7 papers
FlexDraft: Flexible Speculative Decoding via Attention Tuning and Bonus-Guided Calibration
Yaojie Zhang, Jianuo Huang, Junlong Ke +5
Speculative decoding accelerates memory-bound LLM inference without quality degradation by using a fast drafter to propose multiple candidate tokens and the target model to verify…
CKT-WAM: Parameter-Efficient Context Knowledge Transfer Between World Action Models
Yuhua Jiang, Yijun Guo, Hongbing Yang +7
World action models (WAMs) provide a powerful generative framework for embodied control, yet transferring knowledge across heterogeneous WAMs remains challenging due to mismatched…
Auto-FlexSwitch: Efficient Dynamic Model Merging via Learnable Task Vector Compression
Junqi Gao, Dazhi Zhang, Zhichang Guo +3
Model merging has attracted attention as an effective path toward multi-task adaptation by integrating knowledge from multiple task-specific models. Among existing approaches, dyna…
MARS: Scaling Multi-Agent Tree Search via Reinforcement Learning for Code Generation
Pengfei Li, Shijie Wang, Fangyuan Li +7
Reinforcement learning (RL) paradigms have demonstrated strong performance on reasoning-intensive tasks such as code generation. However, limited trajectory diversity often leads t…
DARE: Diffusion Large Language Models Alignment and Reinforcement Executor
Jingyi Yang, Yuxian Jiang, Xuhao Hu +3
Diffusion large language models (dLLMs) are emerging as a compelling alternative to dominant autoregressive models, replacing strictly sequential token generation with iterative de…
WIST: Web-Grounded Iterative Self-Play Tree for Domain-Targeted Reasoning Improvement
Fangyuan Li, Pengfei Li, Shijie Wang +4
Recent progress in reinforcement learning with verifiable rewards (RLVR) offers a practical path to self-improvement of language models, but existing methods face a key trade-off:…