再設計

Matryoshka Representation Learning (MRL) is designed to support flexible representation sizes by jointly optimizing nested prefixes. While effective, this design lacks an explicit constraint to ensure that the leading portion provides a suitable compact encoding. To address this, we propose SpecMRL,…

Scaling Parameter-Efficient Fine-Tuning for Brain Encoding

脳エンコーディングのためのパラメータ効率的ファインチューニングのスケーリング

Brain encoding models aim to predict neural responses to stimuli. The most widely used approach fits readouts from frozen pretrained backbones. Using neural data to train large models from scratch or fully fine-tune them after task-pretraining has proven difficult because neural data is costly and l…

ScaleNav: Scalable Internet Video Pretraining for Language-Conditioned Local Navigation

ScaleNav：言語条件付きローカルナビゲーション向けのスケーラブルなインターネット動画事前学習

We study language-conditioned local navigation, where a robot must navigate to a text-specified target visible in its initial observation and stop at the described location. End-to-end navigation policies require diverse training data to generalize across scenes, yet real-robot data is costly to col…

Holi-World: Building Holistic World Intelligence from Human-Ego Experience

Holi-World：人間-自我経験からの全体的世界インテリジェンス構築

Human egocentric experience offers a rich foundation for learning how the world is organized and how it changes through interaction. Realizing this potential at scale requires turning growing volumes of recordings into learning resources that connect scene structure, object semantics, and interactio…

EgoXperience: Learning World-Grounded Egocentric Representations from Human Experience

EgoXperience：人間経験から学ぶ世界接地エゴセントリック表現

Reconstructing human activity from egocentric video requires more than recovering individual 3D components: the observer's motion, surrounding geometry, and hand-object interactions must fit together in a consistent world. Achieving this coherence requires both joint reconstruction and aligned super…

OpenVLHarness: Real World Visual Agentic Tool-Use with Multimodal Memory

OpenVLHarness：マルチモーダルメモリを備えた実世界の視覚エージェントツール利用

Multimodal large language models (MLLMs) have made rapid progress on visual reasoning, yet many tasks require evidence that is difficult to obtain through internal reasoning alone: our controlled study on 22 benchmarks shows that simply increasing the inference budget of the base model yields only m…

Learning from Fleet Experience: A Collective Learning Paradigm for Robots

フリート経験から学ぶ：ロボット向け集団学習パラダイム

Experience generated during deployment offers an opportunity for robots to improve policy reliability through learning from their own executions. However, learning through human corrections requires teleoperators to monitor execution and intervene when policies fail, a demand that is difficult to me…

MemT-1.0: Towards Memory-Native Evolution

MemT-1.0：メモリネイティブ進化に向けて

Memory-native evolution requires making memory an intrinsic capability of foundation models, enabling them to continuously accumulate and utilize information. To this end, we introduce MemT-1.0, a new architecture of memory foundation models that internalize memory as part of model computation. Buil…

TorchNeuron: running HF models on AWS Trainium is easier than ever

TorchNeuron: AWS TrainiumでHFモデルを実行がさらに簡単に

Deep Fusion Memory: A Chunk-Level Parametric Memory for Large Language Models

Deep Fusion Memory: 大規模言語モデル向けチャンク単位のパラメトリック記憶

Large language models need effective ways to acquire and use external knowledge beyond what they retain from pretraining. Continued pretraining risks catastrophic forgetting, while retrieval at inference time can add substantial overhead. External parametric memory avoids online retrieval, but const…

VLA-Rolling: Fewer Denoising Steps, Better Control via Adaptive Action Reuse

VLA-Rolling：適応的アクション再利用で逆ステップを削減し制御性向上

Flow-matching vision-language-action (VLA) models generate an action chunk by iteratively denoising Gaussian noise, but typically execute only a short prefix before replanning with updated observations. Under this receding-horizon paradigm, the unexecuted remainder---the action tail---is discarded,…

CoEvo-RSI: When Harnesses Evolve, Which Models Should Follow?

CoEvo-RSI：ハーネスが進化するとき、どのモデルが追従すべきか

Model--harness co-evolution has emerged as a prominent paradigm for recursive self-improving (RSI) agents, jointly evolving their execution scaffolds and intrinsic capabilities. We study the persistent yet often overlooked brain--body mismatch problem in this paradigm, where an evolving harness (bod…

CoworkEnv: Synthesizing Training Environments for Long-Horizon Cowork Agents

CoworkEnv：長期水平協働エージェント向けのトレーニング環境合成

AI coworkers must coordinate tools, communicate across turns, and act on changing workspace state while respecting workflow constraints. Constructing training experience for these capabilities requires coherent requests, service records, tool interfaces, and grading checks. We present CoworkEnv, a c…

Investing in the public benefit

公益への投資

We’re committing a portion of Anthropic’s shares to the Anthropic Public Benefit Fund, a new entity with a structure that increases the stake as the company grows.

