Research map / Detection and segmentation / Open-set and few-shot detection
Tokenize Anything via Prompting
Ting Pan, Lulu Tang, Xinlong Wang, Shiguang Shan
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 concept-prediction loss (KL to CLIP concept distribution) on semantic token (ablation) | TAP promptable tokenizer · VG region captioning (partial-FT) | CIDEr: 105.2 -> 154.7 | Improved |
| use concept alignment instead of feature alignment for CLIP priors (ablation) | TAP promptable tokenizer · LVIS zero-shot classification | AP: 39.1 -> 59.1 | Improved |
| use concept alignment instead of feature alignment (ablation) | TAP promptable tokenizer · COCO zero-shot classification | AP: 62.0 -> 77.0 | Improved |
| use semantic token instead of mask token to prompt text decoder (ablation) | TAP captioning · VG region captioning (partial-FT) | CIDEr: 138.9 -> 154.7 | Improved |
| add semantic prediction to pre-training (ablation) | TAP promptable tokenizer · COCO/LVIS segmentation | mask AP: 46.1 -> 46.0 (COCO), 43.2 -> 43.0 (LVIS) | No clear effect |
| scale text decoder depth and width (ablation) | TAP-L captioning · VG region captioning | CIDEr: 153.2 -> 154.7 | No clear effect |
| unfreeze image encoder-decoder and fine-tune end to end (full-FT) | TAP-L captioning · Visual Genome region captioning | CIDEr: 154.7 -> 164.7 | Improved |
| increase point prompts from 1 to 5 instead of box prompt (ablation) | TAP-L classification · LVIS | AP_cls: 48.1 -> 58.7 (box 59.1) | Improved |
| use compact TAP-L tokenizer instead of 5B EVA-CLIP crop classifier | Region classification · LVIS zero-shot classification | AP: 64.3 -> 59.1 | Hurt |
Closest papers by meaning
- Subobject-level Image Tokenization (ICML 2025 · similarity 0.88)
- Perceive Anything: Recognize, Explain, Caption, and Segment Anything in Images and Videos (NeurIPS 2025 · similarity 0.88)
- Segment Anyword: Mask Prompt Inversion for Open-Set Grounded Segmentation (ICML 2025 · similarity 0.88)
- Segment and Caption Anything (CVPR 2024 · similarity 0.87)
- Visual In-Context Prompting (CVPR 2024 · similarity 0.87)
- Hugging Visual Prompt and Segmentation Tokens: Consistency Learning for Fine-Grained Visual Understanding in MLLMs (CVPR 2026 · 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.
