DeCoVec: Building Decoding Space based Task Vector for Large Language Models via In-Context Learning
Feiyang Li, Yile Wang · 2026
Read DeCoVec: Building Decoding Space based Task Vector for Large Language Models via In-Context Learning by Feiyang Li, Yile Wang (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2604-11129.txt.
Description
Li and Wang propose DeCoVec, a training-free framework building decoding-space task vectors for language models. Task vectors encode behaviours as directions in model spaces, offering a steering tool. Existing approaches require fine-tuning or invasive changes to internal states, limiting freedom. DeCoVec captures the task essence as the gap between few-shot and zero-shot logit distributions, then steers generation by placing this vector into decoding. Tests across seven models show it consistently beats few-shot baselines, with gains up to five and a half points.
Catalog record
| CREATOR | Feiyang Li, Yile Wang |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 47141 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2604.11129