Large language model (LLM) agents tackle complex tasks through reasoning and interaction with external environments. Agent skills provide persistent, transferable capabilities that guide task planning, inform contextual decisions, support tool use, and enable executable procedures. Yet existing skill-evolution methods...
Ye-Wei Zhao, De-Rong Xu, Wei Chen et al.· 0 citations
A memory can be outdated for present-day reasoning yet indispensable for reconstructing the past. Relevance-based retrieval overlooks this tension: semantic relevance, evidence-grounded validity, and query-specific eligibility are distinct. We propose CALM, a conflict-aware lifecycle memory governance framework that de...
This work introduces SAKI (Supervision Allocation with KL-constrained Interpolation), which combines a KL-constrained teacher-guided rollout with maximal coupling and reuses realized accept/correction events to route token-level supervision.
Miteto Wei, Xiao-Han Wang, Ze-Hao Chen et al.· 0 citations
Temperature-Grouped Reinforcement Learning (TGRL), which turns temperature-induced diversity into an explicit training signal, and reaches equivalent accuracy up to 36% faster than strong RLVR baselines without expanding the rollout budget.
Zi-Han Lin, Xiao-Han Wang, Jie Cao et al.· 0 citations
Experiments in controlled multi-solution environments and continual post-training settings across multiple model scales demonstrate that CoKL achieves a more favorable balance between target-task improvement and prior-capability retention than existing regularization methods.
Li Wang, Xiao-Dong Lu, Xiao-Han Wang et al.· 0 citations
OSOL, which designates a focus domain at each iteration, uses the preceding checkpoint footprint to rank token-level rebound risk, and applies a drift-ranked, adaptively scaled correction within the standard GRPO update, is proposed, which suppresses the targeted cross-step output backtracking component.
Zi-Han Lin, Xiao-Han Wang, Jie Cao et al.· 1 citation
HiDiffTIR is proposed, a Hierarchical Difficulty-aware policy optimization framework for multi-turn TIR that consistently improves multi-turn TIR performance and tool invocation accuracy over strong RL baselines, highlighting the necessity of difficulty-aware credit assignment for effective policy optimization in tool-...
Yu-Can Guo, Xiao-Han Wang, Miao Su et al.· 0 citations
BASM is proposed, which augments each skill with explicit boundary fields, which transforms each retrieved skill from an unconditional action template into state-conditioned guidance: the agent applies the skill when its conditions hold, suppresses inapplicable tool calls when they do not, and issues targeted repairs w...
Zi-Han Lin, Zhen-Yu Chen, Jiawen Wei et al.· 1 citation
Inspired by the human brain, which balances plasticity and stability through complementary episodic storage and gradual consolidation, UniMem is proposed, a self-routing framework for autonomous memory management that consistently outperforms baselines while maintaining execution fidelity.
