ToxReason - Pathway Alignment: leaderboard
ToxReason (2026) tests mechanistic chemical toxicity reasoning: given a query molecule with experimental initiating-event evidence from structurally similar compounds, the model predicts liver, heart and kidney toxicity and explains how molecular initiating events lead to the adverse outcome along an Adverse Outcome Pathway. Needleman-Wunsch alignment between the reasoning steps of the model and the reference pathway steps, from 0 to 1; higher is better.
Metric: Needleman-Wunsch alignment score (0-1) between the model's reasoning steps and the reference adverse outcome pathway steps, combining semantic similarity and causal order, on the ToxReason test set (query chemicals with curated human CTD toxicity labels and adverse outcome pathway context, eight retrieved activation and inhibition examples from similar compounds), zero-shot, temperature 0; higher is better. Source: arxiv.org. Saturation forecast: Around 2031. 11 models tracked. Updated 7 Oct 2026.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5.1 | 0.513 | #240 |
| 2 | Gemma 3 27B (IT) | 0.498 | #1324 |
| 3 | GPT-4o | 0.494 | #751 |
| 4 | O3 | 0.488 | #237 |
| 5 | O4 Mini | 0.487 | #368 |
| 6 | Qwen 2.5 14B Instruct | 0.484 | #1775 |
| 7 | Qwen 3 4B Instruct | 0.484 | #1696 |
| 8 | GPT-5 | 0.475 | #178 |
| 9 | Llama 3.1 70B Instruct | 0.473 | #1571 |
| 10 | DeepSeek R1 Distill Llama 70B | 0.458 | #1669 |
| 11 | Llama 3.1 8B Instruct | 0.396 | #4600 |
No result here: #1 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.
Interactive version: theaggregate.ai/benchmark?slug=toxreason-pathway-alignment · How It Works · Data refreshed daily, snapshot 2026-11-21.