CoRAL Study - Trip-Planning: leaderboard
CoRAL (2026): main common evaluation of multimodal model outputs; LLaDA-V planner and Qwen2.5-VL executor are jointly trained as a distinct hybrid system. Caption quality uses GPT-4o/scene-graph judging; throughput is not a capability score. Accuracy (0–1).
Metric: Accuracy (0–1). Source: arxiv.org. Saturation forecast: Around December 2027. 11 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Qwen 2.5 VL 7B | 0.12 | #645 |
| 2 | InternVL3-8B | 0.1 | #622 |
| 3 | MiniCPM-V-2.6 | 0.05 | #846 |
No result here: #4 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=coral-study-trip-planning · How It Works · Data refreshed daily, snapshot 2026-10-22.