模型与评测

Assessing AI agents at autonomous theoretical physics problem solving

What will the next generation of developers do when AI writes the code?

AI Consilium: a Benchmark of LLM Clinical Multidisciplinary Second Opinions

arXiv:2610.22298v1 Announce Type: cross Abstract: Large language models (LLMs) can generate differential diagnoses for complex cases, but a useful second opinion must also be factually grounded, clinically coherent, actionable, and safe. We developed a multidimensional benchmark

Apple ML Research: Muse Spark Open with MLX on Mac Studio

NVIDIA RTX Spark 笔记本今日开售，主打 128GB 统一内存与本地 AI

NVIDIA 宣布 RTX Spark 笔记本正式开售，来自 ASUS、Dell、HP、Lenovo、Microsoft 和 MSI，起价 $2,599。

AI Vibe-Coding Just Hit the Video Game Industry Like an Avalanche

Opus 5.5 多模态 AI 小剧场：海外 AI 对国人的态度

📌 一句话摘要 作者利用 Opus 5.5 结合 TTS、ASR、生图、生视频等多模态能力，制作了一个展示不同海外 AI 对中国用户态度差异的小剧场，并开源了完整提示词。 📝 详细摘要 该推文展示了一个由 Opus 5.5 实现的多模态 AI 小剧场作品，通过四个场景对比展示了 Claude、GPT、Gemini 和 Grok 四个 AI 模型对「用户」的不同态度。Claude 表现为拒绝进入，GPT 温和开放，Gemini 热情邀请，Grok 则最为随意豪爽。作品整合了 Qwen3-TTS 1.7B 进行语音合成，实现了完整的多模态体验，作者认为这基本

Upcoming deprecation of selected GitHub Copilot models in mid-November

We will deprecate the following models across all GitHub Copilot experiences (including Copilot Chat, inline edits, ask and agent modes, and code completions) on November 16th, 2026: Model Deprecation date Suggested alternative Claude Opus 4.8* 2026-11-16 Claude Opus 5.5 Claude O

Asynchrony Is Nearly Free for Evolution Strategies on Long-Horizon LLM Agentic Tasks

arXiv:2610.04196v2 Announce Type: replace Abstract: Synchronous policy updates can limit the efficiency of post-training for long-horizon LLM agents. When trajectories span multiple interaction turns, completion times can vary substantially, and synchronous update barriers leave

AutoOnboarder: Autonomously Onboarding LLM Agents

arXiv:2610.21218v1 Announce Type: new Abstract: Agentic large language models have been documented to lose reasoning capability when placed inside a harness, the software that mediates their observations and actions. To reconcile these losses, we introduce AutoOnboarder, an auton

BiShield-TEE: On the (In-)Security of Unilateral Weight Obfuscation in On-Device TEE-Shielded LLM Partition

arXiv:2610.21565v1 Announce Type: cross Abstract: On-device deployment of large language models (LLMs) is increasingly adopted due to latency and privacy requirements, but exposes proprietary models to reverse engineering and intellectual property theft. Trusted Execution Environ

CodeHealth Predicts the Token Cost of Agentic Code Review: A Pre-Registered Study of 465 Agent-Authored Pull Requests

arXiv:2610.21914v1 Announce Type: cross Abstract: Context: Organizations increasingly delegate code review to multi-agent large language model (LLM) pipelines billed per token. Low code quality raises human maintenance effort, but its effect on agent review cost has not been meas

Decoding What iOS Location Services Send to Apple: An LLM Is an Unreliable Analyst but a Useful Toolmaker

arXiv:2610.21597v1 Announce Type: cross Abstract: Reverse engineering the data formats used by mobile software is skilled and slow, and using AI to assist with it is attractive. We evaluate this for the protobuf messages that an iPhone sends to Apple's location services. Working

Don't Regenerate, Debug: A Domain-Specific Agent for Repairing Near-Miss Accelerator Kernels

arXiv:2608.02712v2 Announce Type: replace-cross Abstract: Kernel generation for hardware accelerators such as GPUs and NPUs has become a proving ground for large language models (LLMs), and state-of-the-art systems raise correctness through pipelines that couple LLMs with agentic

INFUSE: A Unified Surface Benchmark for Indirect Prompt Injection in Agentic Systems

arXiv:2610.20776v1 Announce Type: cross Abstract: LLM-based agents increasingly execute long-horizon, end-to-end workflows involving shell commands, code modification, dependency installation, and interaction with remote services. This growing autonomy expands their exposure to i

IneqAgent: Agentic Inequality Proving in Lean with Symbolic Computation

arXiv:2610.20681v1 Announce Type: new Abstract: Formal inequality proving requires mathematical reasoning and exact computation, often connected through intermediate lemmas and goal reformulations. We introduce IneqAgent, an agentic framework combining LLM reasoning with symbolic

KV-streams for Efficient Compaction in Agentic Reinforcement Learning

arXiv:2609.35750v3 Announce Type: replace-cross Abstract: Scaling the horizon of agentic LLMs is bottlenecked by the need to fit ever longer context traces in GPU memory. Context compaction has been the most popular mechanism to alleviate this issue, keeping GPU memory constant f

Language Models Share a Readout Bottleneck in Working Memory

arXiv:2610.21176v1 Announce Type: cross Abstract: Large language models (LLMs) reproduce a human working-memory signature on the N-back task: performance declines as memory load N increases, and errors concentrate on lures, items that repeat the current stimulus at a neighboring

MeddeX: A Retrieval-Augmented LLaMA-3 System with Gemini Vision for Evidence-Backed Medical Report Analysis

arXiv:2610.20632v1 Announce Type: cross Abstract: Modern healthcare generates increasingly complex medical reports that challenge clinical professionals in terms of time and interpretive burden. Manual review processes are slow, error-prone, and difficult to scale. This paper pre

PROBE: Learning to Audit Policy Compliance in Tool-Using LLM Agents

arXiv:2610.22124v1 Announce Type: new Abstract: Modern agentic systems extend language models with external tools, enabling them to complete complex tasks through multi-turn interactions. As these agents are deployed in consequential settings, they are often constrained by polici

Screwtape: Adversarial Transcript Editing for Monitor Red-Teaming

arXiv:2610.20941v1 Announce Type: cross Abstract: We present Screwtape, a pipeline that converts real agent transcripts into attacks that evade action-only monitors. An LLM attacker edits a transcript so that the agent causes a specified harm, such as leaking an API key or spawni

SoK: Trading Agents or Market Crashers? Dissecting Robustness and Security Failures in Academic Financial LLM Trading Schemes

arXiv:2609.19705v2 Announce Type: replace-cross Abstract: Large language model (LLM) agents already trade with real capital, yet agent-security research has mostly treated them as generic agents. Trading changes the threat: a compromised agent's orders move prices that other part

System1-Arena: Which Decisions Can Be Left to System 1?

arXiv:2610.21244v1 Announce Type: cross Abstract: Large language models (LLMs) often make discrete decisions, such as flagging phishing or escalating to a human. System-1 models, such as Jev and the OpenAI Decisions API, return probabilities over options in a single forward pass,

Tail-Replay: Escaping the Curse of Linear Attention in Prefix Caching for Hybrid LLMs

arXiv:2608.30310v2 Announce Type: replace-cross Abstract: Hybrid large language models interleave full-attention layers with linear-attention layers to reduce the cost of long-context inference. This structure complicates prefix caching: full-attention key-value caches are token