Relational Knowledge Distillation Using Fine-tuned Function Vectors

Andrea Kang, Yingnian Wu, Hongijng Lu · 2026

Read Relational Knowledge Distillation Using Fine-tuned Function Vectors by Andrea Kang, Yingnian Wu, Hongijng Lu (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-2601-08169.txt.

Description

Andrea Kang, Yingnian Wu, and Hongjing Lu study how models store relations between concepts. Their 2026 arXiv paper builds on causal mediation work in in-context learning. That work showed a small set of attention heads encodes task representation. The result is stored in a compact form called the function vector. Kang and co-authors use fine-tuned function vectors for relational knowledge distillation. The goal is to move relational skill from one model setup to another. Relations sit at the core of how intelligent systems make sense of the world.

Catalog record

CREATORAndrea Kang, Yingnian Wu, Hongijng Lu
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE64 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2601-08169

Browse the full library · About the Tower of Babel