#LLM Agent

2026

• 09-03 EN

  HIERA: Teaching LLM Agents Where to Optimize, Not Just How — Workload-Aware Planning Across GPU Implementation Spaces

• 09-03 中

  HIERA 阅读笔记：让 LLM Agent 先学会「在哪一层优化」，再谈「怎么优化」——GPU 算子实现空间的分层规划

• 08-31 EN

  ContextPilot: Teaching LLM Agents to Manage Their Own Context with Fine-Grained RL

• 08-31 中

  ContextPilot 阅读笔记：让 Agent 用细粒度强化学习学会自己管理上下文

• 08-17 EN

  AgentRewind: Giving Long-Horizon LLM Agents an Undo Button That Actually Undoes Things

• 08-17 中

  AgentRewind 阅读笔记：给长程 LLM Agent 装一个真正能撤销的悔棋键

• 08-10 EN

  GRAFT: Global Optimization and Inference-Time Region Grafting for Agentic Workflows

• 08-10 中

  阅读笔记：GRAFT——给智能体工作流做「全局定骨架、局部动手术」的推理时嫁接

• 07-28 EN

  Branching Policy Optimization: Turning Sandbox Snapshots Into a Variance-Reduction Primitive

• 07-28 中

  Branching Policy Optimization：把沙盒快照变成一个方差削减原语

• 07-27 EN

  The Regression Tax: Why Adding Skills to an LLM Agent Also Breaks Tasks It Used to Solve

• 07-27 中

  Regression Tax（回归税）：给 LLM Agent 加 Skill 为什么既能帮忙又能拆台

• 07-25 EN

  When Does Adding More Agents Actually Help? An Information-Bottleneck Answer

• 07-25 中

  多智能体到底什么时候才有用？一个信息瓶颈视角的答案

• 07-20 EN

  SEED: When Hindsight Becomes a Training Signal, Not Just a Better Prompt

• 07-20 中

  SEED 阅读笔记：当「事后诸葛亮」变成训练信号，而不只是更好的提示词

• 07-13 EN

  Long-Horizon-Terminal-Bench: Dense Reward Grading Exposes Agent Long-Horizon Execution Limits

• 07-13 中

  Long-Horizon-Terminal-Bench 阅读笔记：密集奖励评估揭露 Agent 长程执行瓶颈

• 07-07 EN

  RSPO: Reward-Swap Policy Optimization for Multi-Turn LLM Agents

• 07-07 中

  RSPO：多轮LLM智能体的奖励交换策略优化

• 06-22 EN

  MRAgent: Why Memory Should Be Reconstructed, Not Retrieved

• 06-22 中

  MRAgent：记忆应该被重建，而不是被检索

• 06-15 EN

  Parallel-Synthesis: Direct KV-Cache Synthesis for Parallel Branches in LLM-Agent Workflows

• 06-15 中

  Parallel-Synthesis：让 LLM 综合智能体直接消费并行分支的 KV 缓存

• 06-08 EN

  ExpWeaver: How LLM Agents Learn from Past Experience in Latent Space

• 06-08 中

  ExpWeaver：LLM 智能体如何在隐空间中从经验中学习

• 06-01 EN

  AutoSci: A Memory-Centric Agentic System for the Full Scientific Research Lifecycle

• 06-01 中

  AutoSci：以记忆为中心的全科研生命周期自主智能体系统

• 05-25 EN

  CodeAct: Executable Code Actions Elicit Better LLM Agents

• 05-25 中

  CodeAct：用可执行代码驱动更强的 LLM Agent

• 05-18 EN

  Why Single-Agent LLMs Beat Multi-Agent Systems on Multi-Hop Reasoning — A Budget-Controlled Story

• 05-18 中

  思考预算锁死之后，单 Agent 为什么打过多 Agent —— 阅读笔记

• 05-11 EN

  MASPO: Joint Prompt Optimization for LLM-based Multi-Agent Systems

• 05-11 中

  MASPO：面向 LLM 多智能体系统的联合提示词优化

• 05-01 EN

  Low-Rank Optimization Trajectories for LLM RLVR Acceleration: A Technical Review of NExt

• 04-27 EN

  Agentic World Modeling: Foundations, Capabilities, Laws, and Beyond — Technical Review

• 04-22 EN

  SAGE: Training-Free Semantic Evidence Composition for Edge-Cloud Inference Under Hard Uplink Budgets

• 04-13 EN

  Toolformer: Language Models Can Teach Themselves to Use Tools — Deep Technical Review

• 04-13 中

  Toolformer：让语言模型自己学会“什么时候调用工具”——深度阅读笔记

• 04-12 EN

  Voyager: An Open-Ended Embodied Agent with Large Language Models — Deep Technical Review

• 04-12 中

  Voyager：一个能在 Minecraft 中持续成长的 LLM 具身智能体 —— 深度阅读笔记

• 04-11 EN

  Language Agent Tree Search (LATS): Unifying Reasoning, Acting, and Planning in Language Models — Deep Technical Review

• 04-11 中

  LATS（Language Agent Tree Search）：把推理、行动、规划统一到同一个语言模型代理框架里 — 深度阅读笔记

• 03-30 EN

  Chain-of-Thought Prompting Elicits Reasoning in LLMs — In-Depth Technical Review

• 03-23 EN

  MiRA: A Subgoal-driven Framework for Improving Long-Horizon LLM Agents — Technical Review

• 03-16 EN

  MetaGPT: When LLM Agents Form a Software Company — Multi-Agent Collaboration Done Right

• 03-09 EN

  AutoGen: Microsoft's Framework for Building Multi-Agent Conversations That Actually Work

• 03-09 EN

  Generative Agents: 25 AI Characters Living in a Simulated Town — Believable Human Behavior from LLMs

• 03-06 EN

  SWE-agent: Turning LLMs Into Autonomous Software Engineers That Fix Real GitHub Issues

• 02-23 EN

  Self-Refine: Teaching LLMs to Critique and Improve Their Own Output — No Extra Training Needed

• 02-20 EN

  Reflexion: LLM Agents That Learn from Failure Through Verbal Self-Reflection

• 02-18 EN

  GLM-5 Technical Review: From Vibe Coding to Agentic Engineering

• 02-16 EN

  Tree of Thoughts: Deliberate Problem Solving with Large Language Models — Technical Review

• 02-09 EN

  ReAct Technical Review: From Reasoning Ability to Executable Reasoning