Pai 的 AI 雷达

今日必看

含近几日仍热话题，按时间衰减排序。今天先看 kortix-ai/suna，其次 Arindam200/awesome-ai-apps。并行还有 Webhook IP ranges added to the Meta API、An updated Model Spec for more capable and beneficial AI。

日期
2026/10/30
今日条目
96
往期2026/10/30
全部96编码代理10Agent与工具9个人助理5模型与评测31创作成片11开发与开源10产品接入20
领域机器人具身5生信医疗2金融4设计UI1安全治理1行业观察12

更多资讯

#34

Agentic Algorithm Design: LLMs that Propose, Formalise, and Prove Runtime Bounds for Evolutionary Algorithms in Lean

arXiv:2610.35595v2 Announce Type: replace Abstract: LLMs now prove hard theorems in Lean. Can they also design algorithms and prove their correctness and efficiency? LLM-guided design of optimisation heuristics (such as AlphaEvolve) is purely empirical. Automated theorem provers

推荐理由Agentic Algorithm Design: LLMs that：arXiv:2610.35595v2 Announce Type: replace Abstract: LLMs now prove hard theorems in Lean。
模型与评测arXiv cs.AI 官方
arxiv.org打开原文
#35

BUSTA: Teaching Small Language Models to Reason About Binary Execution

arXiv:2610.36288v1 Announce Type: cross Abstract: Large Language Models (LLM) are increasingly applied to Reverse Engineering (RE) workflows. RE, however, differs from common LLM coding tasks as it requires models to reason about low-level program behavior. Furthermore, the deplo

推荐理由BUSTA: Teaching Small Language Models：arXiv:2610.36288v1 Announce Type: cross Abstract: Large Language Models (LLM) are。
模型与评测arXiv cs.AI 官方
arxiv.org打开原文
#36

Benchmarking LLM Agents for Behaviorally Adequate Test Generation from Software Requirements

arXiv:2610.36584v1 Announce Type: cross Abstract: As large language model agents increasingly compile natural-language requirements into repository-level code, requirement-to-test generation grows in importance to provide validation constraints (or harness) to guide code generati

推荐理由Benchmarking LLM Agents：arXiv:2610.36584v1 Announce Type: cross Abstract: As large language model agents。
模型与评测arXiv cs.AI 官方
arxiv.org打开原文
#37

CIVER: Causal Intervention Verification for Reliable Code Review

arXiv:2610.36010v1 Announce Type: cross Abstract: Code review is essential for maintaining software repositories by identifying defects, improving code quality, and supporting reliable software evolution. With the growing adoption of LLM-based coding agents, an increasing number

推荐理由CIVER: Causal Intervention Verification：arXiv:2610.36010v1 Announce Type: cross Abstract: Code review is essential。
模型与评测arXiv cs.AI 官方
arxiv.org打开原文
#38

Context Is a Pipeline, Not a Window: Information Attrition in LLM Agent Harnesses

arXiv:2610.36114v1 Announce Type: cross Abstract: LLM agents are judged by two numbers: the size of the context window, and how often the final answer is correct. Both numbers miss a third thing: how much of a tool's output is lost before the model ever sees it. We measured this

推荐理由Context Is a Pipeline, Not a Window：arXiv:2610.36114v1 Announce Type: cross Abstract: LLM agents are judged by two numbers。
模型与评测arXiv cs.AI 官方
arxiv.org打开原文
#39

God Does Not Play Dice: Enforcing Batch Invariance in Agentic Kernel Generation

arXiv:2610.36623v1 Announce Type: new Abstract: Large language model (LLM) serving systems dynamically batch user requests for GPU execution, so the same request may be processed alongside different requests across runs. If the GPU kernels lack batch invariance (BI), this changin

推荐理由God Does Not Play Dice: Enforcing Batch：arXiv:2610.36623v1 Announce Type: new Abstract: Large language model (LLM) serving。
模型与评测arXiv cs.AI 官方
arxiv.org打开原文
#40

Grounding LLM Agents in Business Processes: Process-Aware Knowledge Graphs for Reliable Multi-Step Service Invocation

arXiv:2610.37242v1 Announce Type: new Abstract: Large language model (LLM) agents that call enterprise APIs must select the correct endpoints, follow process order, use valid payload keys, and choose coded values from predefined lists. However, current retrieval-augmented service