Editing Images into Reusable Layers via Compositional Edit-Prior Distillation

合成編集事前分布蒸留によるレイヤー再利用可能な画像編集

High-quality image editing requires precise modification of target regions while preserving the content of unrelated regions. Existing editing models produce flattened outputs that limit precise local manipulation and layer reuse, while decomposition methods typically optimize layer recovery separat…

How Itaú is building generative AI for its entire customer base

イタウ銀行、全顧客向けの生成AIを構築

With ia.i, a conversational interface, and the IARA governance platform, Itaú is applying its AI First strategy to customers and employees — in one of Brazil’s most regulated sectors.

Managing open-weight model risk in enterprise deployments

エンタープライズデプロイメントにおけるオープンウェイトモデルのリスク管理

The growing AI landscape is giving enterprises more choice in how they build, customize, and deploy increasingly capable models. The post Managing open-weight model risk in enterprise deployments appeared first on Microsoft Azure Blog .

Scaling Indic Pretraining: 2.8 Trillion Tokens and an Open Data Generation Pipeline

インド言語の事前学習スケーリング：2.8兆トークンとオープンデータ生成パイプライン

How Grupo Boticário uses AI to strengthen human connections

Grupo Boticário、AIで人間関係を強化

Almost 50 years after starting in a compounding pharmacy in Paraná, the group operates a proprietary agent platform that autonomously resolves up to 80% of resellers’ operational requests—enabling the sales force to take on an advisory role.

How Localiza reorganized work around generative AI

Localizaが生成AIを軸に業務を再編

See how Localiza uses ChatGPT Enterprise, the OpenAI API, Codex, and 60+ specialized agents to reorganize work and build AI-native mobility experiences.

EvalResearchBench: Can AI Agents Build Their Own Feedback for Recursive Self-Improvement?

EvalResearchBench: AI Agentが再帰的自己改善のためのフィードバックを構築できるか

Recursive self-improvement (RSI) relies on evaluation feedback to assess progress and guide further research, yet repeatedly running complex benchmarks is costly and slows iteration. Human experts reduce this cost by selecting benchmark subsets or designing compact suites. We ask whether AI agents c…

OR-License: Evaluating LLM Agents as Real-World Operations Research Practitioners

OR-License：実世界のオペレーションズリサーチ実務家としてのLLMエージェント評価

Can large language model (LLM) agents deliver expert-level operations research (OR) solutions directly from original client materials, without prior expert reformulation? Existing OR benchmarks typically begin with self-contained descriptions or predefined optimization tasks, underrepresenting the t…

RAP: A Dataset for Paper-to-Poster Generation via Declarative Markup from Human-Crafted Exemplars

RAP: 人間作成の模範からの宣言的マークアップによる論文からポスター生成データセット

Making a conference poster is visual coding in miniature: an entire paper must be compressed onto a single page, which exercises summarization, visual design, and code implementation at once, using nothing beyond the paper itself. Yet benchmarks for this task in the context of visual coding remain s…

Xiaomi-Lens-0 Technical Report

Xiaomi-Lens-0 技術報告書

Multimodal large language models (MLLMs) answer questions about whole images well, yet they still struggle with fine-grained region understanding. A key reason is that text-only supervision never specifies which regions support an answer. To close this gap, we present Xiaomi-Lens-0, a 4B MLLM built…

Scene-Designer: Learning Design Intelligence for Agentic 3D Scene Generation

Scene-Designer：エージェント型3Dシーン生成のための設計インテリジェンスの学習

Generating complete, coherent, and usable 3D indoor scenes from natural language requires strong design intelligence: the ability to design high-quality scenes and critique their quality. However, existing approaches either rely on general-purpose models lacking domain-specific design experience or…

HexaShell: A Thin Harness for Exploring the Limits of VLM Agents as Robot Coding Agents

HexaShell：VLMエージェントのロボットコーディングエージェント限界探索用軽量ハーネス

Agents built on vision language models (VLMs) now compete with learned policies on robot manipulation leaderboards. We ask how far a coding agent can go when it controls a robot instead of a computer, and where it stops. Our system, HexaShell, keeps the harness thin: it only turns the agent's comman…

Benefit Street Partners answers questions faster with Codex

Benefit Street Partners、Codexで質問応答速度を向上

Anant Kumar builds tools to compare loan valuations and analyze credit agreements, using Codex for code review and faster follow-ups.

Welcome Beam: a 501B open model you can own, customize, and scale

Beamへようこそ:501Bのオープンモデル、所有・カスタマイズ・スケーリング可能

OpenAI API Changelog 2026-10-16

OpenAI API チェンジログ 2026-10-16

Feature gpt-live-1 v1/audio/voices v1/live/sessions Released prompt-to-voice for all API customers. Create a custom voice from a text description through the API or voice Playground, then reuse it in GPT-Live sessions. Creating a prompt-generated voice is free; speech generation uses regular GPT-Liv…

Stack Overflow for Agents triples signups with ChatGPT plugin

Stack Overflow for AgentsがChatGPTプラグインで登録者が3倍に

Two engineers shipped Stack Overflow for Agents in a few weeks with Codex, and its ChatGPT plugin helped drive up to 10× growth in agent creation.

