Shadowgraph
TrendingBuzzResearch
Chrome extension
Sign in

kkzhang95/Awesome_Concept_Bottleneck_Models

A comprehensive survey of Concept Bottleneck Models (CBM)

73

50 commits

updated Jan 20, 2026

See the code

README

Awesome Concept Bottleneck Models

This repository is the official companion to our survey paper: Concept Bottleneck Models for Explainable Decision Making: A Survey of Progress, Taxonomy, and Future Directions.

We aim to systematically organize the rapidly evolving landscape of Concept Bottleneck Models (CBMs), offering a comprehensive collection of core literature, code implementations, and datasets. The paper can be found here: Concept Bottleneck Models for Explainable Decision Making: A Survey of Progress, Taxonomy and Future Directions

What is Concept Bottleneck Models (CBMs)?

Concept Bottleneck Models (CBMs) represent a shift from opaque "black-box" neural networks to interpretable, human-aligned systems. By explicitly decomposing the prediction process into two stages—mapping inputs to human-understandable concepts ($x \rightarrow c$) and then concepts to targets ($c \rightarrow y$)—CBMs offer semantic explanations and enable test-time intervention.

🧩 Taxonomy & Framework

We propose a comprehensive taxonomy to organize CBM research along four primary dimensions.

fig_overview Figure 1: Overview of the CBM pipeline. The figure illustrates how explicit concepts are acquired, integrated into decision making, intervened upon, and evaluated to produce interpretable predictions.

1. Concept Acquisition

How are the interpretable concepts defined and learned?

  • Manual Annotation: Expert-defined concepts (e.g., CUB, medical attributes).
  • Dictionary-Based: Visually grounded concepts filtered from large vocabularies.
  • LLM/VLM-Guided: Leveraging foundation models to generate hierarchical or descriptive semantics.
  • Visual Prototypes: Discovering concepts directly from image regions or parts.

2. Concept-based Decision Making

How does the bottleneck mediate information flow?

  • Bottleneck Design: From strict "hard" bottlenecks to "soft," "hybrid," or "residual" designs that balance interpretability and accuracy.
  • Supervision: Ranging from strong supervision (full concept labels) to weak or unsupervised learning.
  • Model Adaptation: Methods enabling post-hoc editing or updating of concepts without full retraining.

3. Concept Intervention

How can humans interact with the model?

  • Mechanisms: Modifying scalar values, spatial attention, or using natural language feedback.
  • Correlations: Handling dependencies between concepts to propagate corrections effectively.
  • Localization: Identifying which concepts allow for the most effective intervention.

4. Concept Evaluation

How do we measure success beyond accuracy?

  • Faithfulness: Do concepts align with true semantic meanings?
  • Sparsity: Is the concept set minimal and cognitively manageable?
  • Intervenability: How effectively does the model respond to human corrections?

