Text-space optimization adapts large language models (LLMs) by editing external natural-language artifacts rather than model weights, so the optimized artifacts remain inspectable and the model can be treated as a black box. However, most existing text-space methods keep evaluation fixed. On open-ended tasks, this can become a bottleneck: once the solver improves on the criteria a rubric measures, omitted dimensions remain invisible to the optimization signal. Simply evolving the rubric is also unreliable when updates are selected by the current solver's score, because apparent progress can come from making the rubric easier to satisfy. We introduce DecoEvo (Decoupled Co-Evolution), which co-evolves a solver skill and a rubric-generator skill under decoupled objectives without using gold rubrics during optimization. The solver skill is updated using criterion-level feedback, while the rubric-generator skill is revised through complementary audits of requirement coverage and response discrimination that are independent of aggregate solver score. This separation focuses generator updates on newly exposed solver weaknesses, reducing repeated emphasis on criteria the solver already satisfies. Under each benchmark's official evaluation, DecoEvo outperforms all compared methods across five benchmarks and three LLM backbones, yielding 2.8--5.0\% relative gains over SkillOpt in the five-benchmark average.
Jiangwang Chen, Zixin Song, Junlin Liu et al.· 2 citations
Multi-Agent Protocol Distillation (MAPD), a joint distillation and RL framework uses a structured, style-normalized protocol as an intermediate representation that generalizes robustly across diverse proprietary teachers while effectively mitigating the student policy from style drift and verbosity degeneration.
Junlin Liu, Jiangwang Chen, Zixin Song et al.· arXiv.org· 6 citations
PReM (Preserve and Refresh Memory), a context-compression framework that maintains the long context as the model's internal layer-wise KV memory and learns what to preserve and when to refresh it, is introduced.
Bohan Yu, Lei Shen, Chenxi Zhou et al.· arXiv.org· 0 citations
ClawTrack is presented, a dual-assessment benchmark that simultaneously measures what an agent achieves (Task Score) and how it achieves it (Process Score) and finds that process scores effectively attribute success and failure to specific reasoning dimensions, filtering lucky passes invisible to outcome-only evaluation.
Xingjian Wu, Xuhan Zhu, Xing-Chen Liu et al.· arXiv.org· 0 citations
Persistent Consistency Self-Distillation (PCSD) is proposed, which derives token-level distillation weights from the local persistence of teacher-favoring signals, and combines adaptive windows with exponentially decayed aggregation to capture persistent relative teacher support.
Chunji Lv, Yangguang Wei, Junlin Liu et al.· 0 citations
Contrastive Reinforced Policy Optimization (CRPO) is introduced, which reformulates agentic OPSD from a contrastive learning perspective, and conducts group-wise contrast to preserve reliable, fine-grained optimization signals.
Xingjian Wu, Junlin Liu, Xingchen Liu et al.· arXiv.org· 1 citation
We use cookies to run the site and, with your consent, for analytics and to show ads.
See our Cookie Policy.
