# Pre-trained Vision-Language Models Learn Discoverable Visual Concepts

arXiv: https://arxiv.org/abs/2404.12652 · HTML: https://mlsanity.com/paper/2404.12652/

## Key facts

- Cited by 16 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 24 of its 40 references that are in the map.
- First posted on arXiv 19 Apr 2024.

## Placed in

- Multimodal LLMs / vision-language models / [VLM benchmarks and visual reasoning](https://mlsanity.com/mllm/vlm-benchmarks-and-visual-reasoning/)

## Builds on (most cited first, 24 of 24)

- [Flamingo: a Visual Language Model for Few-Shot Learning](https://mlsanity.com/paper/2204.14198/) (2022)
- [BLIP: Bootstrapping Language-Image Pre-training for Unified Vision-Language Understanding and Generation](https://mlsanity.com/paper/2201.12086/) (2022)
- [Qwen-VL: A Versatile Vision-Language Model for Understanding, Localization, Text Reading, and Beyond](https://mlsanity.com/paper/2308.12966/) (2023)
- [ViLBERT: Pretraining Task-Agnostic Visiolinguistic Representations for Vision-and-Language Tasks](https://mlsanity.com/paper/1908.02265/) (2019)
- [Oscar: Object-Semantics Aligned Pre-training for Vision-Language Tasks](https://mlsanity.com/paper/2004.06165/) (2020)
- [Concept Bottleneck Models](https://mlsanity.com/paper/2007.04612/) (2020)
- [When and why vision-language models behave like bags-of-words, and what to do about it?](https://mlsanity.com/paper/2210.01936/) (2022)
- [Winoground: Probing Vision and Language Models for Visio-Linguistic Compositionality](https://mlsanity.com/paper/2204.03162/) (2022)
- [The Neuro-Symbolic Concept Learner: Interpreting Scenes, Words, and Sentences From Natural Supervision](https://mlsanity.com/paper/1904.12584/) (2019)
- [Visual Classification via Description from Large Language Models](https://mlsanity.com/paper/2210.07183/) (2022)
- [Language in a Bottle: Language Model Guided Concept Bottlenecks for Interpretable Image Classification](https://mlsanity.com/paper/2211.11158/) (2022)
- [Neural-Symbolic VQA: Disentangling Reasoning from Vision and Language Understanding](https://mlsanity.com/paper/1810.02338/) (2018)
- [What does a platypus look like? Generating customized prompts for zero-shot image classification](https://mlsanity.com/paper/2209.03320/) (2022)
- [Vid2Seq: Large-Scale Pretraining of a Visual Language Model for Dense Video Captioning](https://mlsanity.com/paper/2302.14115/) (2023)
- [SugarCrepe: Fixing Hackable Benchmarks for Vision-Language Compositionality](https://mlsanity.com/paper/2306.14610/) (2023)
- [DetCLIP: Dictionary-Enriched Visual-Concept Paralleled Pre-training for Open-world Detection](https://mlsanity.com/paper/2209.09407/) (2022)
- [Do Concept Bottleneck Models Learn as Intended?](https://mlsanity.com/paper/2105.04289/) (2021)
- [Does CLIP Bind Concepts? Probing Compositionality in Large Image Models](https://mlsanity.com/paper/2212.10537/) (2022)
- [Interactive Concept Bottleneck Models](https://mlsanity.com/paper/2212.07430/) (2022)
- [Learning Concise and Descriptive Attributes for Visual Recognition](https://mlsanity.com/paper/2308.03685/) (2023)
- [Robust and Interpretable Medical Image Classifiers via Concept Bottleneck Models](https://mlsanity.com/paper/2310.03182/) (2023)
- [Do Vision-Language Pretrained Models Learn Composable Primitive Concepts?](https://mlsanity.com/paper/2203.17271/) (2022)
- [Language-Informed Visual Concept Learning](https://mlsanity.com/paper/2312.03587/) (2023)
- [Automatic Concept Discovery from Parallel Text and Visual Corpora](https://mlsanity.com/paper/1509.07225/) (2015)

