Research map / Vision-language and multimodal / Multimodal LLMs
Hints of Prompt: Enhancing Visual Representation for Multimodal LLMs in Autonomous Driving
Hao Zhou, Zhanning Gao, Zhili Chen, Maosheng Ye, Qifeng Chen, Tongyi Cao, Honggang Qi
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 Hints of Prompt (three hints) to LLaVA-v1.5 | LLaVA-v1.5 · LingoQA | Lingo-Judge: 63.2 -> 67.8 | Improved |
| add HoP to LLaVA-v1.5 | LLaVA-v1.5 · BDD-X (All) | CIDEr: 96.51 -> 102.20 | Improved |
| add HoP to LLaVA-v1.5 | LLaVA-v1.5 · DRAMA | BLEU-4: 56.1 -> 56.3 | Improved |
| train HoP with only 25% data (ablation) | HoP · LingoQA | Lingo-Judge: 64.0 vs full-data LLaVA-v1.5 63.2 | Improved |
| use Joint Cross-Attention instead of concatenation to fuse hints (ablation) | HoP · LingoQA | Lingo-Judge: 58.8 -> 67.8 | Improved |
| use Joint Cross-Attention instead of Self-Cross-Attention (ablation) | HoP · LingoQA | Lingo-Judge: 65.2 -> 67.8 | Improved |
| use Joint Cross-Attention instead of Sequential Cross-Attention (ablation) | HoP · LingoQA | Lingo-Judge: 67.0 -> 67.8 | Improved |
| use Mask2Former instead of GroundingDINO for Semantic hint (ablation) | HoP · LingoQA | Lingo-Judge: 66.4 -> 67.8 | Improved |
| take Question hint from LLM embeddings instead of CLIP text (ablation) | HoP · LingoQA | Lingo-Judge: 65.2 -> 67.8 | Improved |
| increase Semantic hint tokens from 8 to 16 (ablation) | HoP · LingoQA | all metrics | Improved |
| increase Semantic hint tokens from 32 to 64 (ablation) | HoP · LingoQA | all metrics | Hurt |
| use DINOv2 similarity tokens instead of full DINOv2 tokens for Affinity hint (ablation) | HoP · LingoQA | CIDEr: 66.0 -> 71.1 | Mixed |
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.
- ADAPT — ADAPT: Vision-Language Navigation With Modality-Aligned Action Prompts (CVPR 2022) · better on 1 of 1 shared result
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
- STSBench: A Spatio-temporal Scenario Benchmark for Multi-modal Large Language Models in Autonomous Driving (NeurIPS 2025 · similarity 0.87)
- SafeAuto: Knowledge-Enhanced Safe Autonomous Driving with Multimodal Foundation Models (ICML 2025 · similarity 0.87)
- DriveGPT4-V2: Harnessing Large Language Model Capabilities for Enhanced Closed-Loop Autonomous Driving (CVPR 2025 · similarity 0.86)
- NuScenes-QA: A Multi-Modal Visual Question Answering Benchmark for Autonomous Driving Scenario (AAAI 2024 · similarity 0.86)
- MAPLM: A Real-World Large-Scale Vision-Language Benchmark for Map and Traffic Scene Understanding (CVPR 2024 · similarity 0.86)
- Language Prompt for Autonomous Driving (AAAI 2025 · similarity 0.86)
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.
