arXiv cs.AI· Gabriel J. Perin, Lucas Boscaini, Andr\'e Araujo, Nina S. T. Hirata·· 2 天前AI 评分31
无需微调的 TFTV：用前向统计实现 LLM 行为控制
Training-Free Task Vectors for LLM Behavioral Control
AI 导读
研究者提出 Training-Free Task Vectors（TFTVs），无需微调即可计算类任务向量方向，仅用前向传播统计将激活引导向量映射为秩一权重空间编辑，并支持加法学习、减法遗忘与多编辑组合。在 LLM 行为控制任务上，TFTVs 能持续放大、抑制和组合目标行为，同时保留通用知识与解题能力，相比其他编辑与引导基线实现了更强的特质控制与更优或相当的效用保持。代码已在项目网站公开。
正文 · 原文
Abstract:Task vectors enable post-training model editing by identifying semantically meaningful directions in weight space, typically computed as the difference between a fine-tuned model and its pretrained initialization. However, this reliance on fine-tuning makes discovering such directions costly and limits the practicality of post-training model editing. To address this limitation, we introduce Training-Free Task Vectors (TFTVs), a novel method to compute task-vector-like directions without requiring fine-tuning. Our method maps activation steering vectors to rank-one weight-space edits using only forward-pass statistics, while satisfying arithmetic properties that directly support learning via addition, forgetting via subtraction, and the composition of multiple edits. Empirically, we evaluate TFTVs on large language model behavioral control tasks and show that they consistently amplify, suppress, and compose target behaviors while preserving general knowledge and problem-solving skills. We also validate our method against other editing and steering baselines, experimentally demonstrating that TFTVs achieve stronger trait control with better or competitive utility preservation. We hope our work opens new directions for the community in post-training model editing and broader training-free model control. Code is available on the project website: this http URL.
| Subjects: | Machine Learning (cs.LG); Artificial Intelligence (cs.AI) |
| Cite as: | arXiv:2609.09054 [cs.LG] |
| (or arXiv:2609.09054v2 [cs.LG] for this version) | |
| https://doi.org/10.48550/arXiv.2609.09054 arXiv-issued DOI via DataCite |
Submission history
From: Gabriel Perin [view email]
[v1]
Tue, 8 Sep 2026 17:10:28 UTC (683 KB)
[v2]
Wed, 14 Oct 2026 17:32:51 UTC (659 KB)
来源：arXiv cs.AI · arxiv.org