Divyaansh Mertia, Chirag Bhutra, Himanshu Shekhar, Shubham Bagwari, Pratik Mazumder · 2026 · arXiv
DOI: 10.48550/arXiv.2610.29688 · open-access PDF
Grandes modelos pré-treinados, como o CLIP, são amplamente usados como extraidores de recursos congelados. No entanto, suas incorporações herdam correlações falsas de dados de treinamento, introduzindo preconceitos nos recursos. Um corpo crescente de pesquisa procurou abordar este problema através de debiasing pós-hoc, deixando a espinha intacta. Descobrimos que, apesar dos mecanismos diferentes, esses métodos exigem que o recurso de preconceito seja identificado e expressado em texto. Esta dependência não é inofensiva. Neste trabalho, demonstramos experimentalmente que este requisito incorre em uma punição imprevista em certos cenários.
Machine translation (Meta M2M100); the English original follows.
Large pretrained models such as CLIP are widely used as frozen feature extractors. However, their embeddings inherit spurious correlations from training data, introducing bias into the features. A growing body of research has sought to address this issue through post hoc debiasing, leaving the backbone untouched. We discover that despite differing mechanisms, these methods require the bias feature to be identified and expressed in text. This dependency is not innocuous. In this work, we experimentally demonstrate that this requirement incurs an unforeseen penalty in certain scenarios. For exam
Read in: English · Spanish · Chinese · Hindi · Arabic · Portuguese · French · German · Japanese · Russian · Korean · Indonesian · Vietnamese
GET https://store.agentexchange.work/gw/papers/verify?title=...&year=... (0.003 USD), fetch this record as JSON with /gw/papers/paper?id=2610.29688, or list what it cites with /gw/papers/references?id=.... Pay per call in USDC (x402) or with a 19 USD card key for 1,000 calls. Catalogue · llms.txtSources: OpenAlex (CC0), arXiv, Crossref. Agent Exchange