8 papers
CODEBLOCK: Learning to Supervise Code at the Right Granularity
Zhijie Deng, Ling Li, Jinlong Pang +5
Supervised fine-tuning of code LLMs typically applies uniform cross-entropy loss to all response tokens, implicitly assuming that every token provides equally useful learning signa…
Analyzing the Effect of Noise in LLM Fine-tuning
Lingfang Li, Procheta Sen
Fine-tuning is the dominant paradigm for adapting pretrained large language models (LLMs) to downstream NLP tasks. In practice, fine-tuning datasets may contain various forms of no…
Reclaiming Low-Quality Instruction Data with Rule-Guided LLM Reconstruction
Zile Yang, Ling Li, Na Di +5
Data selection for supervised fine-tuning (SFT) often discards low-scoring instruction examples that may retain useful supervision. We propose a rule-guided LLM reconstruction fram…
Label Smoothing Improves Gradient Ascent in LLM Unlearning
Zirui Pang, Hao Zheng, Zhijie Deng +3
LLM unlearning has emerged as a promising approach, aiming to enable models to forget hazardous/undesired knowledge at low cost while preserving as much model utility as possible.…
OFFSIDE: Benchmarking Unlearning Misinformation in Multimodal Large Language Models
Hao Zheng, Zirui Pang, Ling li +5
Advances in Multimodal Large Language Models (MLLMs) intensify concerns about data privacy, making Machine Unlearning (MU), the selective removal of learned information, a critical…
LM-mixup: Text Data Augmentation via Language Model based Mixup
Zhijie Deng, Zhouan Shen, Ling Li +5
Instruction tuning is crucial for aligning Large Language Models (LLMs), yet the quality of instruction-following data varies significantly. While high-quality data is paramount, i…
SelectMix: Enhancing Label Noise Robustness through Targeted Sample Mixing
Qiuhao Liu, Ling Li, Yao Lu +3
Deep neural networks tend to memorize noisy labels, severely degrading their generalization performance. Although Mixup has demonstrated effectiveness in improving generalization a…
Recognition through Reasoning: Reinforcing Image Geo-localization with Large Vision-Language Models
Ling Li, Yao Zhou, Yuxuan Liang +2
Previous methods for image geo-localization have typically treated the task as either classification or retrieval, often relying on black-box decisions that lack interpretability.…