Does CLIP Bind Concepts? Probing Compositionality in Large Image Models

• URL: http://arxiv.org/abs/2212.10537v3
• Date: Fri, 30 Aug 2024 04:51:28 GMT
• Title: Does CLIP Bind Concepts? Probing Compositionality in Large Image Models
• Abstract summary: We focus on the ability of a large pretrained vision and language model (CLIP) to encode compositional concepts. We benchmark them on three synthetic datasets designed to test concept binding. We find that CLIP can compose concepts in a single-object setting, but in situations where concept binding is needed, performance drops dramatically.
• Score: 24.85238978067213
• License: http://creativecommons.org/licenses/by-nc-nd/4.0/
• Abstract: Large-scale neural network models combining text and images have made incredible progress in recent years. However, it remains an open question to what extent such models encode compositional representations of the concepts over which they operate, such as correctly identifying "red cube" by reasoning over the constituents "red" and "cube". In this work, we focus on the ability of a large pretrained vision and language model (CLIP) to encode compositional concepts and to bind variables in a structure-sensitive way (e.g., differentiating "cube behind sphere" from "sphere behind cube"). To inspect the performance of CLIP, we compare several architectures from research on compositional distributional semantics models (CDSMs), a line of research that attempts to implement traditional compositional linguistic structures within embedding spaces. We benchmark them on three synthetic datasets - single-object, two-object, and relational - designed to test concept binding. We find that CLIP can compose concepts in a single-object setting, but in situations where concept binding is needed, performance drops dramatically. At the same time, CDSMs also perform poorly, with best performance at chance level.

Related papers

• Cross-Modal Masked Compositional Concept Modeling for Enhancing Visio-Linguistic Compositionality [18.431462064948178]
  Contrastively trained vision-language models like CLIP, have made remarkable progress in learning joint image-text representations.<n>They often exhibit a "bag-of-words" behavior--struggling to capture the object relations, attribute-object bindings, and word order dependencies.<n>We propose MACCO, a framework that masks compositional concepts in one modality and reconstructs them conditioned on the full contextual information from the other.
  arXiv  Detail & Related papers  (2026-06-11T12:45:25Z)
• Insight: Interpretable Semantic Hierarchies in Vision-Language Encoders [52.94006363830628]
  Language-aligned vision foundation models perform strongly across diverse downstream tasks.<n>Recent works decompose these representations into human-interpretable concepts, but provide poor spatial grounding and are limited to image classification tasks.<n>We propose Insight, a language-aligned concept foundation model that provides fine-grained concepts, which are human-interpretable and spatially grounded in the input image.
  arXiv  Detail & Related papers  (2026-01-20T09:57:26Z)
• CountFormer: A Transformer Framework for Learning Visual Repetition and Structure in Class-Agnostic Object Counting [0.0]
  Humans can effortlessly count diverse objects by perceiving visual repetition and structural relationships rather than relying on class identity.<n>In this work, we introduce CountFormer, a transformer-based framework that learns to recognize repetition and structural coherence for class-agnostic object counting.
  arXiv  Detail & Related papers  (2025-10-27T19:16:02Z)
• SPARC: Concept-Aligned Sparse Autoencoders for Cross-Model and Cross-Modal Interpretability [9.90112908284836]
  We introduce (Sparse Autoencoders for Aligned Representation of Concepts) a new framework that learns a single, unified latent space shared across diverse architectures.<n>On Open Images, dramatically improves concept alignment, achieving a Jaccard similarity of 0.80, more than tripling the alignment compared to previous methods.
  arXiv  Detail & Related papers  (2025-07-07T22:29:00Z)
• Interpreting the linear structure of vision-language model embedding spaces [12.846590038965774]
  We train and release sparse autoencoders (SAEs) on the embedding spaces of four vision-language models.<n>SAEs approximate model embeddings as sparse linear combinations of learned directions, or "concepts"<n>Retraining SAEs with different seeds or different data diet leads to two findings: the rare, specific concepts captured by the SAEs are liable to change drastically, but we also show that commonly-activating concepts are remarkably stable across runs.
