Logits-Based Finetuning
This work addresses the problem of capturing token-level dependencies and linguistic diversity in fine-tuning for researchers and practitioners in natural language processing, representing an incremental improvement over existing methods.
The paper tackles the limitations of traditional supervised fine-tuning for large language models by proposing a logits-based fine-tuning framework that combines teacher logits with ground truth labels, resulting in accuracy gains of 18% on Mawps and 22.7% on TabMWP, with an average improvement of 7.28% across nine mathematical benchmarks.
In recent years, developing compact and efficient large language models (LLMs) has emerged as a thriving area of research. Traditional Supervised Fine-Tuning (SFT), which relies on singular ground truth labels, often fails to capture token-level dependencies and linguistic diversity. To address these limitations, we propose a logits-based fine-tuning framework that integrates the strengths of supervised learning and knowledge distillation. Our approach constructs enriched training targets by combining teacher logits with ground truth labels, preserving both correctness and linguistic diversity. This ensures more reliable and effective training. We constructed a large-scale 1.2M logits dataset and trained a series of science-focused models. Experimental results demonstrate that our method achieves significant improvements, with accuracy gains of 18% on Mawps and 22.7% on TabMWP. Across nine widely used mathematical benchmarks, our method consistently outperforms prior SFT models, achieving an average improvement of 7.28%. Codes are available at https://github.com/dvlab-research/Logits-Based-Finetuning.