PPL
Li Wang
Papers (7)
ZipRL: Adaptive Multi-Turn Context Compression with Hindsight Response ReplayLearning to Adapt SFT Data for Better Reasoning GeneralizationWhen Self-Belief Misleads: Active Label Acquisition for Reinforcement Learning with Verifiable RewardsImplicit Hierarchical GRPO: Decoupling Tool Invocation from Execution for Tool-Integrated Mathematical ReasoningTSFLora: Token-Compressed Split Fine-Tuning for Wireless Edge NetworksMASPO: Joint Prompt Optimization for LLM-based Multi-Agent SystemsLINC: Decoupling Local Consequence Scoring from Hidden Matching in Constructive Neural Routing