Research map / Detection and segmentation / CNN features
Better, Stronger, Faster: Tackling the Trilemma in MLLM-based Segmentation with Simultaneous Textual Mask Prediction
Jiazhen Liu, Mingkuan Feng, Long Chen
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 hybrid attention (bidirectional over mask tokens) (ablation) | STAMP-2B · single-object RES avg | cIoU: 76.2 -> 79.1 | Improved |
| add visually-enhanced mask embedding (fuse patch features) (ablation) | STAMP-2B · single-object RES avg | cIoU: 73.3 -> 79.1 | Improved |
| add SAM refinement (ablation) | STAMP-2B · single-object RES avg | cIoU: 75.3 -> 79.1 | Improved |
| lower input resolution from 896 to 504 (ablation) | STAMP-2B · single-object RES avg | cIoU: 79.1 -> 77.1 | Mixed |
| mix LLaVA VQA data with segmentation data in training (ablation) | STAMP-2B · RefCOCO val | cIoU: 81.9 -> 82.2 | Improved |
| mix segmentation data into VQA training | STAMP-2B · MMBench | accuracy: 66.5 -> 68.7 | No clear effect |
| use embedding-prediction segmentation heads (LISA, READ) with mixed training | LISA-7B · general VQA | accuracy: 0 on MMBench | Hurt |
| swap backbone from Qwen2-VL to LLaVA Vicuna-7B (ablation) | STAMP · single-object RES avg | cIoU: 79.1 (2B) -> 77.0 (Vicuna-7B) | No clear effect |
| use STAMP-7B instead of Text4Seg-13B without SAM | MLLM segmentation · gRefCOCO avg | avg of gIoU/cIoU: 67.8 -> 71.8 | Improved |
| scale STAMP from 2B to 7B | STAMP · single-object RES avg | cIoU: 79.1 -> 80.7 | Improved |
Closest papers by meaning
- SAMTok: Representing Any Mask with Two Words (CVPR 2026 · similarity 0.89)
- ALTo: Adaptive-Length Tokenizer for Autoregressive Mask Generation (NeurIPS 2025 · similarity 0.87)
- Text4Seg: Reimagining Image Segmentation as Text Generation (ICLR 2025 · similarity 0.86)
- F2RVLM: Boosting Fine-grained Fragment Retrieval for Multi-Modal Long-form Dialogue with Vision Language Model (AAAI 2026 · similarity 0.86)
- DenseMLLM: Standard Multimodal LLMs for Dense Prediction (ICML 2026 · similarity 0.86)
- Accelerating Multimodal Large Language Models by Searching Optimal Vision Token Reduction (CVPR 2025 · similarity 0.86)
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.
