EO-Jev - NWPU-RESISC45: leaderboard
EO-Jev Table 1 native SVG:30 EO settings plus general Decision Index retention. Identical closed-form questions and image sets across d 3-flash,Qwen 3.5-9B,EO-Jev and individually named specialist; SAR excluded,training/evaluation images deduplicated. Accuracy except exact-set multi-select,LHRS strict accuracy and deceptive-question correct abstention. d 3-flash andQwen use answer-code readout; specialists use source-documented answer-code or greedy generation.18 in-domain/12 out-of-domain settings retain different source cohorts. Missing specialist reference omitted only where native cell blank; per-row best specialist identified, never merged as one model. Metric: Accuracy (%).
Metric: Accuracy (%). Source: arxiv.org. Saturation forecast: Around February 2027. 4 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Qwen 3.5 9B | 69.8 | #391 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=eo-jev-nwpu-resisc45 · How It Works · Data refreshed daily, snapshot 2026-11-05.