Contrastive Semantic Projection: Faithful Neuron Labeling with Contrastive Examples
For interpretability researchers, this work provides a practical improvement to neuron labeling by leveraging contrastive examples, though it is an incremental extension of existing methods.
The paper introduces Contrastive Semantic Projection (CSP), a method for neuron labeling that uses contrastive examples to improve faithfulness and semantic granularity. Experiments show CSP outperforms state-of-the-art baselines in both faithfulness and specificity, with a case study on melanoma detection.
Neuron labeling assigns textual descriptions to internal units of deep networks. Existing approaches typically rely on highly activating examples, often yielding broad or misleading labels by focusing on dominant but incidental visual factors. Prior work such as FALCON introduced contrastive examples -- inputs that are semantically similar to activating examples but elicit low activations -- to sharpen explanations, but it primarily addresses subspace-level interpretability rather than scalable neuron-level labeling. We revisit contrastive explanations for neuron-level labeling in two stages: (1) candidate label generation with vision language models (VLMs) and (2) label assignment with CLIP-like encoders. First, we show that providing contrastive image sets to VLMs yields candidate labels that are more specific and more faithful. Second, we introduce Contrastive Semantic Projection (CSP), an extension of SemanticLens that incorporates contrastive examples directly into its CLIP-based scoring and selection pipeline. Across extensive experiments and a case study on melanoma detection, contrastive labeling improves both faithfulness and semantic granularity over state-of-the-art baselines. Our results demonstrate that contrastive examples are a simple yet powerful and currently underutilized component of neuron labeling and analysis pipelines.