Learning to Evaluate: Cost-Effective Model Evaluation on Unlabeled Data with Meta-Learning DOI: 10.1145/3770855.3818104 ISBN: 979-8-4007-2258-5/2026/08 Conference: Proceedings of the 32nd ACM SIGKDD Conference on Knowledge Discovery and Data Mining V.2; August 09–13, 2026; Jeju Island, Republic of Korea Proceedings of the 32nd ACM SIGKDD Conference on Knowledge Discovery and Data Mining V.2 (KDD ’26), August 09–13, 2026, Jeju Island, Republic of Korea CCS: Computing methodologies Machine learning
Trinh Pham, Viet Huynh, Hongzhi Yin, Quoc Viet Hung Nguyen, Thanh Tam Nguyen · 2026
Read Learning to Evaluate: Cost-Effective Model Evaluation on Unlabeled Data with Meta-Learning DOI: 10.1145/3770855.3818104 ISBN: 979-8-4007-2258-5/2026/08 Conference: Proceedings of the 32nd ACM SIGKDD Conference on Knowledge Discovery and Data Mining V.2; August 09–13, 2026; Jeju Island, Republic of Korea Proceedings of the 32nd ACM SIGKDD Conference on Knowledge Discovery and Data Mining V.2 (KDD ’26), August 09–13, 2026, Jeju Island, Republic of Korea CCS: Computing methodologies Machine learning by Trinh Pham, Viet Huynh, Hongzhi Yin, Quoc Viet Hung Nguyen, Thanh Tam Nguyen (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-2605-23595.txt.
Description
Pham and colleagues introduce MetaEvaluator, a cheap, model-agnostic framework for fast label-free evaluation of unseen models on unlabeled data. Current routes lean on costly labels, repeated fine-tuning, or assumptions that fail on new models. It meta-learns over a pool of reference models to gain a good starting point for judging unseen models, spreading test cost and removing per-model retraining. Tests show stable, accurate estimates at far lower cost than conventional routes, enabling scalable bench work for new models.
Catalog record
| CREATOR | Trinh Pham, Viet Huynh, Hongzhi Yin, Quoc Viet Hung Nguyen, Thanh Tam Nguyen |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 51972 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2605.23595