📖 Paper List

2025

PublicationPaper TitleCode/Project
PMLR 2025Improving ARDS Diagnosis Through Context-Aware Concept Bottleneck ModelsCode
NeurIPS 2025Sample-efficient Learning of Concepts with Theoretical Guarantees: from Data to Concepts without InterventionsCode
NeurIPS 2025DermaCon-IN: A Multiconcept-Annotated Dermatological Image Dataset of Indian Skin Disorders for Clinical AI ResearchCode
NeurIPS 2025V-CECE: Visual Counterfactual Explanations via Conceptual Edits-
NeurIPS 2025Shortcuts and Identifiability in Concept-based Models from a Neuro-Symbolic Lens-
NeurIPS 2025Object Centric Concept BottlenecksCode
NeurIPS 2025Deferring Concept Bottleneck Models: Learning to Defer Interventions to Inaccurate Experts-
NeurIPS 2025Bayesian concept bottleneck models with llm priorsCode
NeurIPS 2025Debugging Concept Bottleneck Models through Removal and RetrainingCode
NeurIPS 2025Multimodal Concept Bottleneck Models-
NeurIPS 2025An Analysis of Concept Bottleneck Models: Measuring, Understanding, and Mitigating the Impact of Noisy Annotations-
NeurIPS 2025Causally Reliable Concept Bottleneck ModelsCode
ICCV 2025Towards Safer and Understandable Driver Intention PredictionCode
ICCV 2025Intervening in Black Box: Concept Bottleneck Model for Enhancing Human Neural Network Mutual UnderstandingCode
ICCV 2025Semi-supervised Concept Bottleneck ModelsCode
2025Interpretable Gastric Cancer Survival Prediction via Concept Bottleneck Model-
2025Automated Genomic Interpretation via Concept Bottleneck Models for Medical Robotics-
2025Graph Integrated Multimodal Concept Bottleneck Model-
2025Zero-shot Concept Bottleneck ModelsCode
2025Chat-CBM: Towards Interactive Concept Bottleneck Models with Frozen Large Language Models-
2025Graph Concept Bottleneck Models-
2025Towards Reasonable Concept Bottleneck Models-
2025There Was Never a Bottleneck in Concept Bottleneck ModelsCode
ICML 2025Addressing Concept Mislabeling in Concept Bottleneck Models Through Preference OptimizationCode
ICML 2025Editable Concept Bottleneck ModelsCode
ICLR 2025Adaptive Test-Time Intervention for Concept Bottleneck ModelsCode
ICLR 2025Preference optimization for concept bottleneck models-
ICLR 2025Interpretable Vision-Language Survival Analysis with Ordinal Inductive Bias for Computational PathologyCode
ICLR 2025Concept Bottleneck Language Models For Protein Design-
ICLR 2025Concept Bottleneck Large Language ModelsCode
ICLR 2025Counterfactual Concept Bottleneck ModelsCode
AAAI 2025Explanation Bottleneck ModelsCode
AAAI 2025Interpretable Failure Detection with Human-Level ConceptsCode
AAAI 2025V2C-CBM: Building Concept Bottlenecks with Vision-to-Concept TokenizerCode
MICCAI 2025RadAlign: Advancing Radiology Report Generation with Vision-Language Concept AlignmentCode
MICCAI 2025Learning Concept-Driven Logical Rules for Interpretable and Generalizable Medical Image ClassificationCode
MICCAI 2025Explain Any Pathological Concept: Discovering Hierarchical Explanations for Pathology Foundation Models-
MICCAI 2025CytoSAE: Interpretable Cell Embeddings for HematologyCode
MICCAI 2025SurgX: Neuron-Concept Association for Explainable Surgical Phase RecognitionCode
MICCAI 2025Concept-induced Graph Perception Model for Interpretable Diagnosis-
MICCAI 2025CoPA: Hierarchical Concept Prompting and Aggregating Network for Explainable DiagnosisCode
MICCAI 2025MCA-RG: Enhancing LLMs with Medical Concept Alignment for Radiology Report Generation-
CVPR 2025Language Guided Concept Bottleneck Models for Interpretable Continual LearningCode
CVPR 2025Concept Replacer: Replacing Sensitive Concepts in Diffusion Models via Precision LocalizationCode
CVPR 2025Attribute-formed Class-specific Concept Space: Endowing Language Bottleneck Model with Better Interpretability and ScalabilityCode
CVPR 2025Explaining Domain Shifts in Language: Concept erasing for Interpretable Image ClassificationCode
CVPR 2025CoE: Chain-of-Explanation via Automatic Visual Concept Circuit Description and Polysemanticity QuantificationCode
CVPR 2025Discovering Fine-Grained Visual-Concept Relations by Disentangled Optimal Transport Concept Bottleneck Models-
CVPR 2025Precise Fast and Low-cost Concept Erasure in_Value_SpaceCode
CVPR 2025Efficient Fine-Tuning and Concept Suppression for Pruned Diffusion ModelsCode
CVPR 2025Self-Evolving Visual Concept Library using Vision-Language CriticsCode
CVPR 2025Concept Lancet: Image Editing with Compositional Representation Transplant-
CVPR 2025Hybrid Concept Bottleneck ModelsCode
CVPR 2025Localized Concept Erasure for Text-to-Image Diffusion Models Using Training-Free Gated Low-Rank AdaptationCode
CVPR 2025ESC: Erasing Space Concept for Knowledge DeletionCode
CVPR 2025Interpretable Generative Models through Post-hoc Concept BottlenecksCode
CVPR 2025Discovering Hidden Visual Concepts Beyond Linguistic Input in Infant LearningCode
CVPR 2025MC^2: Multi-concept Guidance for Customized Multi-concept GenerationCode
CVPR 2025Interactive Medical Image Analysis with Concept-based Similarity ReasoningCode
CVPR 2025Towards Human-Understandable Multi-Dimensional Concept DiscoveryCode
CVPR 2025ICE: Intrinsic Concept Extraction from a Single Image via Diffusion ModelsCode
CVPR 2025Show and Tell: Visually Explainable Deep Neural Nets via Spatially-Aware Concept Bottleneck Modelscode
TMLR 2025Selective Concept Bottleneck Models Without Predefined ConceptsCode
2025ConceptCLIP: Towards Trustworthy Medical AI Via Concept-Enhanced Contrastive Langauge-Image Pre-trainingCode