  arXiv  Detail & Related papers  (2025-04-16T01:40:06Z)
• Are We Done with Object-Centric Learning? [65.67948794110212]
  Object-centric learning (OCL) seeks to learn representations that only encode an object, isolated from other objects or background cues in a scene. With recent sample-efficient segmentation models, we can separate objects in the pixel space and encode them independently. We address the OOD generalization challenge caused by spurious background cues through the lens of OCL.
  arXiv  Detail & Related papers  (2025-04-09T17:59:05Z)
• Quantifying Interpretability in CLIP Models with Concept Consistency [5.921976812527759]
  We study conceptual consistency of text descriptions for attention heads in CLIP-like models. We propose Concept Consistency Score (CCS), a novel interpretability metric. We find that high CCS heads capture essential concepts and play a key role in out-of-domain detection, concept-specific reasoning, and video-language understanding.
  arXiv  Detail & Related papers  (2025-03-14T05:47:17Z)
• Universal Sparse Autoencoders: Interpretable Cross-Model Concept Alignment [6.614005142754584]
  Universal Sparse Autoencoders (USAEs) are a framework for uncovering and aligning interpretable concepts spanning multiple deep neural networks. USAEs learn a universal concept space that can reconstruct and interpret the internal activations of multiple models at once.
  arXiv  Detail & Related papers  (2025-02-06T02:06:16Z)
• Graph-guided Cross-composition Feature Disentanglement for Compositional Zero-shot Learning [54.08741382593959]
  Disentanglement of visual features of primitives (i.e., attributes and objects) has shown exceptional results in Compositional Zero-shot Learning (CZSL)<n>It is challenging to learn disentangled primitive features that are general across different compositions.<n>We propose the solution of cross-composition feature disentanglement, which takes multiple primitive-sharing compositions as inputs.
  arXiv  Detail & Related papers  (2024-08-19T08:23:09Z)
• Concept Visualization: Explaining the CLIP Multi-modal Embedding Using WordNet [4.597864989500202]
  We propose a novel saliency methodology that explains the CLIP embedding of an image by exploiting the multi-modal nature of the embeddings. ConVis makes use of lexical information from WordNet to compute task-agnostic Saliency Maps for any concept, not limited to concepts the end model was trained on.
  arXiv  Detail & Related papers  (2024-05-23T13:41:17Z)
• CAILA: Concept-Aware Intra-Layer Adapters for Compositional Zero-Shot Learning [14.496173899477283]
  We study the problem of Compositional Zero-Shot Learning (CZSL), which is to recognize novel attribute-object combinations with pre-existing concepts. We propose to insert adapters, a parameter-efficient technique proven to be effective among large language models, into each CLIP encoder layer. We further equip adapters with concept awareness so that concept-specific features of "object", "attribute", and "composition" can be extracted.
  arXiv  Detail & Related papers  (2023-05-26T07:02:57Z)
• Text-Video Retrieval with Disentangled Conceptualization and Set-to-Set Alignment [17.423361070781876]
  We propose the Disentangled Conceptualization and Set-to-set Alignment (DiCoSA) to simulate the conceptualizing and reasoning process of human beings. For disentangled conceptualization, we divide the coarse feature into multiple latent factors related to semantic concepts. For set-to-set alignment, where a set of visual concepts correspond to a set of textual concepts, we propose an adaptive pooling method to aggregate semantic concepts.
