REVIEW 11 cited by
Interpreting CLIP's Image Representation via Text-Based Decomposition
Not yet reviewed by Pith; the record is open.
This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.
SPECIMEN: schema-true, not a live event
T0 review · schema-true
One-sentence machine reading of the paper's core claim.
pith:XXXXXXXX · record.json · timestamp
Interpreting CLIP's Image Representation via Text-Based Decomposition
read the original abstract
We investigate the CLIP image encoder by analyzing how individual model components affect the final representation. We decompose the image representation as a sum across individual image patches, model layers, and attention heads, and use CLIP's text representation to interpret the summands. Interpreting the attention heads, we characterize each head's role by automatically finding text representations that span its output space, which reveals property-specific roles for many heads (e.g. location or shape). Next, interpreting the image patches, we uncover an emergent spatial localization within CLIP. Finally, we use this understanding to remove spurious features from CLIP and to create a strong zero-shot image segmenter. Our results indicate that a scalable understanding of transformer models is attainable and can be used to repair and improve models.
Forward citations
Cited by 11 Pith papers
-
V-SEAM: Visual Semantic Editing and Attention Modulating for Causal Interpretability of Vision-Language Models
V-SEAM combines concept-level visual semantic editing with attention head modulation to identify positive and negative contributors across object, attribute, and relationship levels, then uses this to improve VLM perf...
-
AnchorDiff: Training-Free Concept Grounding for MM-DiTs via Anchor-Based Graph Propagation
AnchorDiff performs training-free concept grounding in multi-modal diffusion transformers by anchor selection followed by graph propagation on attention-derived graphs, reducing concept leakage on a new multi-concept dataset.
-
Letting the neural code speak: Automated characterization of monkey visual neurons through human language
Natural-language descriptions generated and verified through generative models and digital twins capture the selectivity of most neurons in macaque V1 and V4.
-
Letting the neural code speak: Automated characterization of monkey visual neurons through human language
Natural language descriptions generated via a closed-loop pipeline with digital twins capture the selectivity of most neurons in macaque V1 and V4, with synthesized images driving 96% of V4 neurons into the top or bot...
-
BrainExplore: Large-Scale Discovery of Interpretable Visual Representations in the Human Brain
A new automated pipeline decomposes fMRI activity into components and labels them with visual concepts, claiming thousands of interpretable patterns across the human visual cortex.
-
The SuperActivator Mechanism: Transformers Concentrate Reliable Concept Signals in the Tail
Reliable concept presence in transformers is concentrated in the extreme high-activation tail of in-concept tokens; thresholding that tail improves concept detection and localization.
-
Sparse Feature Circuits: Discovering and Editing Interpretable Causal Graphs in Language Models
Sparse feature circuits are introduced as interpretable causal subnetworks in language models, supporting unsupervised discovery of thousands of circuits and a method called SHIFT to improve classifier generalization ...
-
AnimeAdapter: A Modular Adapter for Appearance-Consistent Anime Character Generation
AnimeAdapter is a modular adapter for Stable Diffusion that enables appearance-consistent anime character generation from a single reference image using semantic-selective local attention and pose-aware conditioning, ...
-
MEDIC-AD: Towards Medical Vision-Language Model's Clinical Intelligence
MEDIC-AD adds anomaly-aware and difference tokens to a medical VLM, claiming SOTA lesion detection, temporal tracking, and visual grounding; the zero-shot claim is undermined by likely train/test overlap.
-
CLIP-SVD: Efficient and Interpretable Vision-Language Adaptation via Singular Values
CLIP-SVD performs parameter-efficient adaptation of CLIP by fine-tuning singular values from SVD of weight matrices, reporting SOTA few-shot accuracy on 21 datasets plus a language-based interpretability analysis.
-
AnimeAdapter: A Modular Adapter for Appearance-Consistent Anime Character Generation
AnimeAdapter is a pretrained lightweight adapter for Stable Diffusion that uses semantic-selective local attention from CLIP and pose-aware conditioning to enable zero-shot fine-grained consistent anime character gene...
discussion (0)
Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.