State2State: Environment-Derived Mid-Training for LLM Agents

arXiv:2608.04934 · cs.CL, cs.LG · Submitted 2026-08-05 · Read on arXiv

Xuanyu Lei, Yiqi Zhu, Chenliang Li, Kaiming Liu, Peng Li, Ming Yan, Jieping Ye, Ya-Qin Zhang, Yang Liu

cs.CL, cs.LG

Submitted: 2026-08-05

Comments: Work in progress

Code: https://github.com/THUNLP-MT/State2State

License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/

Terminology

• GRPO — 22× in this paper · explained in 6 episodes
• Verifier — 16× in this paper · explained in 2 episodes
• Reinforcement Learning — 10× in this paper · explained in 4 episodes
• Training Objective — 8× in this paper · explained in 3 episodes
• Benchmark — 7× in this paper · explained in 2 episodes
• Generalization — 7× in this paper · explained in 14 episodes
• Training Objectives — 7× in this paper · explained in 2 episodes
• Distillation — 6× in this paper · explained in 2 episodes
• Agent Early Experience — 5× in this paper
• Fine-Tuning — 5× in this paper · explained in 5 episodes
• Adam — 4× in this paper · explained in 2 episodes
• Verifiable Rewards — 4× in this paper · explained in 2 episodes
• RLVR — 3× in this paper · explained in 2 episodes
• Actor — 2× in this paper · explained in 3 episodes
• Adaptation — 2× in this paper
• Core — 2× in this paper · explained in 2 episodes
• Downstream Tasks — 2× in this paper · explained in 2 episodes
• Online Reinforcement Learning — 2× in this paper · explained in 2 episodes
• Action Space — 1× in this paper · explained in 1 episode
• Benchmarking — 1× in this paper · explained in 4 episodes
• Credit assignment — 1× in this paper · explained in 3 episodes
• Domain Adaptation — 1× in this paper · explained in 5 episodes
• Group Relative Policy Optimization — 1× in this paper
• Group Relative Policy Optimization (GRPO) — 1× in this paper · explained in 10 episodes
• Hyperparameters — 1× in this paper · explained in 2 episodes
• Institute for AI Industry Research — 1× in this paper
• Language Models (LLMs) — 1× in this paper · explained in 2 episodes
• Large Language Models (LLMs) — 1× in this paper · explained in 9 episodes
• Provenance — 1× in this paper · explained in 5 episodes
• Reinforcement Learning with Verifiable Rewards (RLVR) — 1× in this paper · explained in 3 episodes
• Resource Constraints — 1× in this paper · explained in 5 episodes
• Verification — 1× in this paper · explained in 2 episodes
• World Model — 1× in this paper · explained in 8 episodes

Sources

• FireAct: Toward Language Agent Fine-tuning
• Why Do LLM Agents Fail in Exploring New Environments? A World-Modeling Perspective
• Training LLM-Based Agents with Synthetic Self-Reflected Trajectories and Partial Masking
• AgentHER: Hindsight Experience Replay for LLM Agent Trajectory Relabeling
• Group-in-Group Policy Optimization for LLM Agent Training
• Search-R1: Training LLMs to Reason and Leverage Search Engines with Reinforcement Learning
• MobileWorld: Benchmarking Autonomous Mobile Agents in Agent-User Interactive and MCP-Augmented Environments
• Tongyi DeepResearch Technical Report
• GUI-ReWalk: Massive Data Generation for GUI Agent via Stochastic Exploration and Intent-Aware Reasoning
• Scaling Agents via Continual Pre-training
• Kimi K2: Open Agentic Intelligence
• Agentic Critical Training
• Qwen3 Technical Report
• A Survey on LLM Mid-Training
• OpenCUA: Open Foundations for Computer-Use Agents
• DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models
• SynthAgent: Adapting Web Agents with Synthetic Supervision
• HybridFlow: A Flexible and Efficient RLHF Framework
• TaskCraft: Automated Generation of Agentic Tasks
• RAGEN: Understanding Self-Evolution in LLM Agents via Multi-Turn Reinforcement Learning

Related papers

• Exploring Solution Divergence and Its Effect on Large Language Model Problem Solving
• Ishigaki-IDS-Bench: A Benchmark for Generating Information Delivery Specification from BIM Information Requirements
• Subliminal Steering: Stronger Encoding of Hidden Signals
• MedStruct-S: A Benchmark for Key Discovery, Key-Conditioned QA and Semi-Structured Extraction from OCR Clinical Reports
• The End of Transformers? On Challenging Attention and the Rise of Sub-Quadratic Architectures
• Untangling the Mechanisms of Misleading Context in Medical Question Answering