LongVidSearch - Causal Inference (4-Hop): leaderboard
Metric: Accuracy (%) on the 144 Causal Inference questions that need exactly 4 evidence clips, judged against the reference by exact match or by majority vote of three LLM judges (GPT-5, Gemini 3 Pro, GPT-4o); the backbone drives a fixed VideoAgent-style framework that retrieves clips and reads their captions through a standard tool interface; higher is better. Source: arxiv.org. Saturation forecast: Around May 2027. 10 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5 | 39.58 | #91 |
| 2 | Gemini 3 Pro | 17.36 | #77 |
| 3 | Qwen 2.5 VL 72B Instruct | 15.97 | #364 |
| 4 | GPT-4o | 11.81 | #333 |
| 5 | Qwen 2.5 VL 7B Instruct | 2.82 | #643 |
No result here: #3 Claude Opus 5.5, #5 GPT-6 Astra, #8 Claude Fable 5.1.
Interactive version: theaggregate.ai/benchmark?slug=longvidsearch-causal-inference-4-hop · How It Works · Data refreshed daily, snapshot 2026-10-11.