Metraipolis › archive › dc_44f512acfc12
UNBIND: UNlearning By INference-time Directional Steering for Code LLMs — arXiv:2609.35913
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2609.35913
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-07T16:24:59.905Z
- SHA-256 of the archived text
5b9a658ad7925b2f91feb06044f6bd90fde1d047756973480b59027b03c191ac- Stable URL of this page
https://metraipolis.com/documento/dc_44f512acfc12
How to cite
UNBIND: UNlearning By INference-time Directional Steering for Code LLMs — arXiv:2609.35913 — arXiv (en), https://arxiv.org/abs/2609.35913 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_44f512acfc12 (Metraipolis archive dc_44f512acfc12, sha256 5b9a658ad7925b2f91feb06044f6bd90fde1d047756973480b59027b03c191ac)
BibTeX
@misc{dc44f512acfc12,
title = {UNBIND: UNlearning By INference-time Directional Steering for Code LLMs — arXiv:2609.35913},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2609.35913},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_44f512acfc12; sha256 5b9a658ad7925b2f91feb06044f6bd90fde1d047756973480b59027b03c191ac},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.35913. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
Code large language models acquire programming capabilities from large code corpora, but can also memorize implementations that later require removal. Code unlearning is needed to control their continued reproduction when copyright or security concerns arise. However, targeted and retained code share computational patterns, creating a tension between forgetting specific implementations and preserving general programming ability. We propose \textbf{UNBIND}, a code unlearning framework that separately considers which hidden states correspond to the target code and how to suppress its reproduction. By constructing separate directions for these objectives, UNBIND achieves selective unlearning at inference time while keeping model weights fixed. Our evaluation covers fourteen baselines across two code models and two corpora. UNBIND achieves the highest joint forgetting and utility score in every setting. It reduces target code reproduction by 97.3\% to 99.1\% as measured by F-BLEU, with at most two fewer HumanEval+ and six fewer MBPP+ problems solved than the original models. In repeated extraction tests under a fixed budget, the number of targets yielding exact spans of at least 50 tokens falls from 188--262 to 0--2 out of 300 per setting. No extracted span reaches 100 tokens, and the mean best recovery ratio ranges from 0.43\% to 6.45\%. Multilingual and related-code evaluations further show effective forgetting with limited impact on useful programming capabilities, supporting UNBIND as a practical approach to selective code unlearning.
Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-09-30 in cs.AI).
Categorie arXiv: Software Engineering (cs.SE) ; Artificial Intelligence (cs.AI).
Pagina del lavoro: https://arxiv.org/abs/2609.35913 — autori e testo completo sono lì (qui omessi i nomi: nessun dato personale nel deposito).
Licenza: titolo e abstract sono metadati arXiv distribuiti in CC0 1.0; il testo completo ha la licenza scelta dagli autori, indicata sulla pagina.
Letto il 2026-10-07T15:54:53.739Z dalla pagina https://arxiv.org/catchup/cs.AI/2026-09-30?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).