Research map / Vision-language and multimodal / Multimodal LLMs
360° Image Perception with MLLMs: A Comprehensive Benchmark and a Training-Free Method
Huyen Thi Thanh Tran, Van-Quang Nguyen, Farros Alferro, Kang-Jun Liu, Takayuki Okatani
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 |
|---|---|---|---|
| use CMP instead of ERP input | Gemini Pro 2.5 · 360Bench overall | accuracy: 43.9 -> 46.5 | Improved |
| use CMP instead of ERP input | LLaVA-CoT · 360Bench PP-IR | accuracy: +14.1% | Improved |
| use CMP instead of ERP input | Gemini Flash 2.5 · 360Bench PP-IC | accuracy: +12.2% | Improved |
| use ERP instead of CMP input | Gemini Flash 2.5 · 360Bench SR-OV | accuracy: +14.5% | Improved |
| use ERP instead of CMP input | Gemini Pro 2.5 · 360Bench DG | accuracy: +6.8% | Improved |
| add Free360 scene-graph reasoning (hybrid) | Qwen2.5-VL-7B · 360Bench overall | accuracy: 38.1 -> 45.3 | Improved |
| add Free360 scene-graph reasoning | Qwen2.5-VL-7B · 360Bench SR-OV | accuracy: +22.9% | Improved |
| add ZoomEye patch search | Qwen2.5-VL-7B · 360Bench FP-IC | accuracy: -1.3% | Hurt |
| add Omni-CoT view decomposition | Qwen2.5-VL-7B · 360Bench SR-OV | accuracy: +10.6% | Improved |
| add image cropping for attribute extraction (ablation) | Free360 · 360Bench overall | accuracy: 42.7 -> 45.3 | Improved |
| add spherical rotation for inter-entity relations (ablation) | Free360 · 360Bench overall | accuracy: 44.7 -> 45.3 | Improved |
| add entity-view relation detection (ablation) | Free360 · 360Bench overall | accuracy: 42.8 -> 45.3 | Improved |
Closest papers by meaning
- FOCUS: Internal MLLM Representations for Efficient Fine-Grained Visual Question Answering (NeurIPS 2025 · similarity 0.90)
- SpatiaLQA: A Benchmark for Evaluating Spatial Logical Reasoning in Vision-Language Models (CVPR 2026 · similarity 0.90)
- Identifying and Mitigating Position Bias of Multi-image Vision-Language Models (CVPR 2025 · similarity 0.90)
- Language-Image Models with 3D Understanding (ICLR 2025 · similarity 0.90)
- MMSI-Bench: A Benchmark for Multi-Image Spatial Intelligence (ICLR 2026 · similarity 0.90)
- Why Is Spatial Reasoning Hard for VLMs? An Attention Mechanism Perspective on Focus Areas (ICML 2025 · similarity 0.90)
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.
