System One Reranking - MMEB Visual Document - VisRAG SlideVQA: leaderboard
System One multimodal reranking:top16 candidates from Qwen3-VL-Embedding-2B,one-pass choice head;MMEBimage hit1,MMEBvisual-document nDCG5 and ViDoReV3 nDCG10. JEV27BVL is text-decision-tuned; the unmodifiedQwen3.8 control was served under a JEV endpoint alias but remains a distinct checkpoint. No thinking;temp1,topP1,topK0;supervisedQwen rerankers retain their own identity. Score (%).
Metric: Score (%). Source: arxiv.org. Saturation forecast: Around January 2027. 4 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Qwen 3.8 27B (Non-reasoning) | 97.83 | #111 (Qwen 3.8 27B) |
No result here: #4 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=system-one-reranking-mmeb-visual-document-visrag-slidevqa · How It Works · Data refreshed daily, snapshot 2026-10-22.