From Weights to Concepts: Data-Free Interpretability of CLIP via Singular Vector Decomposition

• URL: http://arxiv.org/abs/2603.24653v1
• Date: Wed, 25 Mar 2026 17:59:57 GMT
• Title: From Weights to Concepts: Data-Free Interpretability of CLIP via Singular Vector Decomposition
• Abstract summary: We introduce SITH, a fully data-free, training-free framework that analyzes CLIP's vision transformer in weight space.<n>For each attention head, we decompose its value-output matrix into singular vectors and interpret each one via COMP.<n>We show that SITH yields coherent, faithful intra-head explanations, validated through reconstruction fidelity and interpretability experiments.
• Score: 33.4228178732749
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: As vision-language models are deployed at scale, understanding their internal mechanisms becomes increasingly critical. Existing interpretability methods predominantly rely on activations, making them dataset-dependent, vulnerable to data bias, and often restricted to coarse head-level explanations. We introduce SITH (Semantic Inspection of Transformer Heads), a fully data-free, training-free framework that directly analyzes CLIP's vision transformer in weight space. For each attention head, we decompose its value-output matrix into singular vectors and interpret each one via COMP (Coherent Orthogonal Matching Pursuit), a new algorithm that explains them as sparse, semantically coherent combinations of human-interpretable concepts. We show that SITH yields coherent, faithful intra-head explanations, validated through reconstruction fidelity and interpretability experiments. This allows us to use SITH for precise, interpretable weight-space model edits that amplify or suppress specific concepts, improving downstream performance without retraining. Furthermore, we use SITH to study model adaptation, showing how fine-tuning primarily reweights a stable semantic basis rather than learning entirely new features.

Related papers

• Large Vision-Language Models Get Lost in Attention [51.851592109135716]
  We propose a unified framework grounded in information theory and geometry to quantify the geometric and entropic nature of residual updates.<n>Applying this unified framework reveals a fundamental functional decoupling: Attention acts as a subspace-language operator focused on reconfiguration, whereas FFNs serve as subspace-language operators driving semantic innovation.
  arXiv  Detail & Related papers  (2026-05-07T04:45:52Z)
• Data-Driven Variational Basis Learning Beyond Neural Networks: A Non-Neural Framework for Adaptive Basis Discovery [0.0]
  We develop a non-neural alternative that learns basis functions directly from data through variational optimization.<n>The proposed framework, termed Data Driven Variational Basis Learning (DVBL), treats basis atoms as primary optimization variables.<n>This yields a data-adaptive basis expansion that remains explicit, interpretable, and amenable to rigorous analysis.
  arXiv  Detail & Related papers  (2026-04-17T23:01:34Z)
• Weight-Informed Self-Explaining Clustering for Mixed-Type Tabular Data [63.62853416081748]
  WISE is a framework that unifies representation, feature weighting, clustering, and interpretation.<n>It produces faithful, human-interpretable explanations grounded in the same primitives that drive clustering.
  arXiv  Detail & Related papers  (2026-04-07T13:18:31Z)
• Making Foundation Models Probabilistic via Singular Value Ensembles [56.4174499669573]
  Foundation models have become a dominant paradigm in machine learning, achieving remarkable performance across diverse tasks through large-scale pretraining.<n>Standard approach to quantifying uncertainty, training an ensemble of independent models, incurs prohibitive computational costs that scale linearly with ensemble size.<n>We propose Singular Value Ensemble (SVE), a parameter-efficient implicit ensemble method that builds on a simple, but powerful core assumption.<n>We show that SVE uncertainty quantification achieves comparable to explicit deep ensembles while increasing the parameter count of the base model by less than 1%.
  arXiv  Detail & Related papers  (2026-01-29T18:07:18Z)
• Knowledge-Informed Neural Network for Complex-Valued SAR Image Recognition [51.03674130115878]
  We introduce the Knowledge-Informed Neural Network (KINN), a lightweight framework built upon a novel "compression-aggregation-compression" architecture.<n>KINN establishes a state-of-the-art in parameter-efficient recognition, offering exceptional generalization in data-scarce and out-of-distribution scenarios.
  arXiv  Detail & Related papers  (2025-10-23T07:12:26Z)
• Contrastive Consolidation of Top-Down Modulations Achieves Sparsely Supervised Continual Learning [5.3151394920162645]
  We introduce task-modulated contrastive learning (TMCL), which takes inspiration from the biophysical machinery in the neocortex.<n>TMCL integrates top-down information continually and without supervision.<n>Our experiments show improvements in both class-incremental and transfer learning over state-of-the-art unsupervised approaches.
  arXiv  Detail & Related papers  (2025-05-20T09:31:57Z)
• Structural Entropy Guided Probabilistic Coding [52.01765333755793]
  We propose a novel structural entropy-guided probabilistic coding model, named SEPC.<n>We incorporate the relationship between latent variables into the optimization by proposing a structural entropy regularization loss.<n> Experimental results across 12 natural language understanding tasks, including both classification and regression tasks, demonstrate the superior performance of SEPC.
  arXiv  Detail & Related papers  (2024-12-12T00:37:53Z)
• Disentanglement via Latent Quantization [60.37109712033694]
  In this work, we construct an inductive bias towards encoding to and decoding from an organized latent space. We demonstrate the broad applicability of this approach by adding it to both basic data-re (vanilla autoencoder) and latent-reconstructing (InfoGAN) generative models.
  arXiv  Detail & Related papers  (2023-05-28T06:30:29Z)
• Interpretable Sentence Representation with Variational Autoencoders and Attention [0.685316573653194]
  We develop methods to enhance the interpretability of recent representation learning techniques in natural language processing (NLP) We leverage Variational Autoencoders (VAEs) due to their efficiency in relating observations to latent generative factors. We build two models with inductive bias to separate information in latent representations into understandable concepts without annotated data.
  arXiv  Detail & Related papers  (2023-05-04T13:16:15Z)
• Harnessing the Power of Explanations for Incremental Training: A LIME-Based Approach [6.244905619201076]
  In this work, model explanations are fed back to the feed-forward training to help the model generalize better. The framework incorporates the custom weighted loss with Elastic Weight Consolidation (EWC) to maintain performance in sequential testing sets. The proposed custom training procedure results in a consistent enhancement of accuracy ranging from 0.5% to 1.5% throughout all phases of the incremental learning setup.
  arXiv  Detail & Related papers  (2022-11-02T18:16:17Z)

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.