Long-horizon agentic RL is constrained by the futures visited by the current policy. When rollouts fail to reach useful behaviors, improved credit assignment cannot provide supervision for states and actions that were never encountered. On-policy distillation (OPD) introduces teacher guidance on policy-induced states,...
Ding-Yi Rong, Yue Shi, Chaofan Ma et al.· 0 citations
This work establishes a feedback design strategy that reformulates environments by transitioning from action guidance to observation enrichment during the later stages of both intra-episode exploration and inter-episode evolution, and reveals that FEEs consistently yield performance improvements over standard settings.
Results support a qualified internalized-search reading: under the recipe the authors test, much of the measured RL gain corresponds to a change in sampling efficiency toward operating points the base model can already reach under search.
Wen-He Sun, Cun-Xiang Wang, Zi-Jun Yao et al.· 0 citations
ICAE-Bench, a benchmark for evaluating coding agents under interactive project-building settings, starts from a fuzzy product requirement, simulating the dynamic paradigm with an automated User Agent, and introduces three key designs.
Zhongyuan Peng, Dan Huang, Chuyu Zhang et al.· arXiv.org· 3 citations· ⚡1
We use cookies to run the site and, with your consent, for analytics and to show ads.
See our Cookie Policy.
