arXiv cs.CL· Amit Jadhav, Shaurya Beriwala, Beomjin Kim·· 12 小时前AI 评分31
Undertone-P：用少量评分个性化冻结 VLM 美学评判器
Undertone: Personalizing Frozen VLM Aesthetic Judges from Few Ratings
AI 导读
Undertone-P 提出从其他评分者学习逐人回归权重的高斯先验，只需少量评分即可个性化冻结 VLM 美学评判器，新用户仅需一次闭式求解、无需训练网络。在 PARA 基准上，仅用先验即达 Spearman 相关 .640，超过已发表最佳的 10 评分（.581）和 100 评分（.620）结果；加入个人 10 和 100 条评分后分别升至 .648 和 .676。
正文 · 原文
Abstract:Image quality assessment predicts a mean opinion score (MOS), but for image aesthetics the mean hides the individual, so personalized assessment must start from the few images a new user rates. Fitting each person from scratch fails with ten ratings; networks fine-tuned per person carry over poorly to new kinds of images. We present Undertone-P, which learns from other raters a Gaussian prior over per-person regression weights on read-outs of frozen vision-language model (VLM) judges: the probability-weighted rating and the hidden state at the score token. Since people differ from the crowd in ways other raters reveal, a few ratings need only choose among those ways: a new user costs one closed-form solve, and no network is trained. On Ryu and Yanaka's PARA benchmark, the prior alone reaches a Spearman correlation of .640, above the best published 10- and 100-rating results (.581 and .620); 10 and 100 of a person's ratings raise it to .648 and .676. These ratings add most where tastes depart from the crowd: with 100 ratings, +.067 for the raters who agree least with it and +.206 on LAPIS artworks, against +.037 overall. Without labels, the judges' expected scores predict PARA's MOS at .835 (best published model we ran: .694). PARA-trained networks still lead under PARA's original protocol, likely through a stronger end-to-end MOS model.
| Comments: | v2: substantially revised and retitled; supersedes v1 ("Feed the Panel Dimensions, Not Verdicts: Rubric-Decomposed Fusion of Vision-Language Aesthetic Judges"). 4 pages + references, 2 figures, 2 tables |
| Subjects: | Computer Vision and Pattern Recognition (cs.CV); Computation and Language (cs.CL); Machine Learning (cs.LG) |
| Cite as: | arXiv:2609.27110 [cs.CV] |
| (or arXiv:2609.27110v2 [cs.CV] for this version) | |
| https://doi.org/10.48550/arXiv.2609.27110 arXiv-issued DOI via DataCite |
Submission history
From: Amit Jadhav [view email]
[v1]
Tue, 22 Sep 2026 22:07:52 UTC (85 KB)
[v2]
Thu, 5 Nov 2026 00:33:14 UTC (96 KB)
来源：arXiv cs.CL · arxiv.org