  arXiv  Detail & Related papers  (2023-05-20T15:48:47Z)
• Occlusion-Aware Instance Segmentation via BiLayer Network Architectures [73.45922226843435]
  We propose Bilayer Convolutional Network (BCNet), where the top layer detects occluding objects (occluders) and the bottom layer infers partially occluded instances (occludees) We investigate the efficacy of bilayer structure using two popular convolutional network designs, namely, Fully Convolutional Network (FCN) and Graph Convolutional Network (GCN)
  arXiv  Detail & Related papers  (2022-08-08T21:39:26Z)
• Share With Thy Neighbors: Single-View Reconstruction by Cross-Instance Consistency [59.427074701985795]
  Single-view reconstruction typically rely on viewpoint annotations, silhouettes, the absence of background, multiple views of the same instance, a template shape, or symmetry. We avoid all of these supervisions and hypotheses by leveraging explicitly the consistency between images of different object instances. Our main contributions are two approaches to leverage cross-instance consistency: (i) progressive conditioning, a training strategy to gradually specialize the model from category to instances in a curriculum learning fashion; (ii) swap reconstruction, a loss enforcing consistency between instances having similar shape or texture.
  arXiv  Detail & Related papers  (2022-04-21T17:47:35Z)
• Complex-Valued Autoencoders for Object Discovery [62.26260974933819]
  We propose a distributed approach to object-centric representations: the Complex AutoEncoder. We show that this simple and efficient approach achieves better reconstruction performance than an equivalent real-valued autoencoder on simple multi-object datasets. We also show that it achieves competitive unsupervised object discovery performance to a SlotAttention model on two datasets, and manages to disentangle objects in a third dataset where SlotAttention fails - all while being 7-70 times faster to train.
  arXiv  Detail & Related papers  (2022-04-05T09:25:28Z)
• Learning Graph Embeddings for Compositional Zero-shot Learning [73.80007492964951]
  In compositional zero-shot learning, the goal is to recognize unseen compositions of observed visual primitives states. We propose a novel graph formulation called Compositional Graph Embedding (CGE) that learns image features and latent representations of visual primitives in an end-to-end manner. By learning a joint compatibility that encodes semantics between concepts, our model allows for generalization to unseen compositions without relying on an external knowledge base like WordNet.
  arXiv  Detail & Related papers  (2021-02-03T10:11:03Z)
• Visual Concept Reasoning Networks [93.99840807973546]
  A split-transform-merge strategy has been broadly used as an architectural constraint in convolutional neural networks for visual recognition tasks. We propose to exploit this strategy and combine it with our Visual Concept Reasoning Networks (VCRNet) to enable reasoning between high-level visual concepts. Our proposed model, VCRNet, consistently improves the performance by increasing the number of parameters by less than 1%.
  arXiv  Detail & Related papers  (2020-08-26T20:02:40Z)
• Compositional Convolutional Neural Networks: A Robust and Interpretable Model for Object Recognition under Occlusion [21.737411464598797]
  We show that black-box deep convolutional neural networks (DCNNs) have only limited robustness to partial occlusion. We overcome these limitations by unifying DCNNs with part-based models into Compositional Convolutional Neural Networks (CompositionalNets) Our experiments show that CompositionalNets improve by a large margin over their non-compositional counterparts at classifying and detecting partially occluded objects.
  arXiv  Detail & Related papers  (2020-06-28T08:18:19Z)
• Linguistically Driven Graph Capsule Network for Visual Question Reasoning [153.76012414126643]
  We propose a hierarchical compositional reasoning model called the "Linguistically driven Graph Capsule Network" The compositional process is guided by the linguistic parse tree. Specifically, we bind each capsule in the lowest layer to bridge the linguistic embedding of a single word in the original question with visual evidence. Experiments on the CLEVR dataset, CLEVR compositional generation test, and FigureQA dataset demonstrate the effectiveness and composition generalization ability of our end-to-end model.
  arXiv  Detail & Related papers  (2020-03-23T03:34:25Z)
• Segmentation and Recovery of Superquadric Models using Convolutional Neural Networks [2.454342521577328]
  We present a (two-stage) approach built around convolutional neural networks (CNNs) In the first stage, our approach uses a Mask RCNN model to identify superquadric-like structures in depth scenes. We are able to describe complex structures with a small number of interpretable parameters.
  arXiv  Detail & Related papers  (2020-01-28T18:17:48Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.