Driving · 2026-10-19
STRIDE: Evaluating Spatiotemporal Reasoning in Driving Edge Cases
Lieqi Liu, Rui gao, Jia-Chen Gu, Wenbo Hu, Zhaobin Mo, Ahmadreza Moradipari, Nejib Ammar, Wei Wang, Nanyun Peng
AI-assisted summary based on the paper. Consult the linked source for methods and limitations. Benchmark numbers appear after editor verification.
TL;DR
STRIDE is an object-centric driving benchmark for evaluating spatiotemporal reasoning in long-tail, safety-critical edge cases. It tests six task families (spatial/temporal perception & understanding, memory, extrapolation, trajectory prediction, and scene-context awareness) using linked multiple-choice and free-form visual question answering. Across 12 models, the best multiple-choice accuracy is 43.4% (GPT-6-Astra) versus ~20% random-choice baseline.
Why it matters
The authors argue that edge-case evaluation for driving must connect past spatial observations to future consequences, not just assess perception or simple temporal patterns. STRIDE provides fine-grained, capability-level diagnostics via task-family separation, showing that strong scene description VQA performance can coexist with weak safety-critical judgments and that reconstructing object motion is harder than recalling past positions.
Method
- Introduce STRIDE, an object-centric benchmark built from multi-view temporal driving data, with six task families spanning spatial and temporal reasoning.
- Use linked evaluation formats: multiple-choice questions (MCQ) and free-form visual question answering (OEQ) that probe the same capabilities through intermediate judgments and driving explanations.
- Curate long-tail/safety-critical scenes (unusual object behaviors and complex traffic interactions) and generate reference answers from structured annotations with automated procedures plus limited language-model assistance and human verification.
Limitation
The paper provides no explicit limitations section; the closest stated constraint is that STRIDE evaluates vision-language models under a unified evaluation protocol with deterministic or near-deterministic decoding (unless otherwise stated).
Abstract (from arXiv)
In autonomous driving, perceiving surrounding objects is not enough: failing to understand how they move and interact over time can lead to dangerous decisions. Evaluating vision-language models (VLMs) therefore requires testing whether they connect past spatial observations to future consequences. Yet existing benchmarks often assess edge cases through basic perception questions or test temporal understanding in simple scenarios, leaving complex spatiotemporal reasoning underexplored. We introduce STRIDE (SpatioTemporal Reasoning In Driving Edge Cases), an object-centric benchmark for long-tail, safety-critical driving scenarios. STRIDE covers six task families: spatial perception, spatial understanding, temporal memory, temporal extrapolation, trajectory prediction, and scene-context awareness. It pairs multiple-choice questions probing individual spatiotemporal judgments with related free-form visual question answering (VQA) requiring detailed scene explanations and justifications of driving decisions. We prioritize scenes with unusual object behaviors and complex traffic interactions, generate reference answers from structured dataset annotations through automated procedures, and use limited language-model assistance and human verification to ensure annotation quality. Across 12 proprietary, open-weight, and driving-specialized models, even the best performer, GPT-6-Astra, achieves only 43.4% average multiple-choice accuracy, versus an approximately 20% random-choice baseline. Further analysis reveals that strong performance in describing driving scenes does not necessarily translate to accurate safety-critical judgments, and that models struggle more with reconstructing object motion than recalling past positions. STRIDE thus exposes spatiotemporal driving limitations that aggregate VQA scores can conceal.
Related papers
- Explicit Geometric Chain-of-Thought for Vision-Language-Action in Autonomous Driving
- Think Densely, Act Sparsely: Latent Expert Cognitive Chains for Vision-Language-Action Autonomous Driving
- Mind the Temporal Gap: Evaluating Video Models for Temporal Understanding in Driving
- DriveWorld3D Evidence Constrained 3D World Modeling and Generative Simulation for Real-World Autonomous Driving
- doPlan: A Variable-Horizon Dataset for Multi-Stage Language-Conditioned Planning in Autonomous Driving
Publication record
Bibliographic evidence describes where and when a record appeared. It does not validate scientific claims.
First public release: 2026-10-16 · Indexed by Pub-AI: 2026-10-19 UTC
STRIDE: Evaluating Spatiotemporal Reasoning in Driving Edge Cases
Preprint posted: 2026-10-16 · Original record ↗
- arxiv
- 2610.22795
Explicit source versions (1)
- v1 · 2026-10-16
- Preprint available
Source date: 2026-10-16 · First detected: 2026-10-25 UTC · Historical baseline
Source evidence (1)
- Official evidence ↗ · atom:published (official metadata)
Abstract · arxiv
In autonomous driving, perceiving surrounding objects is not enough: failing to understand how they move and interact over time can lead to dangerous decisions. Evaluating vision-language models (VLMs) therefore requires testing whether they connect past spatial observations to future consequences. Yet existing benchmarks often assess edge cases through basic perception questions or test temporal understanding in simple scenarios, leaving complex spatiotemporal reasoning underexplored. We introduce STRIDE (SpatioTemporal Reasoning In Driving Edge Cases), an object-centric benchmark for long-tail, safety-critical driving scenarios. STRIDE covers six task families: spatial perception, spatial understanding, temporal memory, temporal extrapolation, trajectory prediction, and scene-context awareness. It pairs multiple-choice questions probing individual spatiotemporal judgments with related free-form visual question answering (VQA) requiring detailed scene explanations and justifications of driving decisions. We prioritize scenes with unusual object behaviors and complex traffic interactions, generate reference answers from structured dataset annotations through automated procedures, and use limited language-model assistance and human verification to ensure annotation quality. Across 12 proprietary, open-weight, and driving-specialized models, even the best performer, GPT-6-Astra, achieves only 43.4% average multiple-choice accuracy, versus an approximately 20% random-choice baseline. Further analysis reveals that strong performance in describing driving scenes does not necessarily translate to accurate safety-critical judgments, and that models struggle more with reconstructing object motion than recalling past positions. STRIDE thus exposes spatiotemporal driving limitations that aggregate VQA scores can conceal.