Style or Content? Evaluating Style Classifiers with Controlled Content Overlap

• URL: http://arxiv.org/abs/2606.07103v1
• Date: Fri, 05 Jun 2026 09:53:51 GMT
• Title: Style or Content? Evaluating Style Classifiers with Controlled Content Overlap
• Abstract summary: We study a problem with a controlled content overlap setup built on parallel Bible translations.<n>We define the overlap parameter $$ as the normalized residual of mutual information between content identity and style label.<n>We show that low-overlap models degrade when content cues are removed, while high-overlap models transfer more robustly.
• Score: 16.209371959740952
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Style classifiers can use content cues that correlate with style labels in naturally collected data, yet we lack a systematic way to measure this reliance. We study this problem with a controlled content overlap setup built on parallel Bible translations. Specifically, we define the overlap parameter $α$ as the normalized residual of mutual information between content identity and style label, so that it measures how much content is shared across style classes: from no shared content ($α=0$) to fully shared content ($α=1$). Cross-overlap evaluation of RoBERTa-based classifiers shows that low-overlap models degrade when content cues are removed, while high-overlap models transfer more robustly. A cross-style content retrieval probe further shows that content becomes less recoverable as $α$ increases, with training dynamics showing this removal occurs gradually. Together, these results suggest that controlled overlap provides a simple diagnostic for separating style learning from content shortcuts.

Related papers

• Style-CCL: Content-Preserving Style Transfer via Curriculum Continual Learning [52.33837657928263]
  Style transfer, given content and style references, remains challenging for Diffusion Transformers.<n>We propose Style-CCL, a Multi-Stage Curriculum Continual Learning framework that trains SC-DiT from semantic (easy) to texture (hard) styles.<n>Our framework achieves state-of-the-art performance in three core metrics: style similarity, content consistency, and aesthetic quality.
  arXiv  Detail & Related papers  (2026-06-04T13:02:04Z)
• Mind the Style Gap: Meta-Evaluation of Style and Attribute Transfer Metrics [38.04931225191056]
  This paper presents a large meta-evaluation of metrics for evaluating style and attribute transfer.<n>We find that meta-evaluation studies on existing datasets lead to misleading conclusions about the suitability of metrics for content preservation.<n>We introduce a new, challenging test set specifically designed for evaluating content preservation metrics for style transfer.
  arXiv  Detail & Related papers  (2025-02-20T20:16:34Z)
• Isolating authorship from content with semantic embeddings and contrastive learning [49.15148871877941]
  Authorship has entangled style and content inside.<n>We present a technique to use contrastive learning with additional hard negatives synthetically created using a semantic similarity model.<n>This disentanglement technique aims to distance the content embedding space from the style embedding space, leading to embeddings more informed by style.
  arXiv  Detail & Related papers  (2024-11-27T16:08:46Z)
• StyleDistance: Stronger Content-Independent Style Embeddings with Synthetic Parallel Examples [48.44036251656947]
  Style representations aim to embed texts with similar writing styles closely and texts with different styles far apart, regardless of content.<n>We introduce StyleDistance, a novel approach to training stronger content-independent style embeddings.
  arXiv  Detail & Related papers  (2024-10-16T17:25:25Z)
• Few-shot Image Generation via Style Adaptation and Content Preservation [60.08988307934977]
  We introduce an image translation module to GAN transferring, where the module teaches the generator to separate style and content. Our method consistently surpasses the state-of-the-art methods in few shot setting.
  arXiv  Detail & Related papers  (2023-11-30T01:16:53Z)
• $Z^*$: Zero-shot Style Transfer via Attention Rearrangement [27.185432348397693]
  This study shows that vanilla diffusion models can directly extract style information and seamlessly integrate the generative prior into the content image without retraining. We adopt dual denoising paths to represent content/style references in latent space and then guide the content image denoising process with style latent codes.
  arXiv  Detail & Related papers  (2023-11-25T11:03:43Z)
• ALADIN-NST: Self-supervised disentangled representation learning of artistic style through Neural Style Transfer [60.6863849241972]
  We learn a representation of visual artistic style more strongly disentangled from the semantic content depicted in an image. We show that strongly addressing the disentanglement of style and content leads to large gains in style-specific metrics.
  arXiv  Detail & Related papers  (2023-04-12T10:33:18Z)
• Enhancing Content Preservation in Text Style Transfer Using Reverse Attention and Conditional Layer Normalization [15.444996697848266]
  A common approach is to map a given sentence to content representation that is free of style, and the content representation is fed to a decoder with a target style. Previous methods in filtering style completely remove tokens with style at the token level, which incurs the loss of content information. We propose to enhance content preservation by implicitly removing the style information of each token with reverse attention, and thereby retain the content.
  arXiv  Detail & Related papers  (2021-08-01T12:54:46Z)
• Arbitrary Style Transfer via Multi-Adaptation Network [109.6765099732799]
  A desired style transfer, given a content image and referenced style painting, would render the content image with the color tone and vivid stroke patterns of the style painting. A new disentanglement loss function enables our network to extract main style patterns and exact content structures to adapt to various input images.
  arXiv  Detail & Related papers  (2020-05-27T08:00:22Z)
• Adversarial Disentanglement with Grouped Observations [0.0]
  We consider the disentanglement of the representations of the relevant attributes of the data (content) from all other factors of variations (style) This work supplements these algorithms with a method that eliminates the content information in the style representations. Experimental results and comparisons on image datasets show that the resulting method can efficiently separate the content and style related attributes and generalizes to unseen data.
  arXiv  Detail & Related papers  (2020-01-14T13:21:25Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.