FoundationalASSIST - Most Common Distractor: leaderboard
FoundationalASSIST (Worden et al., 2026) is 1.7 million ASSISTments mathematics interactions of 5,000 students with question text, the students' actual answers and skill tags. For 234 multiple-choice problems the model predicts which wrong option students chose most often (zero-shot); accuracy in percent against a random baseline of 35.8, higher is better.
Metric: Accuracy (%; which wrong option students chose most often, 234 multiple-choice problems with at least 10 responses; random 35.8). Source: arxiv.org. Saturation forecast: Around April 2027. 4 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Qwen 3 Next 80B A3B Instruct | 47.9 | #763 |
| 2 | Llama 3.3 70B Instruct | 43.6 | #1353 |
| 3 | Qwen 3 Next 80B A3B (Thinking) | 40.2 | #591 |
| 4 | GPT-OSS-120B | 35.5 | #699 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=foundationalassist-most-common-distractor · How It Works · Data refreshed daily, snapshot 2026-11-12.