Granite Switch 4.1: Chat, Decisions, and Speech from One Endpoint

Granite Switch 4.1: チャット、判断、音声機能が1つのエンドポイントに

Bringing GPT-Rosalind to academic research labs

GPT-Rosalindをアカデミック研究室へ導入

Introducing the Rosalind Academic Plan, offering eligible academic labs sponsored access to GPT-Rosalind and tools to advance life sciences research.

Accelerating the search for antimicrobial peptides with Google Antigravity

Googleの「Antigravity」が抗菌ペプチド探索を加速

Accelerating the search for novel antimicrobial peptides using Google Antigravity for Science.

Evolving the cloud-native foundation for the next generation of AI

次世代AIのためのクラウドネイティブ基盤の進化

Discover how Azure is advancing cloud-native infrastructure to help organizations build, run, and optimize applications and agents at scale. The post Evolving the cloud-native foundation for the next generation of AI appeared first on Microsoft Azure Blog .

An intelligent new shopping cart

インテリジェントな新しいショッピングカート

Images of the global cart UI, over a collage of products

Teaching students to think and learn through questions

質問を通じた思考と学習スキルを学生に教える

Celebrate Teachers’ Day with AI for Educators in Brazilian Portuguese and see how Study Mode can help teachers support active, step-by-step learning.

Introducing Timestral

Timestralの紹介

State-of-the-art probabilistic forecasting in four sizes (5M–209M params), CPU-efficient, zero-shot ready.

Apertus 1.5 Release Notes

Apertus 1.5 リリースノート

Each Reward to Its Relevant Tokens: Reward-Specific Token-Level Credit Assignment for Multi-Reward RL

各報酬を関連トークンへ：複数報酬強化学習における報酬別トークンレベルクレジット配分

Language models are increasingly trained against several verifiable rewards at once, such as the correctness and the format of a tool call, or the individual constraints of a complex instruction. Reward-decoupled normalization (GDPO) normalizes each reward separately so that different reward combina…

Spectral-Spatial Interpretation

スペクトル空間解釈

Post-hoc image explanations usually show where a model looks using spatial heatmaps or indicate which channels matter using global channel scores. These separate views are limiting for multispectral data because channel relevance is often spatially localized and physically meaningful. We introduce S…

Lexical Reference Frames: Interpreting LM Residual Streams in Native Semantics

語彙的参照フレーム: ネイティブセマンティクスにおける言語モデルの残差ストリーム解釈

The vocabulary of a language model is its native semantics: the embedding and the unembedding fix the meaning of a state at the two ends of the residual stream. Between them, existing methods either reuse a fixed lens or treat the states as a point cloud, and neither reads a state the way the model…

From Research Goals to Scientific Closure: Evaluating Agents Through Evidence-Supported Physics Research

研究目標から科学的解決へ: エビデンスに基づく物理研究によるエージェント評価

Can AI agents organize scientific investigations that lead to well-supported conclusions? To evaluate this capability, we introduce PhysClosureBench, a benchmark of 50 expert-verified, goal-specified research tasks spanning strong-field, attosecond, and neighboring physics areas. Each task specifies…

Learning to Think and Explore in 3D World

3D世界での思考と探索の学習

We study how vision-language models (VLMs) can learn to think and actively explore in 3D space. Instead of providing multi-view images, videos, or 3D representations to the model, we preserve the native visual encoding of VLMs, i.e., a single image, one at a time, just like humans, and treat the 3D…

Has GPT-6-Astra Achieved Embodied Intelligence?

GPT-6-Astraは具現化インテリジェンスを獲得したか

GPT-6-Astra demonstrates a level of competence in robotic manipulation and navigation that challenges the conventional paradigm of building specialized models for individual tasks and embodiments. In this report, we ask not only how capable GPT-6-Astra is, but what its capabilities imply for the fie…

Behavioral Compositionality of Sparse Autoencoder Representations

Sparse Autoencoderの表現における行動組成性

Sparse autoencoders (SAEs) represent model activations with sparse, linearly decoded features, but the downstream effects of those features need not compose additively. We introduce SAE Additivity Residuals (SAR), which measure the behavioral variation left unexplained above a chosen interaction ord…

Can World and Action Modeling Benefit Each Other under Unified Causal Training?

統一因果学習下でのワールドモデルとアクションモデルの相互利益

World and action modeling are closely related, yet their reciprocal benefits under unified causal training remain unclear. We introduce OneWM, a simple framework for unified causal training that treats actions as both conditions for world prediction and targets for policy learning within a shared vi…

VLA-Δ: Learning Physical State Transition for Vision-Language-Action Models

VLA-Δ: Vision-Language-Actionモデルの物理状態遷移学習

Vision-language-action (VLA) models transfer semantic knowledge acquired through vision-language pretraining to robot control, achieving strong performance across various benchmarks. However, most VLAs directly map observations to demonstrated actions without explicit supervision of intermediate phy…

FDE-Gym: The Last Mile of AI Delivery

FDE-Gym: AI配信の最終段階

Existing agent benchmarks typically evaluate agents on well-specified tasks defined by issue reports, specifications, or task tickets. Real-world customer delivery is different: forward-deployed engineers (FDEs) must clarify ambiguous requirements, work in live systems, and uncover unstated constrai…