Summary of AI Updates

CLASHarness: Benchmarking Multi-Agent Conflict in Computer Use via a Controllable and Verifiable Generation Pipeline

CLASHarness：制御可能で検証可能な生成パイプラインによるコンピュータ使用時のマルチエージェント競合ベンチマーク

Vision-language models (VLMs) are increasingly deployed in multi-agent systems to automate complex computer-use workflows, yet existing benchmarks mostly assume consistent, cooperative environments. In contrast, real-world collaboration rarely runs on consistent information: peer or sub-agents frequ…

Many Right Answers, One Training Example: Removing Responder Noise from GRPO for LLM Recommenders

複数の正解、単一の訓練例：LLM推奨システムのGRPOから応答者ノイズを除去

A recommendation request on Reddit draws several replies, each naming different titles. When an LLM recommender is trained on such threads with reinforcement learning (RL), which replies should a generated list be rewarded against? The state-of-the-art recipe, Rank-GRPO, picks one reply per training…

FlashVSR+: Efficient Video Super-Resolution via Hybrid Attention and Reward-Yield Distribution Matching

FlashVSR+: ハイブリッド注意とReward-Yield分布マッチングを活用した効率的動画超解像

Recent video super-resolution (VSR) models achieve efficient one-step generation through diffusion distillation. However, high-resolution inference remains costly, and distilled models are difficult to align with human preferences. These challenges call for improvements in both model architecture an…

Tev: Scaling Long-Horizon Agentic Environments with Dense Progress Rewards

Tev: 密度の高い進捗報酬で長期ホライズンエージェント環境をスケーリング

Long-horizon agents have drawn growing attention for their ability to work autonomously -- from hours to days -- on economically valuable tasks, a capability essential to automated discovery and recursive self-improvement (RSI). Yet open-source models still struggle on long-horizon tasks, largely fo…

GreenWAM: A World Foundation Model for Robotics

GreenWAM: ロボティクス向けワールドファウンデーションモデル

A robot needs a world model that predicts how its actions will change the scene and a policy that chooses those actions. Models are usually built and evaluated for only one of these roles, and every robot records its actions in its own format. We ask whether one robot world model can serve both role…

AesCode: Aesthetic Code Generation with Decoupled Cross-Modal Rewards

AesCode：分離されたクロスモーダル報酬による美的コード生成

Generating information-rich visual artifacts such as slides, posters, and dashboards as code is appealing because the outputs are structured, editable, and verifiable. Yet code models cannot see how their choices of layout and color come together on the canvas, often limiting spatial planning and ae…

VideoJev: System One Model for Video Understanding and Interaction

VideoJev：動画理解とインタラクション向けSystem Oneモデル

Visual agents must turn video observations into timely decisions to track progress and choose what to do next. Building on the emerging System One paradigm, we introduce VideoJev, a family of models that cast video understanding and interaction as structured decision tasks. We adapt pretrained multi…

Jeva: System One Models from First Principles

Jeva：第一原理から導くSystem Oneモデル

System one models enable direct, low-latency decision-making. Yet recent practice of predicting an answer from serialized text does not, by itself, respect the structure of a decision problem. We argue that current designs underemphasize four important structural requirements. (i) null-aware: Models…

Urbex: Agentic Spatial Grounding for City-Scale 3D Scenes

Urbex：都市規模の3Dシーンに対するエージェント型空間グラウンディング

City-scale 3D grounding is commonly formulated as candidate ranking, which requires precomputed instance proposals and becomes inefficient in large outdoor scenes. We propose Urbex, an agentic vision-language framework that instead casts grounding as active spatial search. Urbex represents a city-sc…

Agent-Reflex: Teaching Language Agents to Act Reflectively

Agent-Reflex：言語エージェントに反射的な行動を学習させる

Large language models have evolved from single-turn chatbots into interactive agents capable of long-horizon reasoning. In real deployments, MCP tool services can return noisy feedback and users may intervene to correct the agent, but existing methods provide limited supervision for reflection learn…

Gemini API リリースノート 2026-10-13

Gemini 4 Argon ( gemini-4-argon ) generally available (GA) : Released Gemini 4 Argon ( gemini-4-argon ), our most capable model built for complex coding, agentic, math, and science tasks. For API changes, pricing, and migration guidance, see What's new in Gemini 4 Argon .

Deep dive into Azure Maia 300 architecture

Azure Maia 300アーキテクチャの詳細解説

Maia 300 operates as a single platform to meet the demands of cloud-scale agentic inference. The post Deep dive into Azure Maia 300 architecture appeared first on Microsoft Azure Blog .

