Research map / Vision-language and multimodal / CLIP and zero-shot
Tuning Multi-mode Token-level Prompt Alignment across Modalities
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| Use token-level multi-mode OT alignment instead of MaPLe's projection-based prompts | MaPLe · Base-to-New average | harmonic mean H: 78.55 -> 79.25 | Helped |
| Use ALIGN instead of CoOp | CoOp · Base-to-New average | new accuracy: 63.22 -> 75.51 | Helped |
| Use ALIGN instead of MaPLe | MaPLe · Cross-dataset transfer average | accuracy: 66.30 -> 67.15 | Helped |
| Use ALIGN instead of MaPLe | MaPLe · ImageNet-Sketch | accuracy: 49.15 -> 49.96 | Helped |
| Use ALIGN instead of MaPLe | MaPLe · ImageNet-R | accuracy: 76.98 -> 76.16 | Hurt |
| Use ALIGN instead of MaPLe | MaPLe · EuroSAT 1-shot | accuracy: 66.67 -> 53.23 | Hurt |
| Use ALIGN instead of MaPLe | MaPLe · DTD base-to-new H | harmonic mean H: 68.16 -> 65.28 | Hurt |
| Use ALIGN instead of MaPLe | MaPLe · Stanford Cars base-to-new H | harmonic mean H: 73.47 -> 76.80 | Helped |
| Add hierarchical OT with multiple prompts (more parameters, slower inference) vs MaPLe | ALIGN · Complexity | inference fps: 282 -> 62 | Hurt |
| Run OT operations in parallel on GPU compared with CoCoOp | ALIGN · Complexity | inference fps: 37 -> 62 | Helped |
| Vary beta weighting prompt-level vs token-level OT in [0, 0.2, 0.5, 0.7, 1.0]; combining both (ablation) | ALIGN · Base-to-New | accuracy | Helped |
| Tune beta per dataset (ablation) | ALIGN · Base-to-New | accuracy | Helped |
Closest papers by meaning
- Amend to Alignment: Decoupled Prompt Tuning for Mitigating Spurious Correlation in Vision-Language Models (ICML 2024 · similarity 0.91)
- LAMM: Label Alignment for Multi-Modal Prompt Learning (AAAI 2024 · similarity 0.90)
- SA²VP: Spatially Aligned-and-Adapted Visual Prompt (AAAI 2024 · similarity 0.90)
- Learning to Learn Better Visual Prompts (AAAI 2024 · similarity 0.89)
- MmAP: Multi-Modal Alignment Prompt for Cross-Domain Multi-Task Learning (AAAI 2024 · similarity 0.89)
- VIPAMIN: Visual Prompt Initialization via Embedding Selection and Subspace Expansion (NeurIPS 2025 · 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.