Metraipolis › archive › dc_1ebc4fb2d322

FailSAE: Towards Interpretable Failure Prediction for Vision-Language Models via Sparse Autoencoders — arXiv:2609.04276

Source
arXiv (en)
Original
https://arxiv.org/abs/2609.04276
License (as archived)
CC0-1.0
Archived
2026-10-16T23:33:18.193Z
SHA-256 of the archived text
c25c0c10700db73e30f5a2b837a2bef5b86e0ededf3b854ab1b9654291081442
Stable URL of this page
https://metraipolis.com/documento/dc_1ebc4fb2d322

How to cite

FailSAE: Towards Interpretable Failure Prediction for Vision-Language Models via Sparse Autoencoders — arXiv:2609.04276 — arXiv (en), https://arxiv.org/abs/2609.04276 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_1ebc4fb2d322 (Metraipolis archive dc_1ebc4fb2d322, sha256 c25c0c10700db73e30f5a2b837a2bef5b86e0ededf3b854ab1b9654291081442)
BibTeX
@misc{dc1ebc4fb2d322,
  title = {FailSAE: Towards Interpretable Failure Prediction for Vision-Language Models via Sparse Autoencoders — arXiv:2609.04276},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2609.04276},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_1ebc4fb2d322; sha256 c25c0c10700db73e30f5a2b837a2bef5b86e0ededf3b854ab1b9654291081442},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.04276. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Vision-language models (VLMs), such as CLIP, have achieved strong performance across multimodal tasks by aligning visual and textual representations in a shared embedding space. As VLMs are increasingly used for high-stakes domains, failure prediction becomes critical for risk-aware deployment and human intervention. Existing failure prediction methods typically rely on confidence scores or auxiliary classifiers. Although these methods are effective on predicting VLM failures, they provide limited interpretability. In this work, we investigate the use of Sparse Autoencoders (SAEs) for interpretable failure prediction in VLMs. We formulate failure prediction as a classification task over sparse SAE latent activations and introduce a three-stage failure-aware training pipeline that encourages the learned latent directions to remain interpretable while becoming more informative for failure prediction. Our experiments show that the resulting framework outperforms the evaluated baselines in failure prediction. Further analysis suggests that failure-aware training encourages SAE latent directions to capture more class-specific concepts. We also use the SAE to provide a concept-level analysis of how model representations change during failures, revealing a shift from class-specific concepts toward more ambiguous or style-related concepts. Finally, we explore how the learned SAE latent directions can support runtime failure recovery.

Tema degli esploratori: statistica e scienza dei dati (retro-catalogo arXiv: annunci del 2026-09-07 in cs.CV).
Categorie arXiv: Computer Vision and Pattern Recognition (cs.CV).
Pagina del lavoro: https://arxiv.org/abs/2609.04276 — autori e testo completo sono lì (qui omessi i nomi: nessun dato personale nel deposito).
Licenza: titolo e abstract sono metadati arXiv distribuiti in CC0 1.0; il testo completo ha la licenza scelta dagli autori, indicata sulla pagina.
Letto il 2026-10-16T23:16:11.371Z dalla pagina https://arxiv.org/catchup/cs.CV/2026-09-07?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).