mlsanityData to 14 Oct 2026

LatentDiff: Scaling Semantic Dataset Comparison to Millions of Images

arXiv 2605.00899 ↗First posted 28 Apr 2026cs.CV cs.LGcited by 0

Not on the map. Why?

Abstract (arXiv)

We present LatentDiff, a scalable framework for semantic dataset comparison that operates directly in the latent space of pretrained vision encoders. By combining sparse autoencoder-based divergence testing with density ratio estimation, LatentDiff identifies interpretable semantic differences between datasets at a fraction of the computational cost of caption-based alternatives. We also introduce Noisy-Diff, a benchmark capturing realistic sparse distribution shifts that cause existing methods to struggle. Experiments demonstrate that LatentDiff achieves superior accuracy while remaining robust to settings where an extremely small fraction of images (from 5% to <1% ) differ semantically.

Key facts. No citing papers in the map yet (data through 14 Oct 2026). Builds on 27 of its 41 references that are in the map. First posted on arXiv 28 Apr 2026.

Builds on

27 of 41 references.

References · 27

  1. 2021Learning Transferable Visual Models From Natural Language Supervisioncited by 33,744
  2. 2014Microsoft COCO: Common Objects in Contextcited by 17,608
  3. 2019Sentence-BERT: Sentence Embeddings using Siamese BERT-Networkscited by 9,170
  4. 2019Benchmarking Neural Network Robustness to Common Corruptions and Perturbationscited by 2,877
  5. 2020The Many Faces of Robustness: A Critical Analysis of Out-of-Distribution Generalizationcited by 1,850
  6. 2019Do ImageNet Classifiers Generalize to ImageNet?cited by 1,515
  7. 2023Sparse Autoencoders Find Highly Interpretable Features in Language Modelscited by 1,420
  8. 2019Natural Adversarial Examplescited by 1,408
  9. 2020WILDS: A Benchmark of in-the-Wild Distribution Shiftscited by 1,375
  10. 2022DINO: DETR with Improved DeNoising Anchor Boxes for End-to-End Object Detectioncited by 1,319
  11. 2024Scaling and evaluating sparse autoencoderscited by 553
  12. 2020Measuring Robustness to Natural Distribution Shifts in Image Classificationcited by 454
15 more
  1. 2018Failing Loudly: An Empirical Study of Methods for Detecting Dataset Shiftcited by 243
  2. 2022Domino: Discovering Systematic Errors with Cross-Modal Embeddingscited by 156
  3. 2022MetaShift: A Dataset of Datasets for Evaluating Contextual Distribution Shifts and Training Conflictscited by 81
  4. 2025Sparse Autoencoders Learn Monosemantic Features in Vision-Language Modelscited by 71
  5. 2023Goal Driven Discovery of Distributional Differences via Language Descriptionscited by 62
  6. 2025Interpreting CLIP with Hierarchical Sparse Autoencoderscited by 60
  7. 2022Describing Differences between Text Distributions with Natural Languagecited by 54
  8. 2023Describing Differences in Image Sets with Natural Languagecited by 53
  9. 2025Steering CLIP's vision transformer with sparse autoencoderscited by 38
  10. 2022Towards Explaining Distribution Shiftscited by 26
  11. 2023FACTS: First Amplify Correlations and Then Slice to Discover Biascited by 26
  12. 2022GSCLIP : A Framework for Explaining Distribution Shifts in Natural Languagecited by 14
  13. 2023VLSlice: Interactive Vision-and-Language Slice Discoverycited by 7
  14. 2024LADDER: Language-Driven Slice Discovery and Error Rectification in Vision Classifierscited by 7
  15. 2023Cross-GAN Auditing: Unsupervised Identification of Attribute Level Similarities and Differences between Pretrained Generative Modelscited by 2

Built on by

Not cited yet.

Similar, not cited

Close in wording, no citation either way

  1. 2025OmniDiff: A Comprehensive Benchmark for Fine-grained Image Difference Captioning
  2. 2024OneDiff: A Generalist Model for Image Difference Captioning
  3. 2025Both Semantics and Reconstruction Matter: Making Representation Encoders Ready for Text-to-Image Generation and Editing
  4. 2026DiffCap-Bench: A Comprehensive, Challenging, Robust Benchmark for Image Difference Captioning
  5. 2026VDiff-Bench: A Challenging Benchmark for Fine-Grained Image Difference Identification
  6. 2026Boosting Latent Diffusion Models via Disentangled Representation Alignment
  7. 2025Semantics Lead the Way: Harmonizing Semantic and Texture Modeling with Asynchronous Latent Diffusion
  8. 2026DataComp-VLM: Improved Open Datasets for Vision-Language Models