6 citations · 37 across the 30 of their papers we have counts for
33 papers
Rethinking the Trust Region in LLM Reinforcement Learning
Penghui Qi, Xiangxin Zhou, Zichen Liu +4
Reinforcement learning (RL) has become a cornerstone for fine-tuning Large Language Models (LLMs), with Proximal Policy Optimization (PPO) serving as the de facto standard algorith…
Defeating the Training-Inference Mismatch via FP16
Penghui Qi, Zichen Liu, Xiangxin Zhou +4
Reinforcement learning (RL) fine-tuning of large language models (LLMs) often suffers from instability due to the numerical mismatch between the training and inference policies. Wh…
Nonparametric Data Attribution for Diffusion Models
Yutian Zhao, Chao Du, Xiaosen Zheng +2
Data attribution for generative models seeks to quantify the influence of individual training examples on model outputs. Existing methods for diffusion models typically require acc…
Variational Reasoning for Language Models
Xiangxin Zhou, Zichen Liu, Haonan Wang +5
We introduce a variational reasoning framework for language models that treats thinking traces as latent variables and optimizes them through variational inference. Starting from t…
Language Models Can Learn from Verbal Feedback Without Scalar Rewards
Renjie Luo, Zichen Liu, Xiangyan Liu +5
LLMs are often trained with RL from human or AI feedback, yet such methods typically compress nuanced feedback into scalar rewards, discarding much of their richness and inducing s…
Continual Reinforcement Learning by Planning with Online World Models
Zichen Liu, Guoji Fu, Chao Du +2
Continual reinforcement learning (CRL) refers to a naturalistic setting where an agent needs to endlessly evolve, by trial and error, to solve multiple tasks that are presented seq…
Reinforcing General Reasoning without Verifiers
Xiangxin Zhou, Zichen Liu, Anya Sims +6
The recent paradigm shift towards training large language models (LLMs) using DeepSeek-R1-Zero-style reinforcement learning (RL) on verifiable rewards has led to impressive advance…
Lifelong Safety Alignment for Language Models
Haoyu Wang, Zeyu Qin, Yifei Zhao +4
LLMs have made impressive progress, but their growing capabilities also expose them to highly flexible jailbreaking attacks designed to bypass safety alignment. While many existing…
Optimizing Anytime Reasoning via Budget Relative Policy Optimization
Penghui Qi, Zichen Liu, Tianyu Pang +3
Scaling test-time compute is crucial for enhancing the reasoning capabilities of large language models (LLMs). Existing approaches typically employ reinforcement learning (RL) to m…
FlowReasoner: Reinforcing Query-Level Meta-Agents
Hongcheng Gao, Yue Liu, Yufei He +6
This paper proposes a query-level meta-agent named FlowReasoner to automate the design of query-level multi-agent systems, i.e., one system per user query. Our core idea is to ince…
Understanding R1-Zero-Like Training: A Critical Perspective
Zichen Liu, Changyu Chen, Wenjun Li +5
DeepSeek-R1-Zero has shown that reinforcement learning (RL) at scale can directly enhance the reasoning capabilities of LLMs without supervised fine-tuning. In this work, we critic…
Improving Your Model Ranking on Chatbot Arena by Vote Rigging
Rui Min, Tianyu Pang, Chao Du +3
Chatbot Arena is a popular platform for evaluating LLMs by pairwise battles, where users vote for their preferred response from two randomly sampled anonymous models. While Chatbot…
Sample-Efficient Alignment for LLMs
Zichen Liu, Changyu Chen, Chao Du +2
We study methods for efficiently aligning large language models (LLMs) with human preferences given budgeted online feedback. We first formulate the LLM alignment problem in the fr…
Denial-of-Service Poisoning Attacks against Large Language Models
Kuofeng Gao, Tianyu Pang, Chao Du +3
Recent studies have shown that LLMs are vulnerable to denial-of-service (DoS) attacks, where adversarial inputs like spelling errors or non-semantic prompts trigger endless outputs…
Scaling up Masked Diffusion Models on Text
Shen Nie, Fengqi Zhu, Chao Du +5
Masked diffusion models (MDMs) have shown promise in language modeling, yet their scalability and effectiveness in core language tasks, such as text generation and language underst…
LightTransfer: Your Long-Context LLM is Secretly a Hybrid Model with Effortless Adaptation
Xuan Zhang, Fengzhuo Zhang, Cunxiao Du +4
Scaling language models to handle longer contexts introduces substantial memory challenges due to the growing cost of key-value (KV) caches. Motivated by the efficiency gains of hy…
Meta-Unlearning on Diffusion Models: Preventing Relearning Unlearned Concepts
Hongcheng Gao, Tianyu Pang, Chao Du +3
With the rapid progress of diffusion-based content generation, significant efforts are being made to unlearn harmful or copyrighted concepts from pretrained diffusion models (DMs)…
When Attention Sink Emerges in Language Models: An Empirical View
Xiangming Gu, Tianyu Pang, Chao Du +5
Language Models (LMs) assign significant attention to the first token, even if it is not semantically important, which is known as attention sink. This phenomenon has been widely a…
A Closer Look at Machine Unlearning for Large Language Models
Xiaojian Yuan, Tianyu Pang, Chao Du +3
Large language models (LLMs) may memorize sensitive or copyrighted content, raising privacy and legal concerns. Due to the high cost of retraining from scratch, researchers attempt…
Cheating Automatic LLM Benchmarks: Null Models Achieve High Win Rates
Xiaosen Zheng, Tianyu Pang, Chao Du +3
Automatic LLM benchmarks, such as AlpacaEval 2.0, Arena-Hard-Auto, and MT-Bench, have become popular for evaluating language models due to their cost-effectiveness and scalability…
Improved Techniques for Optimization-Based Jailbreaking on Large Language Models
Xiaojun Jia, Tianyu Pang, Chao Du +5
Large language models (LLMs) are being rapidly developed, and a key component of their widespread deployment is their safety-related alignment. Many red-teaming efforts aim to jail…
Bootstrapping Language Models with DPO Implicit Rewards
Changyu Chen, Zichen Liu, Chao Du +5
Human alignment in large language models (LLMs) is an active area of research. A recent groundbreaking work, direct preference optimization (DPO), has greatly simplified the proces…
Chain of Preference Optimization: Improving Chain-of-Thought Reasoning in LLMs
Xuan Zhang, Chao Du, Tianyu Pang +3
The recent development of chain-of-thought (CoT) decoding has enabled large language models (LLMs) to generate explicit logical reasoning paths for complex problem-solving. However…
Improved Few-Shot Jailbreaking Can Circumvent Aligned Language Models and Their Defenses
Xiaosen Zheng, Tianyu Pang, Chao Du +3
Recently, Anil et al. (2024) show that many-shot (up to hundreds of) demonstrations can jailbreak state-of-the-art LLMs by exploiting their long-context capability. Nevertheless, i…