Minor SFT loss for LLM fine-tune to increase performance and reduce model deviation

TL;DR

This paper introduces MinorSFT, a new loss function for supervised fine-tuning of large language models, aiming to enhance performance and minimize deviation from the original model.

Contribution

It proposes a novel training metric and loss function inspired by DPO and MinorDPO to improve fine-tuning efficiency and model stability.

Findings

MinorSFT reduces model deviation during fine-tuning

Improves alignment with human preferences

Enhances training effectiveness

Abstract

Instruct LLM provide a paradigm used in large scale language model to align LLM to human preference. The paradigm contains supervised fine tuning and reinforce learning from human feedback. This paradigm is also used in downstream scenarios to adapt LLM to specific corpora and applications. Comparing to SFT, there are many efforts focused on RLHF and several algorithms being proposed, such as PPO, DPO, IPO, KTO, MinorDPO and etc. Meanwhile most efforts for SFT are focused on how to collect, filter and mix high quality data. In this article with insight from DPO and MinorDPO, we propose a training metric for SFT to measure the discrepancy between the optimized model and the original model, and a loss function MinorSFT that can increase the training effectiveness, and reduce the discrepancy between the optimized LLM and original LLM.