10 papers
On the Role of Reasoning Patterns in the Generalization Discrepancy of Long Chain-of-Thought Supervised Fine-Tuning
Zhaoyi Li, Xiangyu Xi, Zhengyu Chen +6
Supervised Fine-Tuning (SFT) on long Chain-of-Thought (CoT) trajectories has become a pivotal phase in building large reasoning models. However, how CoT trajectories from different…
Scaling Reasoning Hop Exposes Weaknesses: Demystifying and Improving Hop Generalization in Large Language Models
Zhaoyi Li, Jiatong Li, Gangwei Jiang +3
Chain-of-thought (CoT) reasoning has become the standard paradigm for enabling Large Language Models (LLMs) to solve complex problems. However, recent studies reveal a sharp perfor…
What Makes a Good Reasoning Chain? Uncovering Structural Patterns in Long Chain-of-Thought Reasoning
Gangwei Jiang, Yahui Liu, Zhaoyi Li +5
Recent advances in reasoning with large language models (LLMs) have popularized Long Chain-of-Thought (LCoT), a strategy that encourages deliberate and step-by-step reasoning befor…
Learning to Substitute Components for Compositional Generalization
Zhaoyi Li, Gangwei Jiang, Chenwang Wu +3
Despite the rising prevalence of neural language models, recent empirical evidence suggests their deficiency in compositional generalization. One of the current de-facto solutions…
Unlocking the Power of Function Vectors for Characterizing and Mitigating Catastrophic Forgetting in Continual Instruction Tuning
Gangwei Jiang, Caigao Jiang, Zhaoyi Li +5
Catastrophic forgetting (CF) poses a significant challenge in machine learning, where a model forgets previously learned information upon learning new tasks. Despite the advanced c…
Mitigating the Language Mismatch and Repetition Issues in LLM-based Machine Translation via Model Editing
Weichuan Wang, Zhaoyi Li, Defu Lian +3
Large Language Models (LLMs) have recently revolutionized the NLP field, while they still fall short in some specific down-stream tasks. In the work, we focus on utilizing LLMs to…
Refine Large Language Model Fine-tuning via Instruction Vector
Gangwei Jiang, Zhaoyi Li, Defu Lian +1
Fine-tuning large language models (LLMs) can cause them to lose their general capabilities. However, the intrinsic mechanisms behind such forgetting remain unexplored. In this pape…
Benchmarking and Improving Compositional Generalization of Multi-aspect Controllable Text Generation
Tianqi Zhong, Zhaoyi Li, Quan Wang +4
Compositional generalization, representing the model's ability to generate text with new attribute combinations obtained by recombining single attributes from the training data, is…
Understanding and Patching Compositional Reasoning in LLMs
Zhaoyi Li, Gangwei Jiang, Hong Xie +3
LLMs have marked a revolutonary shift, yet they falter when faced with compositional reasoning tasks. Our research embarks on a quest to uncover the root causes of compositional re…
Towards Anytime Fine-tuning: Continually Pre-trained Language Models with Hypernetwork Prompt
Gangwei Jiang, Caigao Jiang, Siqiao Xue +4
Continual pre-training has been urgent for adapting a pre-trained model to a multitude of domains and tasks in the fast-evolving world. In practice, a continually pre-trained model…