OVEarth-Bench: A New Benchmark for Open-Vocabulary Earth Observation
earth-insights · hf · 2026-07-31
Existing open-vocabulary Earth observation (EO) evaluations are often limited by narrow category vocabularies or query forms. To address this, researchers introduced OVEarth-Bench, extending evaluation in two directions: category breadth via broad hierarchical coverage with positive/negative expressions, and query diversity via vocabulary, referring, and reasoning queries.
The benchmark supports mask and box localization under a unified zero-shot protocol. Evaluations reveal that current methods remain limited; MLLM-based methods achieve the strongest overall performance; and EO-specific methods generally underperform general models.
More from Research
- Meta's ROCS Paradigm Boosts Recommendation Retrieval QPS Up to 3x — _reachsumit · 2026-07-31
- With Mandatory Reviewing, Low-Quality Reviews in AI Conferences Are No Longer Justifiable — Kwangryeol · 2026-07-31
- HiLaR: Optimizing LLM Recommendation Reasoning via Hierarchical RL — _reachsumit · 2026-07-31
- Kuaishou's Feedback-Driven Framework Distills LLM Policies for LLM-Free Serving — _reachsumit · 2026-07-31
- Tencent Proposes CCFormer: Efficient Long-Sequence Modeling for Industrial Recommenders — _reachsumit · 2026-07-31
- H3-Omni Architecture: Native 2K Resolution and In-Context Regeneration — bdsqlsz · 2026-07-31