Self-Distillation with Verbal Regularization

言語正則化による自己蒸留

On-policy self-distillation (OPSD) trains a large language model using supervision from a teacher instantiated from the same model as the student. The teacher is conditioned on privileged information (PI), such as verified solutions. The student, which does not observe PI, learns from the teacher's…

Multics: Generative Multi-Player World Model with Hierarchical Communication

Multics：階層的コミュニケーション付き生成型マルチプレイヤーワールドモデル

Multi-player interactive world models aim to simulate shared environments from multiple players' perspectives, but achieving robust generation while scaling to more players remains challenging. Existing approaches that rely on fully-connected communication (e.g., Solaris) incur quadratic attention c…

X-Ragent: Tool-Augmented Reinforcement Learning for Evidence-Grounded Chest X-Ray Reasoning

X-Ragent：胸部X線推論のためのツール拡張強化学習

Chest X-ray (CXR) interpretation often requires integrating visual findings with task-specific evidence rather than relying on direct visual recognition alone. While tool-augmented medical agents can supply such evidence via external modules, they often lack explicit mechanisms for learning when aux…

Recursive Self-Improvement through Multi-Agent Self-Supervision

マルチエージェント自己教師あり学習による再帰的自己改善

Apertus 1.5: From Open Text Models to Open Multimodal Reasoners

Apertus 1.5: オープンテキストモデルからオープンマルチモーダル推論モデルへ

Language models are becoming infrastructure, and infrastructure built in the open, by public institutions, on documented data, is the alternative to dependence on a few providers; Apertus shows this is viable at scale. Apertus 1.5 extends the fully open Apertus family from text to multimodal reasoni…

BridgeWAM: Bridging Video and Action Experts via Latent Bridge Queries

BridgeWAM：潜在ブリッジクエリでビデオと行動専門家を統合

World Action Models (WAMs) commonly couple video and action experts through dense, layer-wise Mixture-of-Transformers (MoT) attention, exposing action prediction to irrelevant visual features while enforcing matched expert depths. This coupling hinders selective visual-feature use and introduces red…

Learning Where to Split: Dynamic Band Splitting for Audio Source Separation

学習分割箇所：音声ソース分離のための動的バンド分割

Audio source separation models must preserve fine spectral structure while keeping the cost of the separator tractable. Band-split architectures reduce this cost by manually partitioning the spectrum into predefined frequency bands and encoding each band as a token, but the same boundaries are used…

KV4-RL: Unlocking Faster Asynchronous Rollouts for RLVR with INT4 KV Cache Quantization

KV4-RL: INT4 KVキャッシュ量子化によるRL非同期ロールアウトの高速化

Asynchronous rollout improves the efficiency of reinforcement learning with verifiable rewards (RLVR) by reducing under-utilization during long-tail generation. Sustained concurrency also concentrates context processing into fewer decode steps, increasing the KV workload per step. We characterize th…

Last Secure Code Benchmark: Is the Code You Vibe Secure?

Last Secure Code Benchmark: あなたの「Vibe Coded」コードは安全か

AI coding agents now write much of the code that ships, yet whether that code is secure remains unclear. We introduce the Last Secure Code Benchmark, a comprehensive evaluation of secure code generation by coding agents. Built from 150 real vulnerabilities, its 450 tasks span seven widely used langu…

ExeWorldBench: Benchmarking Code-Based Executable 3D World Modeling

ExeWorldBench：コードベース実行可能3Dワールドモデリングのベンチマーク

Coding agents are increasingly evolving beyond reproducing what a world looks like toward constructing executable worlds that capture how a world works. This shift calls for executable world modeling, which integrates two core capabilities: world understanding, reasoning about the underlying structu…

EviScene: Beyond "Is It AI?" for Agentic Image Factuality Assessment

EviScene：エージェント型画像の事実性評価を「それはAIか？」を超えて

Determining whether an image is AI-generated does not establish whether the situation it depicts is factually supported. We formulate Image Factuality Assessment: given an image and a generic verification instruction, a system returns a supported or refuted verdict with an evidence-grounded explanat…

Impactful scheduling for GPU clusters

GPUクラスタの効率的スケジューリング

S-Space: Exploring Spatial Workspace in Multimodal Language Models

S-Space: マルチモーダル言語モデルにおける空間ワークスペースの探索

Multimodal language models (MLLMs) exhibit promising spatial reasoning capabilities, yet the internal mechanisms supporting them remain poorly understood. We uncover S-Space, a linear representation subspace in which object-token representations encode continuous coordinates along horizontal, vertic…

Asana cuts model costs 76x in browser tests with GPT-6.1 Sol

Asana、GPT-6.1 Solでブラウザテストのモデルコストを76倍削減

Using GPT-6 Astra in Codex, Asana made its browser agent 76x cheaper and 5x faster in tests to offer customers more capable models.