2024

PublicationPaper TitleCode/Project
CVPR 2024Incremental Residual Concept Bottleneck ModelsCode
2024Concept-Attention Whitening for Interpretable Skin Lesion Diagnosis-
MICCAI 2024Aligning human knowledge with visual concepts towards explainable medical image classificationCode
MICCAI 2024Evidential concept embedding models: Towards reliable concept explanations for skin disease diagnosisCode
MICCAI 2024Learning a Clinically-Relevant Concept Bottleneck for Lesion Detection in Breast UltrasoundCode
2024Mask-Free Neuron Concept Annotation for Interpreting Neural Networks in Medical DomainCode
ECCV 2024Discover-then-Name: Task-Agnostic Concept Bottlenecks via Automated Concept DiscoveryCode
NeurIPS 2024VLG-CBM: Training Concept Bottleneck Models with Vision-Language GuidanceCode
LNCS 2024AdaCBM: an Adaptive Concept Bottleneck Model for Explainable and Accurate DiagnosisCode
2024Explain Via Any Concept: Concept Bottleneck Model with Open Vocabulary Concepts-
2024Self-eXplainable AI for Medical Image Analysis: A Survey and New Outlooks-
2024Concept Complement Bottleneck Model for Interpretable Medical Image Diagnosis-
NeurIPS 2024Towards Multi-dimensional Explanation Alignment for Medical Classification-
2024Explainable and interpretable multimodal large language models: A comprehensive survey-
AAAI 2024Sparsity-guided holistic explanation for llms with interpretable inference-time interventionCode
NeurIPS 2024Beyond Concept Bottleneck Models: How to Make Black Boxes Intervenable?Code
TMLR 2024CLIP-QDA: an Explainable Concept Bottleneck Model-
ICLR 2024Concept Bottleneck Generative ModelsCode
ICLR 2024Energy-Based Concept Bottleneck Models: Unifying Prediction, Concept Intervention, and Probabilistic InterpretationsCode
ICLR 2024Faithful Vision-Language Interpretation Via Concept Bottleneck ModelsCode
J-BHI 2024Guest Editorial: Trustworthy Machine Learning for Health Informatics-
2024Improving Concept Alignment in Vision-Language Concept Bottleneck ModelsCode
MICCAI 2024Integrating Clinical Knowledge into Concept Bottleneck ModelsCode
Medical Image Analysis 2024Interpretable and Intervenable Ultrasonography-Based Machine Learning Models for Pediatric AppendicitisCode
PAKDD 2024Interpreting Pretrained Language Models Via Concept BottlenecksCode
2024LVLM-Interpret: An Interpretability Tool for Large Vision-Language ModelsCode
AAAI 2024On the Concept Trustworthiness in Concept Bottleneck ModelsCode
NeurIPS 2024Stochastic Concept Bottleneck ModelsCode
2024XCoOp: Explainable Prompt Learning for Computer-Aided Diagnosis via Concept-guided Context Optimization-
ICLR 2024Concept Bottleneck Generative ModelsCode
Sci. Rep 2024Pseudo-class Part Prototype Networks for Interpretable Breast Cancer ClassificationCode
TMLR 2024Reproducibility Study of "LICO: Explainable Models with Language-Image Consistency"Code
2024Post-hoc Part-prototype Networks-
2024Embracing Unimodal Aleatoric Uncertainty for Robust Multimodal Fusion-
TMLR 2024[Re].on the Reproducibility of Post-Hoc Concept Bottleneck ModelsCode
NeurIPS 2024A Theoretical Design of Concept Sets: Improving the Predictability of Concept Bottleneck Models
NeurIPS 2024A Concept-Based Explainability Framework for Large Multimodal ModelsCode
NeurIPS 2024Classifier Clustering and Feature Alignment for Federated Learning under Distributed Concept DriftCode
NeurIPS 2024ConceptMix: A Compositional Image Generation Benchmark with Controllable DifficultyCode
NeurIPS 2024Do LLMs Dream of Elephants (when Told Not To)? Latent Concept Association and Associative Memory in Transformers-
NeurIPS 2024FinCon: A Synthesized LLM Multi-Agent System with Conceptual Verbal Reinforcement for Enhanced Financial Decision MakingCode
NeurIPS 2024Free Lunch in Pathology Foundation Model: Task-specific Model Adaptation with Concept-Guided Feature EnhancementCode
NeurIPS 2024From Causal to Concept-Based Representation Learning-
NeurIPS 2024Interpretable Concept Bottlenecks to Align Reinforcement Learning AgentsCode
NeurIPS 2024Interpretable Concept-Based Memory ReasoningCode
NeurIPS 2024Interpreting CLIP with Sparse Linear Concept Embeddings (Splice)Code
NeurIPS 2024Coarse-to-fine concept bottleneck modelsCode
NeurIPS 2024Learning Discrete Concepts in Latent Hierarchical Models-
2024LG-CAV: Train Any Concept Activation Vector with Language Guidance-
2024Neural Concept BinderCode
NeurIPS 2024No "Zero-Shot" Without Exponential Data: Pretraining Concept Frequency Determines Multimodal Model PerformanceCode
NeurIPS 2024PaCE: Parsimonious Concept Engineering for Large Language ModelsCode
NeurIPS 2024Relational Concept Bottleneck ModelsCode
NeurIPS 2024Uncovering Safety Risks of Large Language Models Through Concept Activation VectorCode

