# LiDAR-guided token selection cuts driving model compute

> A paper proposes using LiDAR and radar geometry to focus an autonomous-driving model on fewer image patches, reporting lower compute use and better benchmark results.

Oossa · 2026-11-02 · https://oossa.com/en/lidar-guided-token-selection-cuts-driving-model-compute

A new arXiv paper proposes Geometry-Aware Token Selection (GATS), which uses LiDAR and radar measurements to choose image patches before an autonomous-driving model processes them. The authors report that at a 10% token budget, GATS used 24 GFLOPs—18 times fewer than the dense baseline—while scoring better on nuScenes and NAVSIM. On 739 intent-critical scenarios, they report a 0.09% collision rate, compared with 0.34% for the dense model. These are results reported in the paper; the supplied abstract does not establish real-world driving performance.

## The facts

- At a 10% token budget, GATS reported 0.537 m nuScenes L2 error versus 0.573 m for the dense baseline.
- The paper was published on arXiv on November 2, 2026.

## Why it matters

For autonomous-driving researchers, the reported compute reduction suggests a way to process less camera data without losing benchmark performance, but its effect on real roads remains unknown.

## Sources & references

1. [Let LiDAR Decide Where to Look: Geometry-Measured Token Selection for Efficient Sparse Latent World Models](https://arxiv.org/abs/2610.40935) – arXiv cs.CV, 2026-11-02

Last updated: 2026-11-02