# When Style Similarity Scores Fail: Diagnosing Raw CSD Cosine in Artist-Style Evaluation

arXiv: https://arxiv.org/abs/2605.09030 · HTML: https://mlsanity.com/paper/2605.09030/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 14 of its 18 references that are in the map.
- First posted on arXiv 9 May 2026.

## Builds on (most cited first, 14 of 14)

- [Learning Transferable Visual Models From Natural Language Supervision](https://mlsanity.com/paper/2103.00020/) (2021)
- [DINOv2: Learning Robust Visual Features without Supervision](https://mlsanity.com/paper/2304.07193/) (2023)
- [Sigmoid Loss for Language Image Pre-Training](https://mlsanity.com/paper/2303.15343/) (2023)
- [FaceNet: A Unified Embedding for Face Recognition and Clustering](https://mlsanity.com/paper/1503.03832/) (2015)
- [Word Translation Without Parallel Data](https://mlsanity.com/paper/1710.04087/) (2017)
- [Supervised Learning of Universal Sentence Representations from Natural Language Inference Data](https://mlsanity.com/paper/1705.02364/) (2017)
- [Estimating the intrinsic dimension of datasets by a minimal neighborhood information](https://mlsanity.com/paper/1803.06992/) (2018)
- [All-but-the-Top: Simple and Effective Postprocessing for Word Representations](https://mlsanity.com/paper/1702.01417/) (2017)
- [Measuring Style Similarity in Diffusion Models](https://mlsanity.com/paper/2404.01292/) (2024)
- [The ArtBench Dataset: Benchmarking Generative Models with Artworks](https://mlsanity.com/paper/2206.11404/) (2022)
- [InstantStyle-Plus: Style Transfer with Content-Preserving in Text-to-Image Generation](https://mlsanity.com/paper/2407.00788/) (2024)
- [UnZipLoRA: Separating Content and Style from a Single Image](https://mlsanity.com/paper/2412.04465/) (2024)
- [Magic Insert: Style-Aware Drag-and-Drop](https://mlsanity.com/paper/2407.02489/) (2024)
- [Low-Rank Continual Personalization of Diffusion Models](https://mlsanity.com/paper/2410.04891/) (2024)

## Abstract (arXiv)

Raw cosine in the 768-dimensional output space of the Contrastive Style Descriptor (CSD) is now widely read as an absolute, calibrated style-fidelity score for text-to-image and style-imitation evaluation. We introduce the discrimination gap, a corpus-internal, prototype-free and threshold-free diagnostic that tests whether contrastive style cosines admit an absolute same-versus-different interpretation on a candidate artist corpus. On a 1799-artwork, 91-artist public-domain corpus, raw CSD cosine yields negative point-estimate gaps for 23/91 artists at the pairwise level (2/91 robust under bootstrap) and for 15/91 in the aggregated-pool scoring regime style-fidelity evaluations typically use. CSLS readout on the frozen backbone reduces the aggregated negative-gap count to 4/91; combined with positional-embedding interpolation to 336 pixels it raises unsupervised pair-verification AUC from 0.883 to 0.905 across 25 artist-disjoint splits. We refer to this diagnostic-driven readout protocol on the frozen backbone (CSLS as default, pos-interp 336 as the stronger optional setting) as CSD+, not a new encoder.A cross-backbone check on CLIP-ViT-L/14, SigLIP-large and DINOv2-Large reproduces the same shared-tradition failure pattern, providing evidence that the residual reflects a shared limitation of the four backbones we tested rather than a CSD-specific artefact. Practical implication: before reporting CSD cosine as an absolute style-fidelity score, run the diagnostic on the candidate corpus; CSLS is the minimal correction when it fails.