Research map / Vision-language and multimodal / Multimodal LLMs
Spatial Preference Rewarding for MLLMs Spatial Understanding
Han Qiu, Peng Gao, Lewei Lu, Xiaoqin Zhang, Ling Shao, Shijian Lu
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 SPR preference optimization | Ferret-7B · REC avg, IoU 0.9 | accuracy: 43.25 -> 52.21 | Improved |
| add SPR preference optimization | Ferret-13B · REC avg, IoU 0.9 | accuracy: 49.55 -> 53.61 | Improved |
| add SPR preference optimization | Ferret-7B · Flickr30k phrase grounding IoU 0.9 | Recall@1: 29.98 -> 33.99 | Improved |
| add SPR preference optimization | Ferret-13B · Flickr30k Entities test | Recall@1: 84.76 -> 83.75 | Hurt |
| add SPR preference optimization | Ferret-13B · Ferret Bench referring reasoning | GPT score: 68.7 -> 72.6 | Improved |
| add SPR preference optimization | Ferret-7B · RefCOCOg region captioning | ROUGE_L: 15.6 -> 20.4 | Improved |
| add SPR preference optimization | LLaVA-OV-7B · LLaVA-Bench | score: 88.9 -> 91.4 | Improved |
| add SPR preference optimization | Ferret-7B · POPE | F1: 85.36 -> 85.69 | Improved |
| add grounded region description refinement (ablation) | SPR on Ferret-7B · REC avg | accuracy: 84.41 -> 84.93 | Improved |
| add grounded region description refinement (ablation) | SPR on Ferret-7B · Flickr30k phrase grounding avg | Recall@1: 81.30 -> 82.44 | Improved |
| set lambda 0.8 instead of 0.0 for score mixing (ablation) | SPR on Ferret-7B · REC avg | accuracy: 84.25 -> 84.93 | Improved |
| set lambda 0.8 instead of 1.0 for score mixing (ablation) | SPR on Ferret-7B · REC avg | accuracy: 84.45 -> 84.93 | Improved |
Closest papers by meaning
- Spatial-MLLM: Boosting MLLM Capabilities in Visual-based Spatial Intelligence (NeurIPS 2025 · similarity 0.91)
- Multi-SpatialMLLM: Multi-Frame Spatial Understanding with Multi-Modal Large Language Models (CVPR 2026 · similarity 0.90)
- SpatialRGPT: Grounded Spatial Reasoning in Vision-Language Models (NeurIPS 2024 · similarity 0.90)
- Fine-Grained Preference Optimization Improves Spatial Reasoning in VLMs (NeurIPS 2025 · similarity 0.89)
- Is A Picture Worth A Thousand Words? Delving Into Spatial Reasoning for Vision Language Models (NeurIPS 2024 · similarity 0.89)
- Why Is Spatial Reasoning Hard for VLMs? An Attention Mechanism Perspective on Focus Areas (ICML 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.