2023

PublicationPaper TitleCode/Project
AAAI 2023Interactive Concept Bottleneck Models-
ICLR 2023Post-hoc Concept Bottleneck ModelsCode
ICLR 2023LABEL-FREE CONCEPT BOTTLENECK MODELSCode
CVPR 2023Language in a Bottle: Language Model Guided Concept Bottlenecks for Interpretable Image ClassificationCode
2023STAIR: Learning Sparse Text and Image Representation in Grounded Tokens-
2023A ChatGPT Aided Explainable Framework for Zero-Shot Medical Image Diagnosis-
ICML 2023A Closer Look at the Intervention Procedure of Concept Bottleneck ModelsCode
MICCAI 2023Concept Bottleneck with Visual Concept Filtering for Explainable Medical Image Classification-
EMNLP 2023Cross-Modal Conceptualization in Bottleneck ModelsCode
NeurIPs 2023Do Concept Bottleneck Models Respect LocalitiesCode
CVPR 2023Learning bottleneck concepts in image classificationCode
ICML 2023Probabilistic Concept Bottleneck ModelsCode
2023Robust and interpretable medical image classifiers via concept bottleneck models-
CVPR 2023Towards Trustable Skin Cancer Diagnosis via Rewriting Model's Decision-

2022

PublicationPaper TitleCode/Project
ICCV 2022Explaining in Style: Training a GAN to Explain a Classifier in StyleSpace-
ICLR 2023CLIP-Dissect: Automatic Description of Neuron Representations in Deep Vision NetworksCode
NeurIPS 2022Concept Embedding ModelsCode
NeurIPS 2022Addressing Leakage in Concept Bottleneck ModelsCode

2021

PublicationPaper TitleCode/Project
ICML 2021Meaningfully Debugging Model Mistakes Using Conceptual Counterfactual ExplanationsCode
NML 2021A case-based interpretable deep learning model for classification of mass lesions in digital mammographyCode
2021Partially Interpretable Estimators (PIE): Black-Box-Refined Interpretable Machine Learning-

2020

PublicationPaper TitleCode/Project
ICML 2020Concept bottleneck modelsCode
NML 2020Concept whitening for interpretable image recognitionCode

2018

PublicationPaper TitleCode/Project
ICML 2018Interpretability Beyond Feature Attribution: Quantitative Testing with Concept Activation Vectors (TCAV)Code
J-BHI 2018Seven-Point Checklist and Skin Lesion Classification Using Multitask Multimodal Neural NetsCode

2017

PublicationPaper TitleCode/Project
AAAI 2017ConceptNet 5.5: An Open Multilingual Graph of General KnowledgeCode

📊 Benchmark

Representative benchmarks for evaluating CBMs. image

📄 Paper

If you find this work helpful, please consider citing:

@article{Wang2026CBMSurvey,
  title   = {Concept Bottleneck Models for Explainable Decision Making: A Survey of Progress, Taxonomy, and Future Directions},
  author  = {Wang, Chunjiang and Li, Fan and Hu, Wenbo and Yan, Rui and Zhang, Kun and Zhou, Shaohua Kevin},
  journal = {ResearchGate Preprint},
  year    = {2026},
  doi     = {10.13140/RG.2.2.30356.16002},
  url     = {https://www.researchgate.net/publication/399898851_Concept_Bottleneck_Models_for_Explainable_Decision_Making_A_Survey_of_Progress_Taxonomy_and_Future_Directions}
}

Similar repos

riverback/Awesome-Concept-Bottleneck-Models

A list of papers about concept bottleneck models (CBMs)

34

rushrukh/awesome-explainable-ai

A repository for summaries of recent explainable AI/Interpretable ML approaches

91

wangyongjie-ntu/Awesome-explainable-AI

A collection of research materials on explainable AI/ML

1,662

AI-in-Transportation-Lab/awesome-mechanistic-interpretability

A carefully curated collection of high-quality libraries, projects, tutorials, research papers, and…

166

adagorgun/awesome-generative-explainability

A collection of research materials on explainable generative models

26

anguyen8/XAI-papers

621

cooperleong00/Awesome-LLM-Interpretability

A curated list of LLM Interpretability related material - Tutorial, Library, Survey, Paper, Blog,…

308

ZFancy/awesome-activation-engineering

A curated list of resources for activation engineering

140

kkzhang95/Awesome_Concept_Bottleneck_Models

A comprehensive survey of Concept Bottleneck Models (CBM)

73

50 commits

updated Jan 20, 2026

See the code

README

Awesome Concept Bottleneck Models

This repository is the official companion to our survey paper: Concept Bottleneck Models for Explainable Decision Making: A Survey of Progress, Taxonomy, and Future Directions.

