Explainable Token-level Noise Filtering for LLM Fine-tuning Datasets

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Yang, Yuchen, Lin, Wenze, Huang, Enhao, Chu, Zhixuan, Zhou, Hongbin, Tao, Lan, Li, Yiming, Qin, Zhan, Ren, Kui
Format: Preprint
Published: 2026
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866911568289071104
author Yang, Yuchen
Lin, Wenze
Huang, Enhao
Chu, Zhixuan
Zhou, Hongbin
Tao, Lan
Li, Yiming
Qin, Zhan
Ren, Kui
author_facet Yang, Yuchen
Lin, Wenze
Huang, Enhao
Chu, Zhixuan
Zhou, Hongbin
Tao, Lan
Li, Yiming
Qin, Zhan
Ren, Kui
contents Large Language Models (LLMs) have seen remarkable advancements, achieving state-of-the-art results in diverse applications. Fine-tuning, an important step for adapting LLMs to specific downstream tasks, typically involves further training on corresponding datasets. However, a fundamental discrepancy exists between current fine-tuning datasets and the token-level optimization mechanism of LLMs: most datasets are designed at the sentence-level, which introduces token-level noise, causing negative influence to final performance. In this paper, we propose XTF, an explainable token-level noise filtering framework. XTF decomposes the complex and subtle contributions of token-level data to the fine-tuning process into three distinct and explicit attributes (reasoning importance, knowledge novelty, and task relevance), which can be assessed using scoring methods, and then masks the gradients of selected noisy tokens accordingly to optimize the performance of fine-tuned LLMs. We conduct extensive experiments on three representative downstream tasks (math, code and medicine) across 7 mainstream LLMs. The results demonstrate that XTF can significantly improve downstream performance by up to 13.7% compared to regular fine-tuning. Our work highlights the importance of token-level dataset optimization, and demonstrates the potential of strategies based on attribute decomposition for explaining complex training mechanisms.
format Preprint
id arxiv_https___arxiv_org_abs_2602_14536
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Explainable Token-level Noise Filtering for LLM Fine-tuning Datasets
Yang, Yuchen
Lin, Wenze
Huang, Enhao
Chu, Zhixuan
Zhou, Hongbin
Tao, Lan
Li, Yiming
Qin, Zhan
Ren, Kui
Computation and Language
Artificial Intelligence
Large Language Models (LLMs) have seen remarkable advancements, achieving state-of-the-art results in diverse applications. Fine-tuning, an important step for adapting LLMs to specific downstream tasks, typically involves further training on corresponding datasets. However, a fundamental discrepancy exists between current fine-tuning datasets and the token-level optimization mechanism of LLMs: most datasets are designed at the sentence-level, which introduces token-level noise, causing negative influence to final performance. In this paper, we propose XTF, an explainable token-level noise filtering framework. XTF decomposes the complex and subtle contributions of token-level data to the fine-tuning process into three distinct and explicit attributes (reasoning importance, knowledge novelty, and task relevance), which can be assessed using scoring methods, and then masks the gradients of selected noisy tokens accordingly to optimize the performance of fine-tuned LLMs. We conduct extensive experiments on three representative downstream tasks (math, code and medicine) across 7 mainstream LLMs. The results demonstrate that XTF can significantly improve downstream performance by up to 13.7% compared to regular fine-tuning. Our work highlights the importance of token-level dataset optimization, and demonstrates the potential of strategies based on attribute decomposition for explaining complex training mechanisms.
title Explainable Token-level Noise Filtering for LLM Fine-tuning Datasets
topic Computation and Language
Artificial Intelligence
url https://arxiv.org/abs/2602.14536