The Physics of Multi-Turn Long-Horizon Planning: From Pre-training to Post-training via Single- and Multi-Teacher On-Policy Agentic Distillation
- 类型：arxiv
- 标识：2607.24720
- 链接：https://arxiv.org/abs/2607.24720
- 主分类：engineering
- 形态：method
- 被引：2
- 被引来源：Semantic Scholar
- S2被引：2
- OpenAlex被引：0
- 影响力被引：0
- TLDR：A unified and controlled multi-turn environment that enables precise control of long-horizon planning across three stages is introduced and it is shown that multi-teacher on-policy distillation (MOPD) integrates capabilities by converging to shared planning-pattern across environments.
- OpenAlex ID：W7171571615
- OpenAlex DOI：10.48550/arxiv.2607.24720
- DOI：10.48550/arxiv.2607.24720
- DOI来源：OpenAlex
- 开放获取：green
- 开放获取链接：https://doi.org/10.48550/arxiv.2607.24720
- OpenAlex更新：2026-08-25
- 副分类：agent
- 待LLM分类：否
- 标题中文：多轮长时程规划的物理机制：从预训练到后训练的单教师与多教师 On-Policy Agentic 蒸馏
- TLDR中文：提出一个统一且受控的多轮环境，可在三阶段精确控制长程规划；研究表明，多教师 on-policy 蒸馏 (MOPD) 通过收敛到跨环境共享的规划模式来整合多种能力。
- 来源文件：
- /inbox/tom/_candidates/2026-07-28-agent-rag-longcontext-candidates.json
- [S2 enrich]
- [OpenAlex backfill]