We aim to systematically organize the rapidly evolving landscape of Concept Bottleneck Models (CBMs), offering a comprehensive collection of core literature, code implementations, and datasets. The paper can be found here: Concept Bottleneck Models for Explainable Decision Making: A Survey of Progress, Taxonomy and Future Directions

What is Concept Bottleneck Models (CBMs)?

Concept Bottleneck Models (CBMs) represent a shift from opaque "black-box" neural networks to interpretable, human-aligned systems. By explicitly decomposing the prediction process into two stages—mapping inputs to human-understandable concepts ($x \rightarrow c$) and then concepts to targets ($c \rightarrow y$)—CBMs offer semantic explanations and enable test-time intervention.

🧩 Taxonomy & Framework

We propose a comprehensive taxonomy to organize CBM research along four primary dimensions.

fig_overview Figure 1: Overview of the CBM pipeline. The figure illustrates how explicit concepts are acquired, integrated into decision making, intervened upon, and evaluated to produce interpretable predictions.

1. Concept Acquisition

How are the interpretable concepts defined and learned?

  • Manual Annotation: Expert-defined concepts (e.g., CUB, medical attributes).
  • Dictionary-Based: Visually grounded concepts filtered from large vocabularies.
  • LLM/VLM-Guided: Leveraging foundation models to generate hierarchical or descriptive semantics.
  • Visual Prototypes: Discovering concepts directly from image regions or parts.

2. Concept-based Decision Making

How does the bottleneck mediate information flow?

  • Bottleneck Design: From strict "hard" bottlenecks to "soft," "hybrid," or "residual" designs that balance interpretability and accuracy.
  • Supervision: Ranging from strong supervision (full concept labels) to weak or unsupervised learning.
  • Model Adaptation: Methods enabling post-hoc editing or updating of concepts without full retraining.

3. Concept Intervention

How can humans interact with the model?

  • Mechanisms: Modifying scalar values, spatial attention, or using natural language feedback.
  • Correlations: Handling dependencies between concepts to propagate corrections effectively.
  • Localization: Identifying which concepts allow for the most effective intervention.

4. Concept Evaluation

How do we measure success beyond accuracy?

  • Faithfulness: Do concepts align with true semantic meanings?
  • Sparsity: Is the concept set minimal and cognitively manageable?
  • Intervenability: How effectively does the model respond to human corrections?

