Research map / Vision-language and multimodal / Human and embodied modeling
Background Blurring Matters: Improving Visual Grounding by Merging Text-Irrelevant Tokens
Ruilin Yao, Shengwu Xiong, Shanshan Yang, Tianyu Zou, Shili Xiong, Yi Rong
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 ToB module | Baseline (DINOv2-B/BERT-B) · RefCOCOg test-u | ACC@0.5: 76.05 -> 80.36 | Improved |
| add ToB module | Baseline (DINOv2-B/BERT-B) · RefCOCO val | ACC@0.5: 85.55 -> 87.73 | Improved |
| add ToB module | CLIP-VG · RefCOCO+ testB | ACC@0.5: 57.62 -> 65.36 | Improved |
| add ToB module | SimVG · RefCOCOg test-u | ACC@0.5: 80.51 -> 81.79 | Improved |
| add ToB module | OneRef · RefCOCOg test-u | ACC@0.5: 83.52 -> 84.64 | Improved |
| add ToB module | SimVG (28K) · RefCOCO testB | ACC@0.5: 87.94 -> 87.43 | Hurt |
| add ToB module | SimVG (28K) · RefCOCOg test-u | ACC@0.5: 86.23 -> 86.77 | Improved |
| add ToMe token merging to baseline (ablation) | Baseline · RefCOCOg test-u | ACC@0.5: 76.05 -> 75.25 | Hurt |
| add ToE token merging to baseline (ablation) | Baseline · RefCOCOg test-u | ACC@0.5: 76.05 -> 74.89 | Hurt |
| add LAPS pruning to baseline (ablation) | Baseline · RefCOCO+ val | ACC@0.5: 75.99 -> 73.54 | Hurt |
| add FitPrune pruning to baseline (ablation) | Baseline · RefCOCO+ val | ACC@0.5: 75.99 -> 61.78 | Hurt |
| increase merged tokens r from 0 to 64 (ablation) | Baseline + ToB · RefCOCOg test-u | ACC@0.5: 79.12 -> 80.36 | Improved |
Closest papers by meaning
- Your Large Vision-Language Model Only Needs A Few Attention Heads For Visual Grounding (CVPR 2025 · similarity 0.90)
- SimVG: A Simple Framework for Visual Grounding with Decoupled Multi-modal Fusion (NeurIPS 2024 · similarity 0.89)
- CLIP Tricks You: Training-free Token Pruning for Efficient Pixel Grounding in Large Vision-Language Models (ICML 2026 · similarity 0.89)
- Referencing Where to Focus: Improving Visual Grounding with Referential Query (NeurIPS 2024 · similarity 0.89)
- Look Around Before Locating: Considering Content and Structure Information for Visual Grounding (AAAI 2025 · similarity 0.88)
- Connecting the Dots: Training-Free Visual Grounding via Agentic Reasoning (AAAI 2026 · similarity 0.88)
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.
