用更一致的轨迹、更少的解码步数「驯服」掩码扩散语言模型，扩散语言模型的推理性能和效率大幅提升
"解码策略" 相关文章
用更一致的轨迹、更少的解码步数「驯服」掩码扩散语言模型，扩散语言模型的推理性能和效率大幅提升
用更一致的轨迹、更少的解码步数「驯服」掩码扩散语言模型，扩散语言模型的推理性能和效率大幅提升
The End of Manual Decoding: Towards Truly End-to-End Language Models
Evaluating Evaluation Metrics -- The Mirage of Hallucination Detection
一致性轨迹强化学习登场：上海AI Lab让扩散语言模型实现全并行少步数推理
一致性轨迹强化学习登场：上海AI Lab让扩散语言模型实现全并行少步数推理
一致性轨迹强化学习登场：上海AI Lab让扩散语言模型实现全并行少步数推理
一致性轨迹强化学习登场：上海AI Lab让扩散语言模型实现全并行少步数推理
Self-Augmented Visual Contrastive Decoding
Energy-Conscious LLM Decoding: Impact of Text Generation Strategies on GPU Energy Consumption
Energy-Conscious LLM Decoding: Impact of Text Generation Strategies on GPU Energy Consumption
Finish First, Perfect Later: Test-Time Token-Level Cross-Validation for Diffusion Large Language Models
Finish First, Perfect Later: Test-Time Token-Level Cross-Validation for Diffusion Large Language Models
华人主导谷歌SLED，论文登顶会！一键让模型学会自救
Taming Masked Diffusion Language Models via Consistency Trajectory Reinforcement Learning with Fewer Decoding Step
Bridging Draft Policy Misalignment: Group Tree Optimization for Speculative Decoding
Mirror-Consistency: Harnessing Inconsistency in Majority Voting
CoVeR: Conformal Calibration for Versatile and Reliable Autoregressive Next-Token Prediction
A Comparative Study of Decoding Strategies in Medical Text Generation
