Atyaephyra at SemEval-2025 Task 4: Low-Rank Negative Preference Optimization
The result's identifiers
Result code in IS VaVaI
<a href="https://www.isvavai.cz/riv?ss=detail&h=RIV%2F00216208%3A11320%2F25%3A10511573" target="_blank" >RIV/00216208:11320/25:10511573 - isvavai.cz</a>
Result on the web
<a href="https://aclanthology.org/2025.semeval-1.187/" target="_blank" >https://aclanthology.org/2025.semeval-1.187/</a>
DOI - Digital Object Identifier
—
Alternative languages
Result language
angličtina
Original language name
Atyaephyra at SemEval-2025 Task 4: Low-Rank Negative Preference Optimization
Original language description
We present a submission to the SemEval 2025 shared task on unlearning sensitive content from LLMs. Our approach employs negative preference optimization using low-rank adaptation. We show that we can utilize this combination to efficiently compute additional regularization terms, which help with unlearning stabilization. The results of our approach significantly exceed the shared task baselines.
Czech name
—
Czech description
—
Classification
Type
D - Article in proceedings
CEP classification
—
OECD FORD branch
10201 - Computer sciences, information science, bioinformathics (hardware development to be 2.2, social aspect to be 5.8)
Result continuities
Project
<a href="/en/project/EH23_025%2F0008691" target="_blank" >EH23_025/0008691: Human-centred AI for a Sustainable and Adaptive Society</a><br>
Continuities
P - Projekt vyzkumu a vyvoje financovany z verejnych zdroju (s odkazem do CEP)<br>I - Institucionalni podpora na dlouhodoby koncepcni rozvoj vyzkumne organizace
Others
Publication year
2025
Confidentiality
S - Úplné a pravdivé údaje o projektu nepodléhají ochraně podle zvláštních právních předpisů
Data specific for result type
Article name in the collection
Proceedings of the 19th International Workshop on Semantic Evaluation (SemEval-2025)
ISBN
979-8-89176-273-2
ISSN
—
e-ISSN
—
Number of pages
8
Pages from-to
1415-1422
Publisher name
Association for Computational Lingustics
Place of publication
Kerrville, TX, USA
Event location
Wien, Austria
Event date
Jul 31, 2025
Type of event by nationality
WRD - Celosvětová akce
UT code for WoS article
—