Loading
Life sciences · Preprint
arXiv · September 8, 2026
Posted before peer review. The findings may change or fail to hold.
This is an unrefereed preprint describing Training-Free Task Vectors (TFTVs), a method to edit large language model behavior without fine-tuning by mapping activation steering to weight-space edits. The authors report that TFTVs enable amplification, suppression, and composition of target behaviors while preserving general knowledge, but the work has not undergone peer review and no quantitative primary outcomes are disclosed in the abstract.
Preprint. Large language models used for behavioral control tasks. Intervention: Training-Free Task Vectors (TFTVs): a method to compute task-vector-like directions by mapping activation steering vectors to rank-one weight-space edits using forward-pass statistics only. Compared with: Other editing and steering baselines (not specifically named in abstract).
TFTVs enable computation of task-vector-like directions without requiring fine-tuning, using only forward-pass statistics Method supports arithmetic properties for learning via addition, forgetting via subtraction, and composition of multiple edits Empirically reported to amplify, suppress, and compose target behaviors while preserving general knowledge and problem-solving skills
Safety was not reported in the material analysed. Check the source before drawing any conclusion about harm.
The source did not state who this applies to in practice.
This is an unrefereed arXiv preprint presenting a novel computational method for language model editing without peer review; the technical contribution is clearly described but lacks independent validation.
As stated by the source record.
Graded across the dimensions that decide whether you should act, each from what the source actually supports. There is no single score, and where a dimension was not assessed it says so.
What is missing. This record has no reported figures. That is a gap in the analysis, not a judgement about the study.
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: tftv-llm.github.io.
Taken from the source record, never inferred. Follow any of these and new work involving them reaches your briefing.