AI Paper Atlas

Research map / Label-efficient and robust learning / Recognition and long-tail

VADv2: End-to-End Autonomous Driving via Probabilistic Planning

Bo Jiang, Shaoyu Chen, Hao Gao, Bencheng Liao, Qian Zhang, Wenyu Liu, Xinggang Wang

ICLR 2026 · Label-efficient and robust learning › Recognition and long-tail · approach: Dense features and foundation priors

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
use probabilistic instead of deterministic planning (ablation)VADv2 · Town05 LongDriving Score: 74.6 -> 85.1Improved
use probabilistic instead of deterministic planning (ablation)VADv2 · Town05 Long (open-loop L2@3s)L2 error: 0.223 -> 0.225No clear effect
use probabilistic instead of deterministic planning, high traffic (ablation)VADv2 · NAVSIM (high traffic density)PDMS: 85.8 -> 87.7Improved
add distribution loss (ablation)VADv2 · open-loop 50k clipsL2 3s (m): 3.153 -> 0.290Improved
add conflict loss (ablation)VADv2 · open-loop 50k clipsCollision 3s (%): 0.039 vs 0.039 (L2 0.291 -> 0.290)Improved
add agent token (ablation)VADv2 · open-loop 50k clipsL2 3s (m): 0.327 -> 0.290Improved
add map token (ablation)VADv2 · open-loop 50k clipsL2 3s (m): 0.332 -> 0.290Improved
add image token (ablation)VADv2 · open-loop 50k clipsL2 3s (m): 0.293 -> 0.290No clear effect
increase vocabulary size from 256 to 4096 (ablation)VADv2 · open-loop L2 3sL2 (m): 0.337 -> 0.290Improved
increase training clips from 1x10^5 to 3x10^6 (ablation)VADv2 · open-loop L2 3sL2 (m): 0.461 -> 0.225Improved
use farthest trajectory sampling (FTS) instead of k-means for vocabulary (ablation)VADv2 · NAVSIM navtestPDMS: 89.0 -> 89.3Improved
select 5th instead of top-1 probability trajectory (ablation)VADv2 · NAVSIM navtestPDMS: 89.3 -> 87.5Hurt

Later papers that compared with this method

4 later accepted papers report VADv2 as a baseline in their results tables; 4 of them report a better number on most of the results they share with it.

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 2 on most shared results.

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

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