AI RESEARCH
Zero-Shot Textual Explanations via Translating Decision-Critical Features
arXiv CS.CV
•
ArXi:2512.07245v2 Announce Type: replace Textual explanations make image classifier decisions transparent by describing the prediction rationale in natural language. Large vision-language models can generate captions but are designed for general visual understanding, not classifier-specific reasoning. Existing zero-shot explanation methods align global image features with language, producing descriptions of what is visible rather than what drives the prediction. We propose TEXTER, which overcomes this limitation by isolating decision-critical features before alignment.