Baichuan2-Sum: Instruction Finetune Baichuan2-7B Model for Dialogue Summarization

TL;DR

This work addresses role-based dialogue summarization by fine-tuning the Baichuan2-7B model with explicit instruction prompts (Agent/User/All) and introducing NEFTune, which injects uniform embedding noise scaled by $\alpha/\sqrt{Ld}$ with $\alpha=5$ and noise in $[-1,1]$. The proposed Baichuan2-Sum achieves state-of-the-art results on CSDS and SAMSUM, validated through both automatic metrics (ROUGE, BLEU, BERTScore) and human judgments, with efficient training on a single RTX 4090. The methodology combines tokenization advantages of Baichuan2, RoPE-based long-context modeling, xFormers for memory-efficient attention, and LoRA-based parameter-efficient fine-tuning to realize strong performance. This work demonstrates the practicality and effectiveness of instruction-following large-language-model techniques for complex, multi-role dialogue summarization and provides publicly available code for reproducibility and extension.

Abstract

Large language models (LLMs) like Llama, Baichuan and Bloom models show remarkable ability with instruction fine-tuning in many natural language tasks. Nevertheless, for the dialogue summarization task, which aims to generate summaries for different roles in dialogue, most of the state-of-the-art methods conduct on small models (e.g Bart and Bert). Existing methods try to add task specified optimization on small models like adding global-local centrality score to models. In this paper, we propose an instruction fine-tuning model: Baichuan2-Sum, for role-oriented diaglouge summarization. By setting different instructions for different roles, the model can learn from the dialogue interactions and output the expected summaries. Furthermore, we applied NEFTune technique to add suitable noise during training to improve the results. The experiments demonstrate that the proposed model achieves the new state-of-the-art results on two public dialogue summarization datasets: CSDS and SAMSUM. We release our model and related codes to facilitate future studies on dialogue summarization task.

Figures (6)

• Figure 1: A customer service dialogue example of CSDS dataset
• Figure 2: Baichuan2 Achitecture with NEFTuning
• Figure 3: Distribution of Samples in CSDS and SAMSUM Datasets
• Figure 4: Token Length Distribution for CSDS Dataset
• Figure 5: Token Length Distribution for SAMSUM Dataset