11 papers
LSR-Ben: A Logical and Scientific Reasoning Benchmark for Evaluating Process Reward Models
Zhouhao Sun, Xuan Zhang, Xiao Ding +10
Currently, process reward models (PRMs) have exhibited remarkable potential for test-time scaling. Since large language models (LLMs) regularly generate flawed intermediate reasoni…
Consolidation or Adaptation? PRISM: Disentangling SFT and RL Data via Gradient Concentration
Yang Zhao, Yangou Ouyang, Xiao Ding +8
While Hybrid Supervised Fine-Tuning (SFT) followed by Reinforcement Learning (RL) has become the standard paradigm for training LLM agents, effective mechanisms for data allocation…
MAESTRO: Meta-learning Adaptive Estimation of Scalarization Trade-offs for Reward Optimization
Yang Zhao, Hepeng Wang, Xiao Ding +8
Group-Relative Policy Optimization (GRPO) has emerged as an efficient paradigm for aligning Large Language Models (LLMs), yet its efficacy is primarily confined to domains with ver…
Tool Zero: Training Tool-Augmented LLMs via Pure RL from Scratch
Yirong Zeng, Xiao Ding, Yutai Hou +9
Training tool-augmented LLMs has emerged as a promising approach to enhancing language models' capabilities for complex tasks. The current supervised fine-tuning paradigm relies on…
Com: A Causal-Guided Benchmark for Exploring Complex Commonsense Reasoning in Large Language Models
Kai Xiong, Xiao Ding, Yixin Cao +7
Large language models (LLMs) have mastered abundant simple and explicit commonsense knowledge through pre-training, enabling them to achieve human-like performance in simple common…
Information Gain-Guided Causal Intervention for Autonomous Debiasing Large Language Models
Zhouhao Sun, Xiao Ding, Li Du +5
Despite significant progress, recent studies indicate that current large language models (LLMs) may still capture dataset biases and utilize them during inference, leading to the p…
UFO-RL: Uncertainty-Focused Optimization for Efficient Reinforcement Learning Data Selection
Yang Zhao, Kai Xiong, Xiao Ding +9
Scaling RL for LLMs is computationally expensive, largely due to multi-sampling for policy optimization and evaluation, making efficient data selection crucial. Inspired by the Zon…
Large Language Models Are Still Misled by Simple Bias Ensembles
Zhouhao Sun, Zhiyuan Kan, Xiao Ding +5
With the evolution of large language models (LLMs), their robustness against individual simple biases has been enhanced. However, we observe that the ensemble of multiple simple bi…
Beyond Similarity: A Gradient-based Graph Method for Instruction Tuning Data Selection
Yang Zhao, Li Du, Xiao Ding +10
Large language models (LLMs) have shown great potential across various industries due to their remarkable ability to generalize through instruction tuning. However, the limited ava…
Supervised Fine-Tuning Achieve Rapid Task Adaption Via Alternating Attention Head Activation Patterns
Yang Zhao, Li Du, Xiao Ding +3
LLMs' performance on complex tasks is still unsatisfactory. A key issue is that presently LLMs learn in a data-driven schema, while the instructions about these complex tasks are b…
Towards Generalizable and Faithful Logic Reasoning over Natural Language via Resolution Refutation
Zhouhao Sun, Xiao Ding, Li Du +4
Large language models (LLMs) have achieved significant performance in various natural language reasoning tasks. However, they still struggle with performing first-order logic reaso…