Sophos cuts threat investigation time by 96% with OpenAI Daybreak

SophosがOpenAI Daybreakで脅威調査時間を96%削減

Discover how Sophos uses OpenAI’s Daybreak to cut cyber-threat investigation time by 96% and automate 52% of MDR cases while preserving human oversight.

OpenAI API Changelog 2026-10-09

OpenAI API Changelog 2026-10-09

Feature v1/agents You can now pre-warm an OpenAI-hosted sandbox before attaching it to an Agents API session. Prewarming improves latency by creating and configuring the sandbox before a user finishes sending their message. This means the user doesn't have to wait for all this work to complete. Open…

OpenMobile-2: Building Versatile Mobile Agents with Scalable Environments and App-Native Tools

OpenMobile-2：スケーラブルな環境とアプリネイティブツールで多機能モバイルエージェントを構築

We introduce OpenMobile-2, a near-frontier mobile agent with fully open training environments and recipes. We make three key advances: (1) Diverse environments with simulated commercial apps: We build MobileGym++, featuring 35 realistic, functionally rich commercial-style apps with cross-app workflo…

The Thinking Leakage in Reinforcement Learning for LLMs

LLMの強化学習における思考リークの分析

Hybrid reasoning models such as the Qwen3 series switch between a THINK mode and a NO-THINK mode through a few chat-template tokens. A popular recipe trains such models with reinforcement learning in NO-THINK mode and reports improved reasoning performance. We provide evidence that much of this gain…

Beyond Domain-Routed Supervision: Cross-Teacher Guarding for Multi-Teacher On-Policy Distillation

Beyond Domain-Routed Supervision: マルチティーチャーOn-Policy Distillationのための交差ティーチャーガーディング

Multi-teacher on-policy distillation (MOPD) extends on-policy distillation (OPD) by routing domain-specific data to specialized teachers, integrating their capabilities through dense token-level supervision on student-generated trajectories. However, domain routing only distinguishes the supervision…

APEX: Asynchronous Prefix Denoising for Extreme Compression

APEX: 極限圧縮向け非同期プレフィックス除ノイズ

Fewer latent tokens promise cheaper diffusion. Yet aggressive latent compression creates a reconstruction-generation dilemma: adding channels improves reconstruction but makes generation harder. We introduce APEX, short for Asynchronous Prefix denoising for EXtreme compression, which pairs a hierarc…

Multi-Stream Attention Residuals

マルチストリーム注意残差

Attention Residuals (AttnRes) enables content-dependent retrieval over network depth, which can be viewed as single-stream softmax attention along the depth axis. Inspired by the success of multi-head attention along the sequence axis, we introduce Multi-Stream Attention Residuals (MSAR), an extensi…

AgentSWE: Can Coding Agents Build the Agent You Actually Want?

AgentSWE: コーディングエージェントは本当に欲しいエージェントを構築できるか

As the production of software passes into the hands of software itself, coding agents are increasingly commissioned to build agents fitted to their users' own workflows. This delegation carries a quiet assumption, that the delivered agent will faithfully serve the requirement it was built for, rathe…

Scaling Agentic Data for Long-Horizon Terminal Intelligence

長期タスク向けエージェント型データスケーリング

Terminal agents are moving beyond local operations toward completing real-world professional tasks. Short trajectories can demonstrate commands and tool calls, but they do not fully show how agents accumulate evidence through exploration, use it to inform judgments, continually adjust strategies bas…

Continually Post-training Small Models via Synthetic Self-distillation

合成自己蒸留による小規模モデルの継続事前学習

Continual post-training faces a dual challenge of acquisition and retention: models must learn new capabilities while preserving prior knowledge. While Self-Distillation Fine-Tuning (SDFT) mitigates this using self-supervision with in-context demonstrations, it relies on strong in-context learning a…

Work2Skill: Capturing Human GUI Experience for Reusable Agent Skill Induction

Work2Skill：人間のGUI操作経験から再利用可能なエージェントスキルを獲得

Computer-use agents can perform increasingly complex GUI tasks, yet unfamiliar software and workflows requiring human domain knowledge remain challenging. Human demonstrations convey procedural knowledge through the same GUI interaction space, but reuse requires interpreting actions in context and s…

