token generation

3 papers

• Corrector Sampling in Language Models
• Reasoning Path Compression: Compressing Generation Trajectories for Efficient LLM Reasoning
• STree: Speculative Tree Decoding for Hybrid State Space Models