mlsanityData to 14 Oct 2026

DEAL: Disentangle and Localize Concept-level Explanations for VLMs

arXiv 2407.14412 ↗First posted 19 Jul 2024cs.CV cs.AI cs.LGcited by 14

On the map: Multimodal LLMs / vision-language models → VLM benchmarks and visual reasoning

Abstract (arXiv)

Large pre-trained Vision-Language Models (VLMs) have become ubiquitous foundational components of other models and downstream tasks. Although powerful, our empirical results reveal that such models might not be able to identify fine-grained concepts. Specifically, the explanations of VLMs with respect to fine-grained concepts are entangled and mislocalized. To address this issue, we propose to DisEntAngle and Localize (DEAL) the concept-level explanations for VLMs without human annotations. The key idea is encouraging the concept-level explanations to be distinct while maintaining consistency with category-level explanations. We conduct extensive experiments and ablation studies on a wide range of benchmark datasets and vision-language models. Our empirical results demonstrate that the proposed method significantly improves the concept-level explanations of the model in terms of disentanglability and localizability. Surprisingly, the improved explainability alleviates the model's reliance on spurious correlations, which further benefits the prediction accuracy.

Key facts. Cited by 14 papers in the map (data through 14 Oct 2026). Placed in VLM benchmarks and visual reasoning, one of 16 directions in Multimodal LLMs / vision-language models. Builds on 56 of its 70 references that are in the map. First posted on arXiv 19 Jul 2024.

Builds on

56 of 70 references.

Closest ancestors · 3

Same direction, newest first

  1. 2022Perceptual Grouping in Contrastive Vision-Language Models18 Oct
  2. 2022BLIP: Bootstrapping Language-Image Pre-training for Unified Vision-Language Understanding and Generation28 Jan
  3. 2021Multi-Grained Vision Language Pre-Training: Aligning Texts with Visual Concepts16 Nov

Other references · 53

  1. 2014Adam: A Method for Stochastic Optimizationcited by 37,240
  2. 2021Learning Transferable Visual Models From Natural Language Supervisioncited by 33,744
  3. 2020Language Models are Few-Shot Learnerscited by 32,419
  4. 2014Microsoft COCO: Common Objects in Contextcited by 17,608
  5. 2022Chain-of-Thought Prompting Elicits Reasoning in Large Language Modelscited by 12,459
  6. 2023Visual Instruction Tuningcited by 7,869
  7. 2018Representation Learning with Contrastive Predictive Codingcited by 7,777
  8. 2019Momentum Contrast for Unsupervised Visual Representation Learningcited by 7,513
  9. 2016Grad-CAM: Visual Explanations from Deep Networks via Gradient-based Localizationcited by 6,733
  10. 2017A Unified Approach to Interpreting Model Predictionscited by 6,555
  11. 2016"Why Should I Trust You?": Explaining the Predictions of Any Classifiercited by 5,070
  12. 2021On the Opportunities and Risks of Foundation Modelscited by 4,033
38 more
  1. 2021Convex optimizationcited by 2,515
  2. 2013Deep Inside Convolutional Networks: Visualising Image Classification Models and Saliency Mapscited by 2,371
  3. 2019ViLBERT: Pretraining Task-Agnostic Visiolinguistic Representations for Vision-and-Language Taskscited by 2,222
  4. 2019Invariant Risk Minimizationcited by 1,854
  5. 2017EuroSAT: A Novel Dataset and Deep Learning Benchmark for Land Use and Land Cover Classificationcited by 1,496
  6. 2017Interpretability Beyond Feature Attribution: Quantitative Testing with Concept Activation Vectors (TCAV)cited by 1,319
  7. 2021Grounded Language-Image Pre-trainingcited by 1,156
  8. 2020Concept Bottleneck Modelscited by 1,032
  9. 2018RISE: Randomized Input Sampling for Explanation of Black-box Modelscited by 838
  10. 2018This Looks Like That: Deep Learning for Interpretable Image Recognitioncited by 737
  11. 2021FLAVA: A Foundational Language And Vision Alignment Modelcited by 649
  12. 2023Capabilities of GPT-4 on Medical Challenge Problemscited by 638
  13. 2022When and why vision-language models behave like bags-of-words, and what to do about it?cited by 610
  14. 2022Winoground: Probing Vision and Language Models for Visio-Linguistic Compositionalitycited by 529
  15. 2021Supervision Exists Everywhere: A Data Efficient Contrastive Language-Image Pre-training Paradigmcited by 405
  16. 2022Visual Classification via Description from Large Language Modelscited by 336
  17. 2021How Much Can CLIP Benefit Vision-and-Language Tasks?cited by 321
  18. 2021Generic Attention-model Explainability for Interpreting Bi-Modal and Encoder-Decoder Transformerscited by 320
  19. 2022Language in a Bottle: Language Model Guided Concept Bottlenecks for Interpretable Image Classificationcited by 312
  20. 2020Self-supervised Equivariant Attention Mechanism for Weakly Supervised Semantic Segmentationcited by 300
  21. 2020Learning to Learn Single Domain Generalizationcited by 296
  22. 2021SMIL: Multimodal Learning with Severely Missing Modalitycited by 245
  23. 2020Evaluating and Aggregating Feature-based Model Explanationscited by 188
  24. 2022CREPE: Can Vision-Language Foundation Models Reason Compositionally?cited by 181
  25. 2019Explainable Machine Learning for Scientific Insights and Discoveriescited by 166
  26. 2022Finetune like you pretrain: Improved finetuning of zero-shot vision modelscited by 150
  27. 2019Interpretations are useful: penalizing explanations to align neural networks with prior knowledgecited by 145
  28. 2020Restricting the Flow: Information Bottlenecks for Attributioncited by 110
  29. 2021PartImageNet: A Large, High-Quality Dataset of Partscited by 107
  30. 2022PyramidCLIP: Hierarchical Feature Alignment for Vision-language Model Pretrainingcited by 104
  31. 2020Right for the Right Concept: Revising Neuro-Symbolic Concepts by Interacting with their Explanationscited by 94
  32. 2022Medical Image Understanding with Pretrained Vision Language Models: A Comprehensive Studycited by 66
  33. 2019Robust Attribution Regularizationcited by 47
  34. 2022Attention Consistency on Visual Corruptions for Single-Source Domain Generalizationcited by 37
  35. 2020Seeing past words: Testing the cross-modal capabilities of pretrained V&L models on counting taskscited by 34
  36. 2018TED: Teaching AI to Explain its Decisionscited by 33
  37. 2021Consistent Explanations by Contrastive Learningcited by 18
  38. 2023Are Data-driven Explanations Robust against Out-of-distribution Data?cited by 14

