# Metadata-Aware Multi-Prompt Reasoning for Zero-Shot Accident Understanding

Source: [Metadata-Aware Multi-Prompt Reasoning for Zero-Shot Accident Understanding](<https://www.sota2.com/research/paper/6a2a4144d33dfb23e783947e>)

Collection index: [llms.txt](<https://www.sota2.com/research/llms.txt>)

Data updated: 2026-06-12T05:00:14.585Z

Authors: Tarandeep Singh, Soumyanetra Pal, Soham Biswas, Nishanth Chandran

Published: 2026-06-10T13:12:40.000Z

## Abstract

In this paper, we address the problem of zero-shot understanding of accidents from surveillance videos by identifying when an impact event occurs, what type of impact it is, and where in the frame it occurs using natural language. We propose a three-stage pipeline that decomposes the accident understanding into when, what, and where. The first stage extracts a short temporal window around the impact using vision-language similarity. In the second stage, we perform metadata-driven multi-prompt reasoning with five complementary views (baseline, motion, geometry, contrast, and tiebreaker) and resolve disagreement via an entropy-gated pairwise adjudicator. Finally, we localize the impact of an open-vocabulary detector queried on the predicted accident type and scene layout, and aggregate detections across keyframes using a score-weighted centroid. Our pipeline achieves a substantial improvement in the harmonic-mean score over a centre-of-frame baseline on the zero-shot ACCIDENT @ CVPR benchmark. We show that decomposing zero-shot video understanding into temporal localization, semantic classification, and spatial grounding enable more reliable reasoning with vision-language models than direct prompting alone.

## Recorded benchmark results

A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage.

- [Accident Understanding on ACCIDENT@CVPR 2026 (Public)](<https://www.sota2.com/research/sota/accident-understanding-on-accident-at-cvpr-2026-public.md>): Accident Understanding; ACCIDENT@CVPR 2026 (Public); LB Score: 38.52.

- [Traffic Accident Understanding on ACCIDENT Leaderboard (private)](<https://www.sota2.com/research/sota/traffic-accident-understanding-on-accident-leaderboard-private.md>): Traffic Accident Understanding; ACCIDENT Leaderboard (private); C: 50.57.

## Original sources and implementation

[Original paper](<https://arxiv.org/abs/2606.12047>)

[PDF](<https://arxiv.org/pdf/2606.12047.pdf>)