- Rating
- 2
- Confidence
- 4
- Informativeness
- 2
- Presentation
- 2
- Soundness
- 2
Read review
Paper summary
This paper proposes CLIP-DCA for robust fine-tuning of CLIP. It adds a domain head and encourages separation from the classification head through soft orthogonality, using synthetic style images, MLLM descriptions, and hidden-state supervision. The method is evaluated on OOP splits and 33 target datasets, with additional unlearning experiments intended to assess dependence on pretrained domain knowledge.
Critical strengths
-
Separating domain awareness from the classification pathway is a sensible design for preserving useful pretrained information. The auxiliary domain head is discarded at inference, retaining the standard CLIP prediction pipeline.
-
The component ablations provide evidence beyond gains from additional data alone. In Table 4, keeping domain descriptions and MLLM hidden-state supervision, adding disentanglement improves average accuracy from to .
Critical weaknesses
-
Generalization beyond CLIP pretraining is not established. The OOD score is relative to ImageNet, and unlearning is a proxy that also degrades general capabilities. Moreover, Tables 1 and 8 describe OOP by cosine similarity , whereas Teterwak et al. (ICLR 2025) discard that range before splitting at . The actual evaluation subsets and claims about reduced pretraining contamination need clarification.
-
Superiority over strong baselines is not established under matched auxiliary supervision. Table 4 supports disentanglement, but Table 2 does not document matching GCC, synthetic data, and MLLM signals for the strong baselines. Table 5 matches GCC access and gives DANN the same images and descriptions, but does not establish matched hidden-state supervision.
Questions for authors
-
What exact OOP filtering rules, backbone, and sample subsets were used in Tables 1 and 8? Please provide the split code or sample lists and clarify whether the threshold discrepancy is a reporting error. What evidence supports interpreting these evaluations as generalization beyond pretraining?
-
Which auxiliary images, descriptions, and hidden-state signals does each baseline use? Do the strongest applicable robust fine-tuning baselines remain inferior when given the same auxiliary supervision as CLIP-DCA?
Thank you for your careful and insightful review! We are encouraged that you found separating domain awareness from the classification pathway to be a sensible design and noted that our component ablations provide clear evidence beyond gains from additional data alone.
W1 / Q1. OOP splits and generalization beyond pretraining
PTeterwak/DomainBed_OOP), where we use leave-one-domain-out training.W2 / Q2. Matched auxiliary supervision
Thank you again for your careful reading and helpful review!