📖 Paper List

2025

PublicationPaper TitleCode/Project
PMLR 2025Improving ARDS Diagnosis Through Context-Aware Concept Bottleneck ModelsCode
NeurIPS 2025Sample-efficient Learning of Concepts with Theoretical Guarantees: from Data to Concepts without InterventionsCode
NeurIPS 2025DermaCon-IN: A Multiconcept-Annotated Dermatological Image Dataset of Indian Skin Disorders for Clinical AI ResearchCode
NeurIPS 2025V-CECE: Visual Counterfactual Explanations via Conceptual Edits-
NeurIPS 2025Shortcuts and Identifiability in Concept-based Models from a Neuro-Symbolic Lens-
NeurIPS 2025Object Centric Concept BottlenecksCode
NeurIPS 2025Deferring Concept Bottleneck Models: Learning to Defer Interventions to Inaccurate Experts-
NeurIPS 2025Bayesian concept bottleneck models with llm priorsCode
NeurIPS 2025Debugging Concept Bottleneck Models through Removal and RetrainingCode
NeurIPS 2025Multimodal Concept Bottleneck Models-
NeurIPS 2025An Analysis of Concept Bottleneck Models: Measuring, Understanding, and Mitigating the Impact of Noisy Annotations-
NeurIPS 2025Causally Reliable Concept Bottleneck ModelsCode
ICCV 2025Towards Safer and Understandable Driver Intention PredictionCode
ICCV 2025Intervening in Black Box: Concept Bottleneck Model for Enhancing Human Neural Network Mutual UnderstandingCode
ICCV 2025Semi-supervised Concept Bottleneck ModelsCode
2025Interpretable Gastric Cancer Survival Prediction via Concept Bottleneck Model-
2025Automated Genomic Interpretation via Concept Bottleneck Models for Medical Robotics-
2025Graph Integrated Multimodal Concept Bottleneck Model-
2025Zero-shot Concept Bottleneck ModelsCode
2025Chat-CBM: Towards Interactive Concept Bottleneck Models with Frozen Large Language Models-
2025Graph Concept Bottleneck Models-
2025Towards Reasonable Concept Bottleneck Models-
2025There Was Never a Bottleneck in Concept Bottleneck ModelsCode
ICML 2025Addressing Concept Mislabeling in Concept Bottleneck Models Through Preference OptimizationCode
ICML 2025Editable Concept Bottleneck ModelsCode
ICLR 2025Adaptive Test-Time Intervention for Concept Bottleneck ModelsCode
ICLR 2025Preference optimization for concept bottleneck models-
ICLR 2025Interpretable Vision-Language Survival Analysis with Ordinal Inductive Bias for Computational PathologyCode
ICLR 2025Concept Bottleneck Language Models For Protein Design-
ICLR 2025Concept Bottleneck Large Language ModelsCode
ICLR 2025Counterfactual Concept Bottleneck ModelsCode
AAAI 2025Explanation Bottleneck ModelsCode
AAAI 2025Interpretable Failure Detection with Human-Level ConceptsCode
AAAI 2025V2C-CBM: Building Concept Bottlenecks with Vision-to-Concept TokenizerCode
MICCAI 2025RadAlign: Advancing Radiology Report Generation with Vision-Language Concept AlignmentCode
MICCAI 2025Learning Concept-Driven Logical Rules for Interpretable and Generalizable Medical Image ClassificationCode
MICCAI 2025Explain Any Pathological Concept: Discovering Hierarchical Explanations for Pathology Foundation Models-
MICCAI 2025CytoSAE: Interpretable Cell Embeddings for HematologyCode
MICCAI 2025SurgX: Neuron-Concept Association for Explainable Surgical Phase RecognitionCode
MICCAI 2025Concept-induced Graph Perception Model for Interpretable Diagnosis-
MICCAI 2025CoPA: Hierarchical Concept Prompting and Aggregating Network for Explainable DiagnosisCode
MICCAI 2025MCA-RG: Enhancing LLMs with Medical Concept Alignment for Radiology Report Generation-
CVPR 2025Language Guided Concept Bottleneck Models for Interpretable Continual LearningCode
CVPR 2025Concept Replacer: Replacing Sensitive Concepts in Diffusion Models via Precision LocalizationCode
CVPR 2025Attribute-formed Class-specific Concept Space: Endowing Language Bottleneck Model with Better Interpretability and ScalabilityCode
CVPR 2025Explaining Domain Shifts in Language: Concept erasing for Interpretable Image ClassificationCode
CVPR 2025CoE: Chain-of-Explanation via Automatic Visual Concept Circuit Description and Polysemanticity QuantificationCode
CVPR 2025Discovering Fine-Grained Visual-Concept Relations by Disentangled Optimal Transport Concept Bottleneck Models-
CVPR 2025Precise Fast and Low-cost Concept Erasure in_Value_SpaceCode
CVPR 2025Efficient Fine-Tuning and Concept Suppression for Pruned Diffusion ModelsCode
CVPR 2025Self-Evolving Visual Concept Library using Vision-Language CriticsCode
CVPR 2025Concept Lancet: Image Editing with Compositional Representation Transplant-
CVPR 2025Hybrid Concept Bottleneck ModelsCode
CVPR 2025Localized Concept Erasure for Text-to-Image Diffusion Models Using Training-Free Gated Low-Rank AdaptationCode
CVPR 2025ESC: Erasing Space Concept for Knowledge DeletionCode
CVPR 2025Interpretable Generative Models through Post-hoc Concept BottlenecksCode
CVPR 2025Discovering Hidden Visual Concepts Beyond Linguistic Input in Infant LearningCode
CVPR 2025MC^2: Multi-concept Guidance for Customized Multi-concept GenerationCode
CVPR 2025Interactive Medical Image Analysis with Concept-based Similarity ReasoningCode
CVPR 2025Towards Human-Understandable Multi-Dimensional Concept DiscoveryCode
CVPR 2025ICE: Intrinsic Concept Extraction from a Single Image via Diffusion ModelsCode
CVPR 2025Show and Tell: Visually Explainable Deep Neural Nets via Spatially-Aware Concept Bottleneck Modelscode
TMLR 2025Selective Concept Bottleneck Models Without Predefined ConceptsCode
2025ConceptCLIP: Towards Trustworthy Medical AI Via Concept-Enhanced Contrastive Langauge-Image Pre-trainingCode

