# A Closer Look at the Explainability of Contrastive Language-Image Pre-training

arXiv: https://arxiv.org/abs/2304.05653 · HTML: https://mlsanity.com/paper/2304.05653/

## Key facts

- Cited by 96 papers in the map (data through 14 Oct 2026).
- Placed in Prompt learning and adaptation of CLIP, one of 16 directions in Multimodal LLMs / vision-language models.
- Builds on 32 of its 51 references that are in the map.
- First posted on arXiv 12 Apr 2023.

## Placed in

- Multimodal LLMs / vision-language models / [Prompt learning and adaptation of CLIP](https://mlsanity.com/mllm/prompt-learning-and-adaptation-of-clip/)

## Builds on (most cited first, 30 of 32)

- [Deep Residual Learning for Image Recognition](https://mlsanity.com/paper/1512.03385/) (2015)
- [Learning Transferable Visual Models From Natural Language Supervision](https://mlsanity.com/paper/2103.00020/) (2021)
- [An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale](https://mlsanity.com/paper/2010.11929/) (2020)
- [Microsoft COCO: Common Objects in Context](https://mlsanity.com/paper/1405.0312/) (2014)
- [Segment Anything](https://mlsanity.com/paper/2304.02643/) (2023)
- [Grad-CAM: Visual Explanations from Deep Networks via Gradient-based Localization](https://mlsanity.com/paper/1610.02391/) (2016)
- [The Cityscapes Dataset for Semantic Urban Scene Understanding](https://mlsanity.com/paper/1604.01685/) (2016)
- [Learning Deep Features for Discriminative Localization](https://mlsanity.com/paper/1512.04150/) (2015)
- [Grad-CAM++: Improved Visual Explanations for Deep Convolutional Networks](https://mlsanity.com/paper/1710.11063/) (2017)
- [Transformer Interpretability Beyond Attention Visualization](https://mlsanity.com/paper/2012.09838/) (2020)
- [RegionCLIP: Region-based Language-Image Pretraining](https://mlsanity.com/paper/2112.09106/) (2021)
- [Extract Free Dense Labels from CLIP](https://mlsanity.com/paper/2112.01071/) (2021)
- [GroupViT: Semantic Segmentation Emerges from Text Supervision](https://mlsanity.com/paper/2202.11094/) (2022)
- [Unmasking Clever Hans Predictors and Assessing What Machines Really Learn](https://mlsanity.com/paper/1902.10178/) (2019)
- [Generic Attention-model Explainability for Interpreting Bi-Modal and Encoder-Decoder Transformers](https://mlsanity.com/paper/2103.15679/) (2021)
- [Text2LIVE: Text-Driven Layered Image and Video Editing](https://mlsanity.com/paper/2204.02491/) (2022)
- [A Simple Baseline for Open-Vocabulary Semantic Segmentation with Pre-trained Vision-language Model](https://mlsanity.com/paper/2112.14757/) (2021)
- [A Simple Framework for Open-Vocabulary Segmentation and Detection](https://mlsanity.com/paper/2303.08131/) (2023)
- [SegCLIP: Patch Aggregation with Learnable Centers for Open-Vocabulary Semantic Segmentation](https://mlsanity.com/paper/2211.14813/) (2022)
- [ReCo: Retrieve and Co-segment for Zero-shot Transfer](https://mlsanity.com/paper/2206.07045/) (2022)
- [DualCoOp: Fast Adaptation to Multi-Label Recognition with Limited Annotations](https://mlsanity.com/paper/2206.09541/) (2022)
- [Large-scale Unsupervised Semantic Segmentation](https://mlsanity.com/paper/2106.03149/) (2021)
- [Learning Open-vocabulary Semantic Segmentation Models From Natural Language Supervision](https://mlsanity.com/paper/2301.09121/) (2023)
- [Open-world Semantic Segmentation via Contrasting and Clustering Vision-Language Embedding](https://mlsanity.com/paper/2207.08455/) (2022)
- [Texts as Images in Prompt Tuning for Multi-Label Image Recognition](https://mlsanity.com/paper/2211.12739/) (2022)
- [Pseudo-mask Matters in Weakly-supervised Semantic Segmentation](https://mlsanity.com/paper/2108.12995/) (2021)
- [Open-Vocabulary Multi-Label Classification via Multi-Modal Knowledge Transfer](https://mlsanity.com/paper/2207.01887/) (2022)
- [Exploring Visual Interpretability for Contrastive Language-Image Pre-training](https://mlsanity.com/paper/2209.07046/) (2022)
- [No Token Left Behind: Explainability-Aided Image Classification and Generation](https://mlsanity.com/paper/2204.04908/) (2022)
- [Hybrid Routing Transformer for Zero-Shot Learning](https://mlsanity.com/paper/2203.15310/) (2022)

