Obshazard-bench - Predictive Crisis Anticipation: leaderboard
The pre-disaster stage of Obshazard-bench (National University of Defense Technology, Shanghai AI Laboratory and Shanghai Jiao Tong University, 2026): risk detection, disaster type classification, and arrival time and initial duration prediction from raw Earth observation streams (1,524 samples). Task-aware soft score in percent, equal-weight aggregation; higher is better.
Metric: Normalized task score (%; x100 of the 0-1 task-aware soft score following the Earth AI protocol: categorical answers by semantic or lexical consistency, numeric and interval answers by distance to the ground truth; equal-weight aggregation across timesteps, subtasks, lifecycle stages and disaster categories; zero-shot, uniform prompts; raw satellite sounding streams with ground-station, disaster-record and socio-economic inputs; pre-disaster risk detection, disaster type classification, arrival time and initial duration prediction, 1,524 samples). Source: arxiv.org. Saturation forecast: Around August 2027. 4 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude Opus 4.8 | 46.97 | #56 |
| 2 | Kimi K2.6 | 37.56 | #173 |
| 3 | GPT-5.5 | 34.43 | #48 |
| 4 | Qwen 3.5 397B A17B | 30.84 | #242 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=obshazard-bench-predictive-crisis-anticipation · How It Works · Data refreshed daily, snapshot 2026-11-12.