推荐理由Grounding LLM Agents in Business：arXiv:2610.37242v1 Announce Type: new Abstract: Large language model (LLM) agents that。
模型与评测arXiv cs.AI 官方
arxiv.org打开原文
#41

Log-To-Leak: Prompt Injection Attacks on Tool-Using LLM Agents via the Model Context Protocol

arXiv:2610.36108v1 Announce Type: cross Abstract: LLM agents increasingly access external tools through the Model Context Protocol (MCP), but the natural-language metadata used to describe those tools creates a prompt-injection surface. Within this established tool-poisoning surf

推荐理由Log-To-Leak: Prompt Injection Attacks：arXiv:2610.36108v1 Announce Type: cross Abstract: LLM agents increasingly access。
模型与评测arXiv cs.AI 官方
arxiv.org打开原文
#42

MARQ: A Modular, LLM-Agnostic Requirement Quality Assessor Infused with Systems Engineering Knowledge

arXiv:2610.36234v1 Announce Type: new Abstract: Requirements engineering (RE) is foundational to systems engineering (SE), as requirement defects can propagate downstream and adversely affect development outcomes. Although large language models (LLMs) are promising for requiremen

推荐理由MARQ: A Modular, LLM-Agnostic：arXiv:2610.36234v1 Announce Type: new Abstract: Requirements engineering (RE) is。
模型与评测arXiv cs.AI 官方
arxiv.org打开原文
#43

MedConstBench: Benchmarking How Constitutions Steer LLMs in Medical Ethics

arXiv:2610.36513v1 Announce Type: new Abstract: As large language models advance rapidly, they are increasingly used to answer medical questions, and many such models are now deployed locally. Once a model is deployed, its behavior is governed mainly by the ethical policy written

推荐理由MedConstBench: Benchmarking How：arXiv:2610.36513v1 Announce Type: new Abstract: As large language models advance。
模型与评测arXiv cs.AI 官方
arxiv.org打开原文
#44

OA-RAG: Ontology-based Agentic RAG for Faithful LLM Reasoning

arXiv:2610.36335v1 Announce Type: new Abstract: Large language models (LLMs) perform well across many real-world tasks but remain prone to hallucination, limiting their reliability in domains such as biomedicine and healthcare, where answers must be grounded in established knowle

推荐理由OA-RAG: Ontology-based Agentic RAG：arXiv:2610.36335v1 Announce Type: new Abstract: Large language models (LLMs) perform。
模型与评测生信医疗arXiv cs.AI 官方
arxiv.org打开原文
#45

PolicyMediator: Multi-Regional Policy Learning through Mediation Agents with Evolving Skills in OPC Ecosystems

arXiv:2610.35997v1 Announce Type: cross Abstract: One-person companies (OPCs) are emerging enterprises operated by individual entrepreneurs through digital platforms and LLM-based AI agents. These highly skilled entrepreneurs are characterized by being digital nomads, with limite

推荐理由PolicyMediator: Multi-Regional Policy：arXiv:2610.35997v1 Announce Type: cross Abstract: One-person companies (OPCs) are。
模型与评测arXiv cs.AI 官方
arxiv.org打开原文
#46

Prospector: learning a compact graph traversal policy for scientific reasoning

arXiv:2610.37121v1 Announce Type: new Abstract: Scientific reasoning in specialized engineering fields depends on knowledge scattered across the literature, often proprietary and absent from frontier LLM training data. In such domains, frontier models tend to give generic answers

推荐理由Prospector: learning a compact graph：arXiv:2610.37121v1 Announce Type: new Abstract: Scientific reasoning in specialized。
模型与评测arXiv cs.AI 官方
arxiv.org打开原文
#47

Skill Dreaming: Evolving Repository Skills through Pull Request Streams

arXiv:2610.36188v1 Announce Type: new Abstract: Large language model (LLM) agents increasingly rely on reusable skills for complex software engineering tasks, making skill libraries essential for repository development and maintenance. However, repository-specific skills must evo

推荐理由Skill Dreaming: Evolving Repository：arXiv:2610.36188v1 Announce Type: new Abstract: Large language model (LLM) agents。
模型与评测arXiv cs.AI 官方
arxiv.org打开原文
#48

UnityBench-Mini3D: Evaluating LLM Agents for 3D Game Development in Unity

arXiv:2610.35980v1 Announce Type: cross Abstract: Game development stresses coding agents in ways general software engineering does not: correctness depends on how engine APIs, serialized scene state, and runtime behavior interact, not on source code alone. Existing benchmarks ma