2024

PublicationPaper TitleCode/Project
CVPR 2024Incremental Residual Concept Bottleneck ModelsCode
2024Concept-Attention Whitening for Interpretable Skin Lesion Diagnosis-
MICCAI 2024Aligning human knowledge with visual concepts towards explainable medical image classificationCode
MICCAI 2024Evidential concept embedding models: Towards reliable concept explanations for skin disease diagnosisCode
MICCAI 2024Learning a Clinically-Relevant Concept Bottleneck for Lesion Detection in Breast UltrasoundCode
2024Mask-Free Neuron Concept Annotation for Interpreting Neural Networks in Medical DomainCode
ECCV 2024Discover-then-Name: Task-Agnostic Concept Bottlenecks via Automated Concept DiscoveryCode
NeurIPS 2024VLG-CBM: Training Concept Bottleneck Models with Vision-Language GuidanceCode
LNCS 2024AdaCBM: an Adaptive Concept Bottleneck Model for Explainable and Accurate DiagnosisCode
2024Explain Via Any Concept: Concept Bottleneck Model with Open Vocabulary Concepts-
2024Self-eXplainable AI for Medical Image Analysis: A Survey and New Outlooks-
2024Concept Complement Bottleneck Model for Interpretable Medical Image Diagnosis-
NeurIPS 2024Towards Multi-dimensional Explanation Alignment for Medical Classification-
2024Explainable and interpretable multimodal large language models: A comprehensive survey-
AAAI 2024Sparsity-guided holistic explanation for llms with interpretable inference-time interventionCode
NeurIPS 2024Beyond Concept Bottleneck Models: How to Make Black Boxes Intervenable?Code
TMLR 2024CLIP-QDA: an Explainable Concept Bottleneck Model-
ICLR 2024Concept Bottleneck Generative ModelsCode
ICLR 2024Energy-Based Concept Bottleneck Models: Unifying Prediction, Concept Intervention, and Probabilistic InterpretationsCode
ICLR 2024Faithful Vision-Language Interpretation Via Concept Bottleneck ModelsCode
J-BHI 2024Guest Editorial: Trustworthy Machine Learning for Health Informatics-
2024Improving Concept Alignment in Vision-Language Concept Bottleneck ModelsCode
MICCAI 2024Integrating Clinical Knowledge into Concept Bottleneck ModelsCode
Medical Image Analysis 2024Interpretable and Intervenable Ultrasonography-Based Machine Learning Models for Pediatric AppendicitisCode
PAKDD 2024Interpreting Pretrained Language Models Via Concept BottlenecksCode
2024LVLM-Interpret: An Interpretability Tool for Large Vision-Language ModelsCode
AAAI 2024On the Concept Trustworthiness in Concept Bottleneck ModelsCode
NeurIPS 2024Stochastic Concept Bottleneck ModelsCode
2024XCoOp: Explainable Prompt Learning for Computer-Aided Diagnosis via Concept-guided Context Optimization-
ICLR 2024Concept Bottleneck Generative ModelsCode
Sci. Rep 2024Pseudo-class Part Prototype Networks for Interpretable Breast Cancer ClassificationCode
TMLR 2024Reproducibility Study of "LICO: Explainable Models with Language-Image Consistency"Code
2024Post-hoc Part-prototype Networks-
2024Embracing Unimodal Aleatoric Uncertainty for Robust Multimodal Fusion-
TMLR 2024[Re].on the Reproducibility of Post-Hoc Concept Bottleneck ModelsCode
NeurIPS 2024A Theoretical Design of Concept Sets: Improving the Predictability of Concept Bottleneck Models
NeurIPS 2024A Concept-Based Explainability Framework for Large Multimodal ModelsCode
NeurIPS 2024Classifier Clustering and Feature Alignment for Federated Learning under Distributed Concept DriftCode
NeurIPS 2024ConceptMix: A Compositional Image Generation Benchmark with Controllable DifficultyCode
NeurIPS 2024Do LLMs Dream of Elephants (when Told Not To)? Latent Concept Association and Associative Memory in Transformers-
NeurIPS 2024FinCon: A Synthesized LLM Multi-Agent System with Conceptual Verbal Reinforcement for Enhanced Financial Decision MakingCode
NeurIPS 2024Free Lunch in Pathology Foundation Model: Task-specific Model Adaptation with Concept-Guided Feature EnhancementCode
NeurIPS 2024From Causal to Concept-Based Representation Learning-
NeurIPS 2024Interpretable Concept Bottlenecks to Align Reinforcement Learning AgentsCode
NeurIPS 2024Interpretable Concept-Based Memory ReasoningCode
NeurIPS 2024Interpreting CLIP with Sparse Linear Concept Embeddings (Splice)Code
NeurIPS 2024Coarse-to-fine concept bottleneck modelsCode
NeurIPS 2024Learning Discrete Concepts in Latent Hierarchical Models-
2024LG-CAV: Train Any Concept Activation Vector with Language Guidance-
2024Neural Concept BinderCode
NeurIPS 2024No "Zero-Shot" Without Exponential Data: Pretraining Concept Frequency Determines Multimodal Model PerformanceCode
NeurIPS 2024PaCE: Parsimonious Concept Engineering for Large Language ModelsCode
NeurIPS 2024Relational Concept Bottleneck ModelsCode
NeurIPS 2024Uncovering Safety Risks of Large Language Models Through Concept Activation VectorCode

