LithoBench - Mechanism Explanation (Open-Ended): leaderboard
LithoBench Mechanism Explanation (Open-Ended) scores vision-language models on 1,200 open-ended questions about causal explanations linking visual evidence to geological processes in GF-2 satellite image patches (LithoBench, 2026). The OPEN_J score, from 0 to 10 and higher is better, blends ROUGE-L and BLEU-4 overlap with the reference answer and a GPT-5.4 judge rating.
Metric: OPEN_J score (0-10) = 0.4 x ROUGE-L x 10 + 0.2 x BLEU-4 x 10 + 0.4 x the GPT-5.4 judge's overall score (0-10), comparing the answer with the reference answer, on the 1,200 open-ended mechanism explanation questions of LithoBench (causal explanations linking visual evidence to geological processes in GF-2 remote-sensing image patches); temperature 0, 1,024 output tokens; higher is better. Source: arxiv.org. Saturation forecast: Around 2031. 11 models tracked. Updated 7 Oct 2026.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Qwen 3.5 397B A17B | 4.28 | #264 |
| 2 | Kimi K2.5 | 4.24 | #266 |
| 3 | InternVL3.5-241B-A28B | 4.12 | #733 |
| 4 | GLM-4.6V | 4.11 | #710 |
| 5 | Intern-S1-Pro | 4.1 | #585 |
| 6 | Grok 4 | 4.09 | #319 |
| 7 | Gemini 3 Pro | 4.09 | #145 |
| 8 | InternVL3.5-8B-HF | 3.96 | |
| 9 | Qwen 3.5 9B | 3.59 | #828 |
| 10 | Qwen 3.5 27B | 3.39 | #375 |
| 11 | Claude Sonnet 4.6 | 2.26 | #173 |
No result here: #1 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.
Interactive version: theaggregate.ai/benchmark?slug=lithobench-mechanism-explanation-open-ended · How It Works · Data refreshed daily, snapshot 2026-11-21.