Finetuning with Scientific Data Increases Hall: leaderboard
Large language models (LLMs) are increasingly used to communicate and explain scientific concepts, yet their tendency to hallucinate poses significant risks in this high stakes use-case. Prior hallucination evaluation work remains largely restricted to the biomedical domain, treats hallucination as a binary task, and has not examined the growing family of scientifically fine-tuned LLMs. We address
Metric: OFS (self-reported). Source: benchmarklist.com. Saturation forecast: Estimated already saturated. 18 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Qwen 3 8B | 67.1 | #674 |
| 2 | Qwen 2.5 14B Instruct | 66.7 | #647 |
| 3 | GPT-4o Mini | 66.1 | #599 |
| 4 | Qwen 3 32B | 66 | #443 |
| 5 | Qwen 2.5 7B Instruct | 65.2 | #862 |
| 6 | Llama 3.1 8B Instruct | 65 | #1035 |
| 7 | tulu-2-7B | 62.7 | #1429 |
| 8 | Llama 2 70B Chat | 62.2 | #1237 |
| 9 | Llama 2 7B Chat | 61.4 | #1499 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=finetuning-with-scientific-data-increases-hall · How It Works · Data refreshed daily, snapshot 2026-10-20.