Test whether the temperature finding transfers to SigLIP’s sigmoid loss

Determine whether the contrastive-temperature finding established for InfoNCE transfers to the sigmoid loss used by SigLIP.

Background

The paper applies ComCLIP’s InfoNCE-based objective to a pretrained SigLIP backbone and reports improved zero-shot classification and retrieval. However, this experiment does not evaluate temperature effects within SigLIP’s native sigmoid loss.

The authors explicitly leave unresolved whether their temperature analysis generalizes from InfoNCE to SigLIP’s sigmoid-loss formulation.

References

Note that this experiment uses our InfoNCE contrastive objective on a SigLIP backbone; it does not test whether the temperature finding transfers to SigLIP's sigmoid loss, which we leave to future work.

— Rethinking Contrastive Loss in CLIP Post-training: A Complementary Framework with Frozen Text Encoder  (2610.11374 - Wang et al., 8 Oct 2026) in Appendix, Section “Post-training Performance Results based on SigLIP”