50 most cited of 53 All 53 (JSON)

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2024: 4; 2025: 5; 2026: 5.

14 papers.

Sort:

By area:

  1. 2024DCBM: Data-Efficient Visual Concept Bottleneck Modelscited by 16
  2. 2025Discovering Fine-Grained Visual-Concept Relations by Disentangled Optimal Transport Concept Bottleneck Modelscited by 12
  3. 2025Interpretable Failure Detection with Human-Level Conceptscited by 7
  4. 2024Beyond Accuracy: Ensuring Correct Predictions With Correct Rationalescited by 6
  5. 2024SeafloorAI: A Large-scale Vision-Language Dataset for Seafloor Geological Surveycited by 6
  6. 2026Inside the Visual Mind: Neuroscience-Motivated Concept Circuits for Interpreting and Steering Vision Transformerscited by 3
  7. 2024Beyond Accuracy: On the Effects of Fine-tuning Towards Vision-Language Model's Prediction Rationalitycited by 2
  8. 2025Causality-guided Prompt Learning for Vision-language Models via Visual Granulationcited by 2
  9. 2025Enhancing Pre-trained Representation Classifiability can Boost its Interpretabilitycited by 2
  10. 2025BiPVL-Seg: Bidirectional Progressive Vision-Language Fusion with Global-Local Alignment for Medical Image Segmentationcited by 1
  11. 2026Select, Hypothesize and Verify: Towards Verified Neuron Concept Interpretationcited by 0
  12. 2026MEDIC-AD: Towards Medical Vision-Language Model's Clinical Intelligencecited by 0
2 more
  1. 2026MedPlex: Deep Vision-Language Co-Adaptation for Clinically Grounded Medical Segmentationcited by 0
  2. 2026SE-MoLoRA: Shared-Expert LoRA Adapters for Domain-Specific Photographic Assessmentcited by 0

Similar, not cited

Close in wording, no citation either way

  1. 2025V-SEAM: Visual Semantic Editing and Attention Modulating for Causal Interpretability of Vision-Language Models
  2. 2026Explaining is Harder Than Predicting Alone: Evaluating Concept-based Explanations of MLLMs as ICL Visual Classifiers
  3. 2026Investigating Relational Reasoning in VLMs
  4. 2025Believing without Seeing: Quality Scores for Contextualizing Vision-Language Model Explanations
  5. 2026When VLMs Destroy What They Know: Spatial Representations Fall Below Lexical Priors
  6. 2023Synthesize, Diagnose, and Optimize: Towards Fine-Grained Vision-Language Understanding
  7. 2024If CLIP Could Talk: Understanding Vision-Language Model Representations Through Their Preferred Concept Descriptions
  8. 2026Logit Lens Supervision for Patch-Level Explanations in Vision-Language Models