Research map / Vision-language and multimodal / Robot action and navigation
ColaVLA: Leveraging Cognitive Latent Reasoning for Hierarchical Parallel Trajectory Planning in Autonomous Driving
Qihang Peng, Xuesong Chen, Chenye Yang, Shaoshuai Shi, Hongsheng 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 latent reasoning (ablation) | ColaVLA · nuScenes ablation | Avg L2 (cm): 32.2 -> 31.3 | Improved |
| add the rethink stage on top of reasoning (ablation) | ColaVLA · nuScenes ablation | Avg L2 (cm): 31.3 -> 30.4 | Improved |
| use Hierarchical Parallel Planner instead of MLP planner (ablation) | ColaVLA · NeuroNCAP | NeuroNCAP Score (avg): 1.05 -> 1.50 | Improved |
| use Hierarchical Parallel Planner instead of diffusion planner (ablation) | ColaVLA · NeuroNCAP | NeuroNCAP Score (avg): 1.02 -> 1.50 | Improved |
| use Hierarchical Parallel Planner instead of diffusion planner (ablation) | ColaVLA · NeuroNCAP side collisions | NeuroNCAP Score (side): 1.43 -> 0.82 | Hurt |
| retain 256 tokens instead of 128 (ablation) | ColaVLA · nuScenes ablation | Avg L2 (cm): 31.2 -> 30.4 | Improved |
| retain 256 tokens instead of 320 (ablation) | ColaVLA · nuScenes ablation | Avg L2 (cm): 31.7 -> 30.4 | Improved |
| use Interpolate multi-scale strategy instead of single scale (ablation) | ColaVLA · nuScenes ablation | Avg L2 (cm): 31.9 -> 30.4 | Improved |
| use Interpolate instead of Sequential multi-scale (ablation) | ColaVLA · nuScenes ablation | Avg L2 (cm): 31.8 -> 30.4 | Improved |
| use Interpolate instead of Reverse multi-scale (ablation) | ColaVLA · nuScenes ablation | Avg L2 (cm): 31.4 -> 30.4 | Improved |
| use single scale instead of Interpolate for 1s horizon (ablation) | ColaVLA · nuScenes ablation | L2 1s (cm): 13.8 -> 14.0 | Hurt |
| do latent reasoning instead of text chain-of-thought | ColaVLA · Inference latency | latency (ms per frame): 3727 -> 727 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 0 of 1 on most shared results.
- EMMA+ — EMMA: Empowering Multi-modal Mamba with Structural and Hierarchical Alignment (ICLR 2025) · better on 0 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
- OneVL: One-Step Latent Reasoning and Planning with Vision-Language Explanation (NeurIPS 2026 · similarity 0.94)
- Drive-R1: Bridging Reasoning and Planning in VLMs for Autonomous Driving with Reinforcement Learning (AAAI 2026 · similarity 0.93)
- LaST-VLA: Thinking in Latent Spatio-Temporal Space for Vision-Language-Action in Autonomous Driving (NeurIPS 2026 · similarity 0.92)
- AutoVLA: A Vision-Language-Action Model for End-to-End Autonomous Driving with Adaptive Reasoning and Reinforcement Fine-Tuning (NeurIPS 2025 · similarity 0.92)
- SOLVE: Synergy of Language-Vision and End-to-End Networks for Autonomous Driving (CVPR 2025 · similarity 0.92)
- SGDrive: Scene-to-Goal Hierarchical World Cognition for Autonomous Driving (CVPR 2026 · similarity 0.92)
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.
