Benchmarking Few-shot Transferability of Pre-trained Models with Improved Evaluation Protocols

Xu Luo, Ji Zhang, Lianli Gao, Heng Tao Shen and Jingkuan Song · 2026

Read Benchmarking Few-shot Transferability of Pre-trained Models with Improved Evaluation Protocols by Xu Luo, Ji Zhang, Lianli Gao, Heng Tao Shen and Jingkuan Song (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-2603-00478.txt.

Description

Luo and four colleagues set a rigorous ruler for few shot transfer, where no shared challenging protocol exists for real world use. Their FEWTRANS benchmark spans ten varied data sets, and their Hyperparameter Ensemble protocol counters the validation set illusion in data scarce settings. Tests show the pretrained model choice dominates results, while fancy transfer methods barely beat simple full fine tuning, which works through tiny distributed tweaks and flexible reshaping of high level meaning. Multimodal models collapse in specialised domains where wording grows rare.

Catalog record

CREATORXu Luo, Ji Zhang, Lianli Gao, Heng Tao Shen and Jingkuan Song
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE61862
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2603.00478

Browse the full library · About the Tower of Babel