arXiv cs.CL· Mengdi Liu, Chenghao Jia, Hong Chang, Mingjie He, Shiguang Shan, Xilin Chen·· 6 天前AI 评分27
JANUS：用协同进化世界模型增强长时程自主研究
JANUS: Enhancing Long-Horizon Autonomous Research with Co-Evolving World Models
AI 导读
JANUS 是一个通过协同进化世界模型驱动的 Predict-Verify-Evolve 循环，统一未来结果预测与过往经验反思的自主研究框架。研究智能体在进化图中提出候选方案，世界模型按价值提升、信息增益和探索效用打分，据此在执行前选择、优化或剪枝分支。在 MLS-Bench-Lite 和真实研究任务上，JANUS 的进化效率与上限均显著优于 SOTA 基线。
正文 · 原文
Abstract:Long-horizon autonomous research has shown promise in accelerating scientific progress through sustained cycles of hypothesis generation, implementation, and validation. However, this Generate-Execute-Feedback paradigm relies on reactive trial-and-error, where costly and time-consuming executions introduce severe feedback delays, throttling evolutionary progress under explosively expanding search space. To address this challenge, we propose JANUS, a framework that unifies future outcomes forecasting and past experience reflection through a Predict-Verify-Evolve loop driven by a co-evolving world model. Within an evolving graph, the research agent proposes candidates while world model scores them by forecasting value improvement, information gain, and exploration utility, then accordingly selects, refines, or prunes branches prior to execution. After receiving ground-truth feedback, JANUS distills experience into evolving skills: hierarchical skills for research agent, exploration-strategy skills for world model, and their coordination protocols. This mechanism enables continual co-evolution: the research agent sharpens candidate generation while the world model refines its predictive judgment over extended horizons. Extensive evaluations on MLS-Bench-Lite and real research tasks demonstrate that JANUS significantly outperforms state-of-the-art baselines, achieving both higher evolving efficiency and ceiling.
| Subjects: | Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Multiagent Systems (cs.MA) |
| Cite as: | arXiv:2610.13194 [cs.AI] |
| (or arXiv:2610.13194v1 [cs.AI] for this version) | |
| https://doi.org/10.48550/arXiv.2610.13194 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Chenghao Jia [view email]
[v1]
Fri, 9 Oct 2026 08:13:28 UTC (5,389 KB)
来源：arXiv cs.CL · arxiv.org