L2CU: Learning To Complement Unseen Users
DILEEPA PITAWELA 1, GUSTAVO CARNEIRO 2, HSIANG-TING CHEN 1 · 2026
Read L2CU: Learning To Complement Unseen Users by DILEEPA PITAWELA 1, GUSTAVO CARNEIRO 2, HSIANG-TING CHEN 1 (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-2601-06119.txt.
Description
AI that complements a specific person's strengths must handle people it has never met. Dileepa Pitawela and 2 co-authors present L2CU in a 2026 paper on human-AI cooperative classification, where a person and a model label items together. Earlier methods assumed one global user model, ignoring individual differences.
L2CU works from sparse, noisy annotations to find representative annotator profiles, each capturing a distinct labeling pattern. A new user is matched to a profile, and a profile-specific model then complements that user's judgments. The combined accuracy improves over treating everyone alike. Tests across image and text datasets, including medical and news data, showed the framework works as a model-agnostic way to lift joint human-AI performance.
Catalog record
| CREATOR | DILEEPA PITAWELA 1, GUSTAVO CARNEIRO 2, HSIANG-TING CHEN 1 |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 54 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2601-06119