Research map / Vision-language and multimodal / Multimodal LLMs
HiLM-D: Enhancing MLLMs with Multi-scale High-Resolution Details for Autonomous Driving
Xinpeng Ding, Jianhua Han, Hang Xu, Wei Zhang, Xiaomeng Li
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 high-resolution spatial encoder (ablation) | HiLM-D · DRAMA-ROLISP | mIoU: 43.7 -> 60.5 | Improved |
| add query-aware detector (ablation) | HiLM-D · DRAMA-ROLISP | mIoU: 47.5 -> 60.5 | Improved |
| add P-Adapter (ablation) | HiLM-D · DRAMA-ROLISP | B4 and mIoU: 53.1 -> 56.3 B4; 53.8 -> 60.5 mIoU | Improved |
| add spatial perception stream to BLIP-2 | BLIP-2 · DRAMA-ROLISP | mIoU: 46.3 -> 59.8 | Improved |
| add spatial perception stream to Video-LLaMA | Video-LLaMA · DRAMA-ROLISP | mIoU: 42.8 -> 59.3 | Improved |
| use last frame instead of first frame in spatial stream (ablation) | HiLM-D · DRAMA-ROLISP | AVG: 54.0 -> 58.4 | Improved |
| sample more frames (two or three) instead of one (ablation) | HiLM-D · DRAMA-ROLISP | B4 / mIoU: 60.5 -> 60.6 mIoU (two); 60.5 -> 60.9 (three) | No clear effect |
| train detection and captioning jointly (ablation) | HiLM-D · DRAMA-ROLISP | B4 and mIoU: 54.3 -> 56.3 B4; 52.5 -> 60.5 mIoU | Improved |
| use multi-scale features I1,I2,I3 instead of I1 only (ablation) | HiLM-D · DRAMA-ROLISP | IoU_S: 26.7 -> 33.5 | Improved |
| use three P-Adapters instead of one (ablation) | HiLM-D · DRAMA-ROLISP | mIoU: 52.6 -> 60.5 | Improved |
| add four P-Adapters instead of three (ablation) | HiLM-D · DRAMA-ROLISP | mIoU: 60.5 -> 60.4 | No clear effect |
| use alpha (ablation) | HiLM-D · DRAMA-ROLISP | B4 and mIoU: 54.7 -> 56.3 B4; 56.5 -> 60.5 mIoU | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- MiniGPT-4 — MiniGPT-4: Enhancing Vision-Language Understanding with Advanced Large Language Models (ICLR 2024) · better on 3 of 3 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- MAPLM: A Real-World Large-Scale Vision-Language Benchmark for Map and Traffic Scene Understanding (CVPR 2024 · similarity 0.89)
- SGDrive: Scene-to-Goal Hierarchical World Cognition for Autonomous Driving (CVPR 2026 · similarity 0.88)
- S4-Driver: Scalable Self-Supervised Driving Multimodal Large Language Model with Spatio-Temporal Visual Representation (CVPR 2025 · similarity 0.88)
- Holistic Autonomous Driving Understanding by Bird's-Eye-View Injected Multi-Modal Large Models (CVPR 2024 · similarity 0.88)
- STSBench: A Spatio-temporal Scenario Benchmark for Multi-modal Large Language Models in Autonomous Driving (NeurIPS 2025 · similarity 0.87)
- SpaceDrive: Infusing Spatial Awareness into VLM-based Autonomous Driving (CVPR 2026 · similarity 0.87)
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.
