LongSCOP Retrieval - RoMa - Hit@3: leaderboard
LongSCOP v1 referenceframeranking,52LVOcases,4queryframes,candidatesfulltimeline,filterbottom30%registration. RoMa variant7 suppliesranks/scores/registrationvisualizations topretrainedVLM; nativeHit@30–1. Reference relevance definedbyobjectivegenerationmetrics; exact combination/perqueryinventorysourceunreported. No videogeneratedquality attribution or guessed aggregate.
Metric: Hit@3 (0–1). Source: arxiv.org. Saturation forecast: Rough model projection: around 2026. 2 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Qwen3.5-27B (LongSCOP Retrieval Run: registration=RoMa;variant=7) | 0.65 | |
| 2 | Qwen3.5-9B (LongSCOP Retrieval Run: registration=RoMa;variant=7) | 0.65 |
Interactive version: theaggregate.ai/benchmark?slug=longscop-retrieval-roma-hit-3 · How It Works · Data refreshed daily, snapshot 2026-11-12.