AI Paper Atlas

Research map / Architectures and efficiency / Attention and transformers

Efficient Equivariant Transformer for Self-Driving Agent Modeling

Scott Xu, Dian Chen, Kelvin Wong, Chris Zhang, Kion Fallah, Raquel Urtasun

CVPR 2026 · Architectures and efficiency › Attention and transformers · approach: Attention head and block redesign · 1 citations (OpenAlex)

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 changedWhereMeasuredEffect
add distance-awareness features to attention queries and keys (ablation)DriveGATr 3M · WOSAC 2% validation splitRMM: 0.7408 -> 0.7478Improved
add invariant adapter after each attention block (ablation)DriveGATr 3M · WOSAC 2% validation splitminADE: 1.6053 -> 1.5798Improved
add invariant adapter after each attention block (ablation)DriveGATr 3M · WOSAC 2% validation splitRMM: 0.7483 -> 0.7478No clear effect
let each agent attend to 8 nearest map tokens instead of 4 (ablation)DriveGATr 3M · WOSAC 2% validation splitRMM: 0.7478 -> 0.7528Improved
let each agent attend to the whole map instead of 8 nearest tokens (ablation)DriveGATr 3M · WOSAC 2% validation splitRMM: 0.7528 -> 0.7617Improved
use DriveGATr equivariance instead of explicit RPE3M transformer agent model · WOSAC 2024 validationRMM: 0.7251 -> 0.7620Improved
use DriveGATr instead of data-augmented non-equivariant Transformer3M transformer agent model · WOSAC 2024 validationminADE: 1.6720 -> 1.4192Improved
use rotary DRoPE instead of full SE(2) equivariance3M transformer agent model · WOSAC 2024 validationRMM: 0.7206 vs 0.7620 for DriveGATr-3MHurt
scale DriveGATr from 3M to 30M parametersDriveGATr · WOSAC 2024 validationRMM: 0.7620 -> 0.7636Improved
add explicit pairwise relative positional encodings to Transformer3M scene-centric Transformer · WOSAC 2024 validationRMM: 0.7257 -> 0.7251No clear effect
add closed-loop fine-tuning (CAT-K) to SMARTSMART-7M · WOSAC 2024 validationRMM: 0.7678 -> 0.7709Improved

Closest papers by meaning

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.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy