5 papers
Shallow Queries, Mature Values: Depth-Asynchronous Self-Speculation for Looped Transformers
Guanghao Li, Zihan Su, Hao Yu +6
Looped Transformers reuse a shared block across recurrent depths, making autoregressive decoding expensive because every generated token requires many sequential recurrent passes.…
DiffuSpec: Unlocking Diffusion Language Models for Speculative Decoding
Guanghao Li, Zhihui Fu, Min Fang +4
As large language models (LLMs) scale up, accuracy improves, but the autoregressive (AR) nature of decoding increases latency since each token requires a serial forward pass. Specu…
SCOUT: Teaching Pre-trained Language Models to Enhance Reasoning via Flow Chain-of-Thought
Guanghao Li, Wenhao Jiang, Mingfeng Chen +6
Chain of Thought (CoT) prompting improves the reasoning performance of large language models (LLMs) by encouraging step by step thinking. However, CoT-based methods depend on inter…
MIGA: Mutual Information-Guided Attack on Denoising Models for Semantic Manipulation
Guanghao Li, Mingzhi Chen, Hao Yu +4
Deep learning-based denoising models have been widely employed in vision tasks, functioning as filters to eliminate noise while retaining crucial semantic information. Additionally…
Zero Token-Driven Deep Thinking in LLMs: Unlocking the Full Potential of Existing Parameters via Cyclic Refinement
Guanghao Li, Wenhao Jiang, Li Shen +2
Resource limitations often constrain the parameter counts of Large Language Models (LLMs), hindering their performance. While existing methods employ parameter sharing to reuse the…