2023

PublicationPaper TitleCode/Project
AAAI 2023Interactive Concept Bottleneck Models-
ICLR 2023Post-hoc Concept Bottleneck ModelsCode
ICLR 2023LABEL-FREE CONCEPT BOTTLENECK MODELSCode
CVPR 2023Language in a Bottle: Language Model Guided Concept Bottlenecks for Interpretable Image ClassificationCode
2023STAIR: Learning Sparse Text and Image Representation in Grounded Tokens-
2023A ChatGPT Aided Explainable Framework for Zero-Shot Medical Image Diagnosis-
ICML 2023A Closer Look at the Intervention Procedure of Concept Bottleneck ModelsCode
MICCAI 2023Concept Bottleneck with Visual Concept Filtering for Explainable Medical Image Classification-
EMNLP 2023Cross-Modal Conceptualization in Bottleneck ModelsCode
NeurIPs 2023Do Concept Bottleneck Models Respect LocalitiesCode
CVPR 2023Learning bottleneck concepts in image classificationCode
ICML 2023Probabilistic Concept Bottleneck ModelsCode
2023Robust and interpretable medical image classifiers via concept bottleneck models-
CVPR 2023Towards Trustable Skin Cancer Diagnosis via Rewriting Model's Decision-

2022

PublicationPaper TitleCode/Project
ICCV 2022Explaining in Style: Training a GAN to Explain a Classifier in StyleSpace-
ICLR 2023CLIP-Dissect: Automatic Description of Neuron Representations in Deep Vision NetworksCode
NeurIPS 2022Concept Embedding ModelsCode
NeurIPS 2022Addressing Leakage in Concept Bottleneck ModelsCode

2021

PublicationPaper TitleCode/Project
ICML 2021Meaningfully Debugging Model Mistakes Using Conceptual Counterfactual ExplanationsCode
NML 2021A case-based interpretable deep learning model for classification of mass lesions in digital mammographyCode
2021Partially Interpretable Estimators (PIE): Black-Box-Refined Interpretable Machine Learning-

2020

PublicationPaper TitleCode/Project
ICML 2020Concept bottleneck modelsCode
NML 2020Concept whitening for interpretable image recognitionCode

2018

PublicationPaper TitleCode/Project
ICML 2018Interpretability Beyond Feature Attribution: Quantitative Testing with Concept Activation Vectors (TCAV)Code
J-BHI 2018Seven-Point Checklist and Skin Lesion Classification Using Multitask Multimodal Neural NetsCode

2017

PublicationPaper TitleCode/Project
AAAI 2017ConceptNet 5.5: An Open Multilingual Graph of General KnowledgeCode

📊 Benchmark

Representative benchmarks for evaluating CBMs. image

📄 Paper

If you find this work helpful, please consider citing:

@article{Wang2026CBMSurvey,
  title   = {Concept Bottleneck Models for Explainable Decision Making: A Survey of Progress, Taxonomy, and Future Directions},
  author  = {Wang, Chunjiang and Li, Fan and Hu, Wenbo and Yan, Rui and Zhang, Kun and Zhou, Shaohua Kevin},
  journal = {ResearchGate Preprint},
  year    = {2026},
  doi     = {10.13140/RG.2.2.30356.16002},
  url     = {https://www.researchgate.net/publication/399898851_Concept_Bottleneck_Models_for_Explainable_Decision_Making_A_Survey_of_Progress_Taxonomy_and_Future_Directions}
}

Similar repos

riverback/Awesome-Concept-Bottleneck-Models

A list of papers about concept bottleneck models (CBMs)

34

rushrukh/awesome-explainable-ai

A repository for summaries of recent explainable AI/Interpretable ML approaches

91

wangyongjie-ntu/Awesome-explainable-AI

A collection of research materials on explainable AI/ML

1,662

AI-in-Transportation-Lab/awesome-mechanistic-interpretability

A carefully curated collection of high-quality libraries, projects, tutorials, research papers, and…

166

adagorgun/awesome-generative-explainability

A collection of research materials on explainable generative models

26

anguyen8/XAI-papers

621

cooperleong00/Awesome-LLM-Interpretability

A curated list of LLM Interpretability related material - Tutorial, Library, Survey, Paper, Blog,…

308

ZFancy/awesome-activation-engineering

A curated list of resources for activation engineering

140