MalruleLib - Same-Template: leaderboard
Misconception prediction within one problem template from MalruleLib (Chen, Liu and Sonkar, ACL 2026): shown one worked problem and only the student's wrong answer, the model predicts the student's answer on another problem from the same template. Exact-match accuracy in percent over 5,006 same-template pairs; higher is better.
Metric: Malrule Reasoning Accuracy (%; predict the student's next answer from one worked mistake; same template, answer only). Source: arxiv.org. Saturation forecast: Estimated already saturated. 9 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-OSS-120B | 77.1 | #699 |
| 2 | Qwen 3 Next 80B A3B (Thinking) | 74.2 | #591 |
| 3 | GPT-OSS-20B | 72.7 | #1150 |
| 4 | Qwen 3 Next 80B A3B Instruct | 69.9 | #763 |
| 5 | Qwen 3 4B | 61.1 | #2297 |
| 6 | Phi-4 | 50 | #1920 |
| 7 | Llama 3.3 70B Instruct | 47.7 | #1353 |
| 8 | Phi-4 Mini Instruct | 26.7 | #6452 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=malrulelib-same-template · How It Works · Data refreshed daily, snapshot 2026-11-12.