activity
20242026
most citedNegative Preference Optimization: From Catastrophic Collapse to Effective Unlearning

3 citations · 5 across the 12 of their papers we have counts for

collaborators
Showing cs.CLShow all

7 papers · 1 filter

cs.CL2026

Robust Reasoning via Dynamic Token Selection for Distribution-Aligned Self-Distillation

Ruiqi Zhang, Lingxiang Wang, Hainan Zhang Zhiming Zheng

Self-distillation improves learning efficiency by rewriting reference answers as training data that better matches the model's own distribution. However, reference answers also int…

cs.CL2026

From Unfamiliar to Familiar: Detecting Pre-training Data via Gradient Deviations in Large Language Models

Ruiqi Zhang, Lingxiang Wang, Hainan Zhang +2

Pre-training data detection for LLMs is essential for addressing copyright concerns and mitigating benchmark contamination. Existing methods mainly focus on the likelihood-based st…

cs.CL2025

Long or short CoT? Investigating Instance-level Switch of Large Reasoning Models

Ruiqi Zhang, Changyi Xiao, Yixin Cao

With the rapid advancement of large reasoning models, long Chain-of-Thought (CoT) prompting has demonstrated strong performance on complex tasks. However, this often comes with a s…

cs.CL2025

How Do LLMs Perform Two-Hop Reasoning in Context?

Tianyu Guo, Hanlin Zhu, Ruiqi Zhang +4

``Socrates is human. All humans are mortal. Therefore, Socrates is mortal.'' This form of argument illustrates a typical pattern of two-hop reasoning. Formally, two-hop reasoning r…

cs.CL2024

Fast Best-of-N Decoding via Speculative Rejection

Hanshi Sun, Momin Haider, Ruiqi Zhang +6

The safe and effective deployment of Large Language Models (LLMs) involves a critical step called alignment, which ensures that the model's responses are in accordance with human p…

cs.CL2024

Simplicity Prevails: Rethinking Negative Preference Optimization for LLM Unlearning

Chongyu Fan, Jiancheng Liu, Licong Lin +4

This work studies the problem of large language model (LLM) unlearning, aiming to remove unwanted data influences (e.g., copyrighted or harmful content) while preserving model util…