## Built on by (most cited first, 30 of 96)

- [CLIPN for Zero-Shot OOD Detection: Teaching CLIP to Say No](https://mlsanity.com/paper/2308.12213/) (2023)
- [ClearCLIP: Decomposing CLIP Representations for Dense Vision-Language Inference](https://mlsanity.com/paper/2407.12442/) (2024)
- [ProxyCLIP: Proxy Attention Improves CLIP for Open-Vocabulary Segmentation](https://mlsanity.com/paper/2408.04883/) (2024)
- [Grounding Everything: Emerging Localization Properties in Vision-Language Transformers](https://mlsanity.com/paper/2312.00878/) (2023)
- [CLIP-Count: Towards Text-Guided Zero-Shot Object Counting](https://mlsanity.com/paper/2305.07304/) (2023)
- [A Comprehensive Survey on Segment Anything Model for Vision and Beyond](https://mlsanity.com/paper/2305.08196/) (2023)
- [Explore the Potential of CLIP for Training-Free Open Vocabulary Semantic Segmentation](https://mlsanity.com/paper/2407.08268/) (2024)
- [CLIP-AD: A Language-Guided Staged Dual-Path Model for Zero-shot Anomaly Detection](https://mlsanity.com/paper/2311.00453/) (2023)
- [VLCounter: Text-aware Visual Representation for Zero-Shot Object Counting](https://mlsanity.com/paper/2312.16580/) (2023)
- [Talking to DINO: Bridging Self-Supervised Vision Backbones with Language for Open-Vocabulary Segmentation](https://mlsanity.com/paper/2411.19331/) (2024)
- [Relax Image-Specific Prompt Requirement in SAM: A Single Generic Prompt for Segmenting Camouflaged Objects](https://mlsanity.com/paper/2312.07374/) (2023)
- [Matte Anything: Interactive Natural Image Matting with Segment Anything Models](https://mlsanity.com/paper/2306.04121/) (2023)
- [MedCLIP-SAM: Bridging Text and Image Towards Universal Medical Image Segmentation](https://mlsanity.com/paper/2403.20253/) (2024)
- [Image Segmentation in Foundation Model Era: A Survey](https://mlsanity.com/paper/2408.12957/) (2024)
- [Training-free Object Counting with Prompts](https://mlsanity.com/paper/2307.00038/) (2023)
- [Self-Calibrated CLIP for Training-Free Open-Vocabulary Segmentation](https://mlsanity.com/paper/2411.15869/) (2024)
- [Token Activation Map to Visually Explain Multimodal LLMs](https://mlsanity.com/paper/2506.23270/) (2025)
- [MISC: Ultra-low Bitrate Image Semantic Compression Driven by Large Multimodal Model](https://mlsanity.com/paper/2402.16749/) (2024)
- [MedGround-R1: Advancing Medical Image Grounding via Spatial-Semantic Rewarded Group Relative Policy Optimization](https://mlsanity.com/paper/2507.02994/) (2025)
- [Text Augmented Spatial-aware Zero-shot Referring Image Segmentation](https://mlsanity.com/paper/2310.18049/) (2023)
- [Textual Query-Driven Mask Transformer for Domain Generalized Segmentation](https://mlsanity.com/paper/2407.09033/) (2024)
- [Is CLIP ideal? No. Can we fix it? Yes!](https://mlsanity.com/paper/2503.08723/) (2025)
- [TagCLIP: A Local-to-Global Framework to Enhance Open-Vocabulary Multi-Label Classification of CLIP Without Training](https://mlsanity.com/paper/2312.12828/) (2023)
- [CLIPer: Hierarchically Improving Spatial Representation of CLIP for Open-Vocabulary Semantic Segmentation](https://mlsanity.com/paper/2411.13836/) (2024)
- [Exploring the Underwater World Segmentation without Extra Training](https://mlsanity.com/paper/2511.07923/) (2025)
- [Vision Transformers with Self-Distilled Registers](https://mlsanity.com/paper/2505.21501/) (2025)
- [As Firm As Their Foundations: Can open-sourced foundation models be used to create adversarial examples for downstream tasks?](https://mlsanity.com/paper/2403.12693/) (2024)
- [Bootstrap Fine-Grained Vision-Language Alignment for Unified Zero-Shot Anomaly Localization](https://mlsanity.com/paper/2308.15939/) (2023)
- [The devil is in the object boundary: towards annotation-free instance segmentation using Foundation Models](https://mlsanity.com/paper/2404.11957/) (2024)
- [Tri-Plane Mamba: Efficiently Adapting Segment Anything Model for 3D Medical Images](https://mlsanity.com/paper/2409.08492/) (2024)

## Abstract (arXiv)

Contrastive language-image pre-training (CLIP) is a powerful vision-language model that has shown great benefits for various tasks. However, we have identified some issues with its explainability, which undermine its credibility and limit the capacity for related tasks. Specifically, we find that CLIP tends to focus on background regions rather than foregrounds, with noisy activations at irrelevant positions on the visualization results. These phenomena conflict with conventional explainability methods based on the class attention map (CAM), where the raw model can highlight the local foreground regions using global supervision without alignment. To address these problems, we take a closer look at its architecture and features. Based on thorough analyses, we find the raw self-attentions link to inconsistent semantic regions, resulting in the opposite visualization. Besides, the noisy activations are owing to redundant features among categories. Building on these insights, we propose the CLIP Surgery for reliable CAM, a method that allows surgery-like modifications to the inference architecture and features, without further fine-tuning as classical CAM methods. This approach significantly improves the explainability of CLIP, surpassing existing methods by large margins. Besides, it enables multimodal visualization and extends the capacity of raw CLIP on open-vocabulary tasks without extra alignment. The code is available at https://github.com/xmed-lab/CLIP_Surgery.