7 papers
FAME: Forecasting Academic Impact via Continuous-Time Manifold Evolution
Jianrong Ding, Jianyuan Zhong, Zhengyan Shi +1
Large Language Models (LLMs) are increasingly used to brainstorm and evaluate research ideas, yet assessing such judgments is fundamentally difficult because the true impact of a n…
Learning to Solve Complex Problems via Dataset Decomposition
Wanru Zhao, Lucas Caccia, Zhengyan Shi +3
Curriculum learning is a class of training strategies that organizes the data being exposed to a model by difficulty, gradually from simpler to more complex examples. This research…
Mixture of Noise for Pre-Trained Model-Based Class-Incremental Learning
Kai Jiang, Zhengyan Shi, Dell Zhang +2
Class Incremental Learning (CIL) aims to continuously learn new categories while retaining the knowledge of old ones. Pre-trained models (PTMs) show promising capabilities in CIL.…
Optimising Language Models for Downstream Tasks: A Post-Training Perspective
Zhengyan Shi
Language models (LMs) have demonstrated remarkable capabilities in NLP, yet adapting them efficiently and robustly to specific tasks remains challenging. As their scale and complex…
Understanding Likelihood Over-optimisation in Direct Alignment Algorithms
Zhengyan Shi, Sander Land, Acyr Locatelli +2
Direct Alignment Algorithms (DAAs), such as Direct Preference Optimisation (DPO) and Identity Preference Optimisation (IPO), have emerged as alternatives to online Reinforcement Le…
Instruction Tuning With Loss Over Instructions
Zhengyan Shi, Adam X. Yang, Bin Wu +3
Instruction tuning plays a crucial role in shaping the outputs of language models (LMs) to desired styles. In this work, we propose a simple yet effective method, Instruction Model…