Research map / Vision-language and multimodal / Human and embodied modeling
See What I Mean: Aligning Vision and Language Representations for Video Fine-grained Object Understanding
Boyuan Sun, Bo-Wen Yin, Yuan-Ming Li, Xihan Wei, Qibin Hou
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 attention-mask alignment training (SWIM) to Qwen2.5-VL-7B | Qwen2.5-VL-7B · VideoRefer-Bench-Q | Avg accuracy: 71.8 -> 78.3 | Improved |
| add attention-mask alignment training (SWIM) to Qwen2.5-VL-7B | Qwen2.5-VL-7B · VideoRefer-Bench-D | Avg score: 3.24 -> 3.78 | Improved |
| supervise multiple layers (6, evenly spaced) instead of a single layer (ablation) | SWIM · VideoRefer-Bench-D | Avg score: 3.43 -> 3.78 | Improved |
| supervise more than six layers (ablation) | SWIM · VideoRefer-Bench-D | Avg score: 3.78 -> 3.77 | No clear effect |
| evenly spaced layers instead of clustered late layers (ablation) | SWIM · VideoRefer-Bench-D | Avg score: 3.76 -> 3.78 | Improved |
| mean fusion instead of Add fusion of layer attention maps (ablation) | SWIM · VideoRefer-Bench-D | Avg score: 3.57 -> 3.78 | Improved |
| mean fusion instead of Pool fusion (ablation) | SWIM · VideoRefer-Bench-D | Avg score: 3.49 -> 3.78 | Improved |
| mean fusion instead of element-wise product fusion (ablation) | SWIM · VideoRefer-Bench-D | Avg score: 3.55 -> 3.78 | Improved |
| BCE loss instead of mIoU, Focal or Dice (ablation) | SWIM · VideoRefer-Bench-D | Avg score: 3.71 -> 3.78 | Improved |
| add SWIM alignment training to Qwen2.5-VL-7B | Qwen2.5-VL-7B · GamePoint@P-1 | GamePoint: 0.329 -> 0.392 | Improved |
| add SWIM alignment training: attention quality metrics | Qwen2.5-VL-7B · attention vs masks | AUC: 0.62 -> 0.67 | Improved |
| replace tagged nouns with synonyms at test time (ablation) | SWIM · VideoRefer-Bench-D | Avg score: 3.78 -> 3.74 | No clear effect |
Closest papers by meaning
- ClawMachine: Learning to Fetch Visual Tokens for Referential Comprehension (ICLR 2025 · similarity 0.90)
- Vision as a Dialect: Unifying Visual Understanding and Generation via Text-Aligned Representations (NeurIPS 2025 · similarity 0.90)
- Same or Not? Enhancing Visual Perception in Vision-Language Models (CVPR 2026 · similarity 0.90)
- Analyzing Fine-Grained Alignment and Enhancing Vision Understanding in Multimodal Language Models (NeurIPS 2025 · similarity 0.90)
- Hugging Visual Prompt and Segmentation Tokens: Consistency Learning for Fine-Grained Visual Understanding in MLLMs (CVPR 2026 · similarity 0.90)
- Synthesize Diagnose and Optimize: Towards Fine-Grained Vision-Language Understanding (CVPR 2024 · 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.
