Research map / Label-efficient and robust learning / Test-time adaptation
BATCLIP: Bimodal Online Test-Time Adaptation for CLIP
Sarthak Maharana, Baoming Zhang, Leonid Karlinsky, Rogerio Feris, Yunhui Guo
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add inter-class separability loss L_sp to entropy minimization (ablation) | BATCLIP (ViT-B/16) · CIFAR-10C | mean accuracy: 60.65 -> 72.69 | Improved |
| add inter-class separability loss L_sp to entropy minimization (ablation) | BATCLIP (ViT-B/16) · CIFAR-100C | mean accuracy: 38.17 -> 41.84 | Improved |
| add inter-class separability loss L_sp to entropy minimization (ablation) | BATCLIP (ViT-B/16) · ImageNet-C | mean accuracy: 24.03 -> 30.08 | Improved |
| add projection matching loss L_pm to entropy minimization (ablation) | BATCLIP (ViT-B/16) · CIFAR-10C | mean accuracy: 60.65 -> 62.60 | Improved |
| add projection matching loss L_pm to entropy minimization (ablation) | BATCLIP (ViT-B/16) · ImageNet-C | mean accuracy: 24.03 -> 25.21 | Improved |
| add projection matching loss L_pm to L_ent + L_sp (full objective) (ablation) | BATCLIP (ViT-B/16) · CIFAR-10C | mean accuracy: 72.69 -> 73.85 | Improved |
| add projection matching loss L_pm to L_ent + L_sp (full objective) (ablation) | BATCLIP (ViT-B/16) · ImageNet-C | mean accuracy: 30.08 -> 30.72 | Improved |
| add L_sp to L_ent + L_pm (full objective) (ablation) | BATCLIP (ViT-B/16) · CIFAR-10C | mean accuracy: 62.60 -> 73.85 | Improved |
| use projection matching loss alone instead of entropy alone (ablation) | BATCLIP (ViT-B/16) · CIFAR-100C | mean accuracy: 38.17 -> 36.63 | Hurt |
| scale L_pm and L_sp by 0.3 instead of 1.0 in the full objective (ablation) | BATCLIP (ViT-B/16) · CIFAR-10C | mean accuracy: 73.31 vs 73.85 | No clear effect |
| use bimodal BATCLIP instead of SAR (vision-only TTA) | BATCLIP vs SAR (ViT-B/16) · CIFAR-10C | mean accuracy: 67.37 -> 73.85 | Improved |
| use BATCLIP instead of WATT-S* with ViT-B/32 | BATCLIP vs WATT-S* (ViT-B/32) · CIFAR-10C | mean accuracy: 70.49 vs 68.26 | Hurt |
Later papers that compared with this method
1 later accepted paper reports BATCLIP as a baseline in their results tables; 1 of them report a better number on most of the results they share with it.
- Probe, Anchor, and Amend: Active Test-Time Adaptation of Vision-Language Models (ECCV 2026 · better on 1 of 1 shared result)
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- TENT — Tent: Fully Test-Time Adaptation by Entropy Minimization (ICLR 2021) · better on 1 of 1 shared result
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- A Hard-to-Beat Baseline for Training-free CLIP-based Adaptation (ICLR 2024 · similarity 0.91)
- SuperCLIP: CLIP with Simple Classification Supervision (NeurIPS 2025 · similarity 0.90)
- What Drives Test-Time Adaptation for CLIP? A Controlled Empirical Study from an Update Perspective (NeurIPS 2026 · similarity 0.89)
- CLIPTTA: Robust Contrastive Vision-Language Test-Time Adaptation (NeurIPS 2025 · similarity 0.89)
- Diffusion Feedback Helps CLIP See Better (ICLR 2025 · similarity 0.89)
- Demystifying CLIP Data (ICLR 2024 · similarity 0.89)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
