OmniStarPro-MIQ (Online): leaderboard
Multi-turn interactive question answering on OmniStarPro (Yang et al., 2026): a continuing user dialogue while the stream plays, which requires coreference resolution and a consistent persona. A GPT-4o judge scores the real-time answers for semantic correctness on a 0-10 scale; higher is better.
Metric: Semantic correctness (0-10; GPT-4o judge score of real-time answers to multi-turn interactive questions during the stream). Source: arxiv.org. Saturation forecast: Rough model projection: around 2026. 5 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | LiveStarPro | 5.81 | |
| 2 | LiveStar | 5.78 | #1076 |
| 3 | MMDuet | 5.62 | |
| 4 | VideoLLM-online | 3.83 | |
| 5 | VideoLLM-MoD | 3.75 |
Interactive version: theaggregate.ai/benchmark?slug=omnistarpro-miq-online · How It Works · Data refreshed daily, snapshot 2026-11-12.