MHGraphBench - Fact Checking: leaderboard
Balanced yes or no questions on whether a triple is supported by the subgraph in MHGraphBench (2026), a knowledge-graph-grounded mental-health benchmark built from a curated PrimeKG subgraph, with letter-only multiple-choice answers. Accuracy in percent; higher is better.
Metric: Accuracy (%) on the 4,000 balanced yes/no triple fact checking items, letter-only answers; the OpenAI API models answer at temperature 0 (at most 120 completion tokens) with strict answer-letter parsing, the open models are scored by forced-choice option-letter log-probabilities; higher is better. Source: arxiv.org. Saturation forecast: Around January 2028. 15 models tracked. Updated 7 Oct 2026.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5 Mini | 64.35 | #364 |
| 2 | GPT-4o | 63.45 | #751 |
| 3 | GPT-5.2 Instant | 63.35 | #304 |
| 4 | GPT-4.1 | 63.32 | #505 |
| 5 | GPT-5.1 Instant | 62.5 | #472 |
| 6 | Qwen 2.5 32B Instruct | 58.4 | #1247 |
| 7 | Qwen 2.5 7B Instruct | 53.18 | #2723 |
| 8 | Mistral 7B Instruct (v0.3) | 52.25 | #6750 |
| 9 | Llama3-Med42-8B | 50.3 | #5084 |
| 10 | Llama3-OpenBioLLM-8B | 50.25 | #6364 |
| 11 | Meditron-7B | 50 | #8733 |
| 12 | BioMistral-7B | 49.9 | #8332 |
| 13 | DeepSeek-R1-Distill-Qwen-7B | 49.6 | #8055 |
| 14 | DeepSeek R1 Distill Qwen 32B | 48.15 | #1834 |
| 15 | Llama 3.1 8B Instruct | 45.32 | #4600 |
No result here: #1 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.
Interactive version: theaggregate.ai/benchmark?slug=mhgraphbench-fact-checking · How It Works · Data refreshed daily, snapshot 2026-11-21.