ConceptKT (Same-Concept Only): leaderboard
Metric: Missing-concept macro-F1 (%) over the concept labels, history selected by keeping only prior responses that share a concept with the target question (Same-Concept Only), ConceptKT's 405 held-out math questions from six students' chronological MathEDU solution records (the first 90% of each student's records are the history), concept labels over 55 Common Core concepts annotated by three mathematics-education experts; the model reads the student's prior solution records with correctness labels in context (chain-of-thought prompt, temperature 0) and predicts the concepts the student will be missing on the target question; higher is better. Source: arxiv.org. Saturation forecast: Estimated already saturated. 4 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | DeepSeek R1 | 17.4 | #245 |
| 2 | O3 Mini | 10.9 | #266 |
| 3 | Llama 4 Maverick Instruct | 10.13 | #439 |
No result here: #3 Claude Opus 5.5, #5 GPT-6 Astra, #8 Claude Fable 5.1.
Interactive version: theaggregate.ai/benchmark?slug=conceptkt-same-concept-only · How It Works · Data refreshed daily, snapshot 2026-10-11.