7 papers
S^3martCirc: Self-supervised Smart Circuit Discovery
Wendy Zheng, Yinhan He, Liang Wu +1
Large Language Models (LLMs) have demonstrated remarkable performance across diverse tasks, from text summarization to question answering. Despite these capabilities, their black-b…
Every Token Leaves a Ripple in the Stream of Thought: Eliciting Model-Internal Token Saliency for Chain-of-Thought Compression
Tianyi Zhao, Yinhan He, Wendy Zheng +1
Chain-of-thought (CoT) reasoning improves multi-step problem solving, but long reasoning traces inflate inference cost. Token-level CoT compression reduces this cost by pruning ful…
Addressing the Reasoning Gap: Mechanistic Circuit-Based Knowledge Editing in Large Language Models
Tianyi Zhao, Yinhan He, Wendy Zheng +1
Deploying Large Language Models (LLMs) in real-world dynamic environments raises the challenge of updating their pre-trained knowledge. While existing knowledge editing methods can…
Wired for Overconfidence: A Mechanistic Perspective on Inflated Verbalized Confidence in LLMs
Tianyi Zhao, Yinhan He, Wendy Zheng +2
Large language models are often not just wrong, but \emph{confidently wrong}: when they produce factually incorrect answers, they tend to verbalize overly high confidence rather th…
IAPO: Information-Aware Policy Optimization for Token-Efficient Reasoning
Yinhan He, Yaochen Zhu, Mingjia Shi +5
Large language models increasingly rely on long chains of thought to improve accuracy, yet such gains come with substantial inference-time costs. We revisit token-efficient post-tr…
SemCoT: Accelerating Chain-of-Thought Reasoning through Semantically-Aligned Implicit Tokens
Yinhan He, Wendy Zheng, Yaochen Zhu +6
The verbosity of Chain-of-Thought (CoT) reasoning hinders its mass deployment in efficiency-critical applications. Recently, implicit CoT approaches have emerged, which encode reas…