Rollout-Time Intervention is introduced, a technique designed to elicit successful trajectories beyond the policy's intrinsic exploration capacity, enabling the post-trained policy to approach the performance of substantially stronger contemporary LLMs and, more importantly, transcend its original capability boundary.
Xi Chen, Rui-Xue Ding, Zhigaoyuan Wang et al.· 0 citations
This work proposes ArenaFlow, a hierarchical credit propagation framework for open-ended agent reinforcement learning that leverages tournament-based relative ranking to derive trajectory-level reward signals and propagates trajectory-level advantages to high-confidence pivotal steps according to tournament survival de...
Qiang Zhang, Rui-Xue Ding, Fan-Rui Zhang et al.· 1 citation
ARISE-RL, a novel full-cycle self-evolution framework that couples a task/rubric Generator and a reasoning Solver through rubric-mediated co-evolution, is proposed and ECR-Bench, an expert-calibrated rubric benchmark suite covering single-tool deep research and multi-tool travel planning is presented.
Fan-Rui Zhang, Rui-Xue Ding, Qiang Zhang et al.· 0 citations
Experimental results show that although current agents can reliably uncover the problems exposed by alerts, they struggle to proactively investigate the disk for silent intrusions and to produce comprehensive, verified remediation plans, with no model achieving complete detection and remediation on any single range.
Le-Han Wang, Boli Chen, Ruixue Ding et al.· arXiv.org· 0 citations
We use cookies to run the site and, with your consent, for analytics and to show ads.
See our Cookie Policy.