CRISP: A Spatiotemporal Camera-Radar Backbone for Driving via Forecasting-Based World-Model Pretraining

arXiv 2026 35.4 method

TLDR

CRISP pretrains a camera-radar backbone for driving by forecasting future LiDAR, improving downstream tasks without LiDAR at inference.

Reasoning

Strengths include a novel forecasting-based pretraining method that leverages privileged LiDAR supervision, practical sensor configuration, and strong downstream task improvements. Weaknesses are reliance on LiDAR during pretraining and evaluation on a single dataset (nuScenes), limiting generalizability claims.

Read-first score

Read-first score 35.4, weighted from topical fit, citation, graph, method, reproducibility, and recency signals. Original total remains 30.

Recency 6%

100

Uses a gentle age decay so recent papers surface without erasing older foundations. 2026

Methodology quality 18%

Screens visible abstract and analysis fields for experiment, dataset, baseline, metric, and limitation evidence. markers=experiment

Reproducibility 18%

Screens links and visible text for paper, code, dataset, artifact, and repository signals. pdf=True; code=False; dataset=False; markers=code,github

Topical relevance 29%

42.9

Uses existing LLM keyword relevance scores normalized to 0-100. world model,world simulator,generative world model,interactive world model,video world model,world dynamics prediction,model-based reinforcement learning world model

Citation impact 18%

Uses OpenAlex-shaped citation metadata as a bibliometric attention signal, separate from paper quality. cited_by_count=0

Citation velocity 12%

Citation velocity estimates citations per publication-year to reduce old-paper bias. velocity=0.00

Field roles

Frontier

Rank sensitivity

Stability: volatile; rank range: 136.

Keyword Scores

world model

world dynamics prediction

generative world model

world simulator

video world model

interactive world model

model-based reinforcement learning world model

Deep Analysis

Innovations

Forecasting-based pretraining for camera-radar fusion using future LiDAR point clouds as privileged supervision
Enhanced radar encoder, radar-enhanced temporal self-attention, and multimodal feature rendering with modality innovation gating
Unified BEV representation learning from camera and radar without requiring LiDAR at deployment

Methodology

CRISP pretrains a spatiotemporal backbone by predicting future LiDAR point clouds from historical multi-view camera images and radar sweeps, using LiDAR only as privileged supervision. It incorporates an enhanced radar encoder, radar-enhanced temporal self-attention, and modality innovation gating to fuse camera and radar features into a bird's-eye-view representation.

Key Results

On nuScenes, CRISP improves long-horizon point cloud forecasting and transfers effectively to downstream tasks including 3D detection, tracking, online mapping, motion forecasting, future occupancy prediction, and planning.

Resources

Paper