## Built on by (most cited first, 16 of 16)

- [ViewSpatial-Bench: Evaluating Multi-perspective Spatial Localization in Vision-Language Models](https://mlsanity.com/paper/2505.21500/) (2025)
- [Explainable and Interpretable Multimodal Large Language Models: A Comprehensive Survey](https://mlsanity.com/paper/2412.02104/) (2024)
- [Discovering Fine-Grained Visual-Concept Relations by Disentangled Optimal Transport Concept Bottleneck Models](https://mlsanity.com/paper/2505.07209/) (2025)
- [Chimera: Diagnosing Shortcut Learning in Visual-Language Understanding](https://mlsanity.com/paper/2509.22437/) (2025)
- [Does VLM Classification Benefit from LLM Description Semantics?](https://mlsanity.com/paper/2412.11917/) (2024)
- [Improving Concept Alignment in Vision-Language Concept Bottleneck Models](https://mlsanity.com/paper/2405.01825/) (2024)
- [Explainable Concept Generation through Vision-Language Preference Learning for Understanding Neural Networks' Internal Representations](https://mlsanity.com/paper/2408.13438/) (2024)
- [Explaining Domain Shifts in Language: Concept erasing for Interpretable Image Classification](https://mlsanity.com/paper/2503.18483/) (2025)
- [Extraction and Analysis of Multimodal Concepts in Vision Language Models through Sparse Autoencoders](https://mlsanity.com/paper/2606.21197/) (2026)
- [SCoCCA: Multi-modal Sparse Concept Decomposition via Canonical Correlation Analysis](https://mlsanity.com/paper/2603.13884/) (2026)
- [Fine-Tuning Florence2 for Enhanced Object Detection in Un-constructed Environments: Vision-Language Model Approach](https://mlsanity.com/paper/2503.04918/) (2025)
- [Contrastive Learning with Enhanced Abstract Representations using Grouped Loss of Abstract Semantic Supervision](https://mlsanity.com/paper/2509.12771/) (2025)
- [Bridging the gap to real-world language-grounded visual concept learning](https://mlsanity.com/paper/2510.21412/) (2025)
- [PrivLEX: Detecting legal concepts in images through Vision-Language Models](https://mlsanity.com/paper/2601.09449/) (2026)
- [COMET: Concept Space Dissection of the Modality Gap in Audio-Text Multimodal Contrastive Embeddings](https://mlsanity.com/paper/2605.29628/) (2026)
- [Latent Cluster Analysis for Vision-Language-Action Models](https://mlsanity.com/paper/2609.02634/) (2026)

## Abstract (arXiv)

Do vision-language models (VLMs) pre-trained to caption an image of a "durian" learn visual concepts such as "brown" (color) and "spiky" (texture) at the same time? We aim to answer this question as visual concepts learned "for free" would enable wide applications such as neuro-symbolic reasoning or human-interpretable object classification. We assume that the visual concepts, if captured by pre-trained VLMs, can be extracted by their vision-language interface with text-based concept prompts. We observe that recent works prompting VLMs with concepts often differ in their strategies to define and evaluate the visual concepts, leading to conflicting conclusions. We propose a new concept definition strategy based on two observations: First, certain concept prompts include shortcuts that recognize correct concepts for wrong reasons; Second, multimodal information (e.g. visual discriminativeness, and textual knowledge) should be leveraged when selecting the concepts. Our proposed concept discovery and learning (CDL) framework is thus designed to identify a diverse list of generic visual concepts (e.g. "spiky" as opposed to "spiky durian"), which are ranked and selected based on visual and language mutual information. We carefully design quantitative and human evaluations of the discovered concepts on six diverse visual recognition datasets, which confirm that pre-trained VLMs do learn visual concepts that provide accurate and thorough descriptions for the recognized objects. All code and models are publicly released.