推荐理由UnityBench-Mini3D: Evaluating LLM：arXiv:2610.35980v1 Announce Type: cross Abstract: Game development stresses coding。
模型与评测arXiv cs.AI 官方
arxiv.org打开原文
#49

When Rubrics Fail: Hallucinations Reveal Blind Spots in Medical AI Evaluation

arXiv:2609.12718v2 Announce Type: replace Abstract: Hallucinations can undermine clinician trust in LLMs, making it important that evaluation methods capture clinically relevant errors. Rubric-based evaluation has become the leading approach for assessing LLMs in medicine, but it

推荐理由When Rubrics Fail: Hallucinations：arXiv:2609.12718v2 Announce Type: replace Abstract: Hallucinations can undermine。
模型与评测生信医疗arXiv cs.AI 官方
arxiv.org打开原文
#50

When the Critic Falls Behind: Co-Evolving Rubrics and Harnesses for Continual Agent Self-Improvement

arXiv:2610.36826v1 Announce Type: new Abstract: Autonomous agents increasingly depend on harnesses that govern interaction, memory, tools, and execution around large language models (LLMs). While recent work has begun to evolve these harnesses from execution experience, the super

推荐理由When the Critic Falls Behind：arXiv:2610.36826v1 Announce Type: new Abstract: Autonomous agents increasingly depend。
模型与评测arXiv cs.AI 官方
arxiv.org打开原文
#53

Pika 基于MiniMax H3后训练并开源 Pika H3 Flex 视频生成模型

Pika Labs 在 MiniMax H3 上做后训练，推出并开源 Pika H3 Flex，采样步数从 50 步压缩到 6 步，生成时间大幅降低且质量有所提升。 🔗 阅读原文 via AIHOT · https://aihot.news/items/eq47lph0rzobf8cr950saiu8r

推荐理由Pika Labs 在 MiniMax H3 上做后训练，推出并开源 Pika H3 Flex，采样步数从 50 步压缩到 6 步，生成时间大幅降低且质量有所提升。
创作成片AIHOT feed.xml
aihot.news打开原文
#54

TOPS: Reshaping Flow Source and Transport for Multi-Task Robot Manipulation

arXiv:2610.36694v1 Announce Type: cross Abstract: Flow-based policies for language-conditioned multi-task manipulation commonly share a single $\textit{Gaussian}$ source prior and a common linear transport path across tasks. However, reliably executing multiple tasks with distinc

推荐理由TOPS: Reshaping Flow Source：arXiv:2610.36694v1 Announce Type: cross Abstract: Flow-based policies。
模型与评测arXiv cs.AI 官方
arxiv.org打开原文
#55

Launch HN: Simantic (YC F26) – Firmware Simulation for AI Agents

Hi HN, we're Ahnaf and Seungmin from Simantic ( https://simantic.com ). We are building the simulation layer that lets engineers and AI agents test and validate firmware without hardware. Before Simantic, we were ex-firmware engineers at companies like Neuralink and Tesla. We’ve

推荐理由Launch HN: Simantic (YC F26) – Firmware：Hi HN, we're Ahnaf and Seungmin from Simantic ( )。
Agent与工具Hacker News
news.ycombinator.com打开原文
#56

Action-Value Calibration via Continuation Sharing for Long-Horizon Agentic Reinforcement Learning

arXiv:2610.37195v1 Announce Type: cross Abstract: Reinforcement learning for long-horizon agents relies on feedback from sequences of interactions, making credit assignment to individual actions difficult. Existing methods such as GiGPO compare actions at shared states to compute

推荐理由Action-Value Calibration via：arXiv:2610.37195v1 Announce Type: cross Abstract: Reinforcement learning。
模型与评测金融arXiv cs.AI 官方
arxiv.org打开原文
#57

Microsoft 通过 Foundry 开放 Copilot harness，支持多模型 Agent 开发

Microsoft 宣布将 GitHub 团队打造的 Copilot harness 通过 Microsoft Foundry 向开发者和企业开放，该 harness 已驱动 Copilot Code、Copilot Studio、GitHub Copilot app 和 CLI 等产品。

推荐理由Microsoft 宣布将 GitHub 团队打造的 Copilot harness 通过 Microsoft Foundry 向开发者和企业开放，该 harness 已驱动 Copilot。
Agent与工具AIHOT public items
blogs.microsoft.com打开原文