Research map / Vision-language and multimodal / CLIP and zero-shot
TecoPrompt: Temporal-Conservative Prompt Learning for Vision-Language Models
zeyi shao, Haowen Hua, Jiaxin Zhang, John See, Zeyd Boukhers, Cong Yang
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 temporal-stable hard label rewriting (ablation) | TecoPrompt · OxfordPets sym noise (avg) | accuracy: 84.44 -> 86.26 | Improved |
| add temporal-stable rewriting (ablation) | TecoPrompt · OxfordPets 70% sym noise | accuracy: 78.22 -> 85.03 | Improved |
| add EMA confidence gate alone (no rewriting) (ablation) | TecoPrompt · OxfordPets sym noise (avg) | accuracy: 84.44 -> 84.67 | No clear effect |
| add EMA on top of rewriting (ablation) | TecoPrompt · OxfordPets sym noise (avg) | accuracy: 86.26 -> 86.96 | Mixed |
| use K=8 stability window instead of K=1 (ablation) | TecoPrompt · OxfordPets sym noise (avg) | accuracy: 80.59 -> 86.96 | Improved |
| use K=8 instead of K=16 (ablation) | TecoPrompt · OxfordPets sym noise (avg) | accuracy: 86.48 -> 86.96 | Improved |
| assign improper group-wise loss (0.7-0.5-0.3) (ablation) | TecoPrompt · OxfordPets 70% sym noise | accuracy: 85.36 -> 58.62 | Hurt |
| use temporally verified hard rewriting vs OT soft targets (ablation) | TecoPrompt · Flowers102 80% asym noise | accuracy: 71.80 -> 73.14 | Improved |
| TecoPrompt vs NLPrompt on real noisy labels | TecoPrompt · Food101N | accuracy: 76.46 -> 78.67 | Improved |
| TecoPrompt at low noise vs NLPrompt | TecoPrompt · DTD 12.5% sym noise | accuracy: 63.0 vs 62.7 | No clear effect |
| temporal overhead of history and EMA | TecoPrompt · Caltech101 16-shot | time per epoch: +0.20 seconds per epoch | No clear effect |
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.
- NLPrompt — NLPrompt: Noise-Label Prompt Learning for Vision-Language Models (CVPR 2025) · better on 10 of 10 shared results
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
- NLPrompt: Noise-Label Prompt Learning for Vision-Language Models (CVPR 2025 · similarity 0.90)
- JoAPR: Cleaning the Lens of Prompt Learning for Vision-Language Models (CVPR 2024 · similarity 0.89)
- Black-Box Test-Time Prompt Tuning for Vision-Language Models (AAAI 2025 · similarity 0.89)
- A Closer Look at the Few-Shot Adaptation of Large Vision-Language Models (CVPR 2024 · similarity 0.87)
- Learning to Prompt with Text Only Supervision for Vision-Language Models (AAAI 2025 · similarity 0.87)
- Frustratingly Easy Test-Time Adaptation of Vision-Language Models (NeurIPS 2024 · similarity 0.87)
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.