ProgramGym: Scaling Training Environments for Long-Horizon Program Reproduction Agents

ProgramGym：長期ホライズンプログラム再現エージェント向けトレーニング環境のスケーリング

Coding agents are increasingly used to build programs from scratch as software engineers do rather than to resolve issues within a repository or to generate a repository from a given requirements document. However, scalable verifiable training environments that force agents to discover what is requi…

EmbodiedCLBench: Evaluating Continual Learning for Self-Evolving Embodied Agents

EmbodiedCLBench：自己進化型実体エージェント向け継続学習の評価

Continual learning requires models to accumulate past experience and transfer it to new tasks. Despite substantial research on continual learning, existing work remains largely confined to non-embodied settings, without extending to embodied environments. To address this gap, we introduce EmbodiedCL…

医師向け検索ツール「Evidence Finder」にSakana Namazuを導入：日本仕様のAIを、医療の現場へ

How Oracle turns days of work into minutes with ChatGPT and Codex

OracleはChatGPTとCodexで数日の業務を数分に短縮

Across recruiting, engineering, and operations, Oracle turns specialist knowledge into fast, repeatable workflows with ChatGPT Work and Codex.

LegalOn halves Codex costs while maintaining development speed

LegalOnがCodexコストを半減、開発速度を維持

LegalOn cut estimated daily Codex costs by 65% while maintaining development speed. It matched Astra, Sol, and Luna to tasks and managed budgets strategically.

Pollo AI turns creative ideas into campaigns with OpenAI

Pollo AIがOpenAIでクリエイティブなアイデアをキャンペーンに変換

With GPT-5.6, GPT-6 Astra, and GPT‑Image‑2.5, Pollo AI helps creators turn bold ideas into detailed images and cinematic video ads.

Right Screen, Wrong Transition: World Models as Verifiers for GUI Agents

Right Screen, Wrong Transition：GUI エージェント検証者としてのWorld Models

A login screen that appears after a tap on Sign in is expected; the same screen after a tap on View order is an attack. For GUI agents, safety is therefore a property of the transition rather than of the screen, and a monitor that inspects only screens can be defeated by reusing a legitimate one. Ju…

Super-Resolution in The Right Latent Space: A Frozen Vision-Foundation Substrate

適切な潜在空間での超解像：凍結ビジョン基盤基質

In an image latent space, the embeddings of high-resolution, natural, and sharp images form a manifold. Degradation of high-resolution images pushes their embeddings off this manifold. Real-world super-resolution (SR) then becomes the task of mapping the degraded embedding back onto this manifold --…

Thinking Inertia: LLMs Keep Thinking When Told Not To

思考慣性: LLMは指示されても思考を続ける

Large Language Models (LLMs) increasingly ship with explicit "thinking modes", yet their counterpart, "no-thinking", has received far less attention. We study LLMs' no-thinking behavior along two axes. a. How to measure no-thinking? Prior work typically defines no-thinking through proxies such as a…

Contextual Correction of Parallel Draftsfor Speculative Decoding

投機的デコーディング向け並列ドラフトの文脈的修正

Speculative decoding accelerates large language model inference by using a lightweight drafter to propose multiple tokens that are subsequently verified in parallel by the target model. Recent diffusion-based drafters make this paradigm particularly attractive: deep attention-based neural backbones…

SepGen: Multi-Stem Audio-Video Separation and Generation in a Single Model

SepGen：単一モデルでマルチステムオーディオ・ビデオ分離と生成

A 4D audio-visual scene comprises a video, the dynamic geometry it depicts, and the sound sources that populate it, each with its own position and trajectory. Rendering such a scene from a novel viewpoint requires that every source be available as an individual waveform, so that it can be localized…

Can AI Agents Learn Their Way to the Top? Evaluating Heuristic Learning in a Long-Running Game Agent Competition

AIエージェントは競争を通じてスキルアップできるか：長期ゲームエージェント競技会でのヒューリスティック学習を評価

Adversarial games have driven advances from heuristic search to reinforcement learning, yet learning and adapting strategies from limited samples remain challenging. AI agents offer an alternative by turning game experience into revisions of executable policies. Building on heuristic learning (HL),…

SoftwareGym: Scaling Computer-Use Agent Training with Diverse Synthesized Software Environments

SoftwareGym：多様な合成ソフトウェア環境でのコンピュータ利用エージェント訓練をスケーリング

Computer-use agents (CUAs) are rapidly improving, yet synthesizing diverse and reliable software environments at scale remains challenging. Native application environments require substantial construction and integration effort, which limits application coverage and increases the risk of application…