K12-Bench - Knowledge Grounding: leaderboard

The Knowledge Grounding task of K12-Bench (2026), built from the K12-KGraph knowledge graph of Chinese K-12 mathematics, physics, chemistry and biology textbooks, as four-option multi-select questions answered zero-shot: map an exercise to the concepts and skills it tests, or a concept or skill to its exercises. Exact match in percent (every option must be chosen correctly); higher is better.

Metric: Knowledge Grounding exact match (%, all-or-nothing over the selected option set) on four-option multi-select questions generated from the K12-KGraph curriculum knowledge graph of Chinese K-12 mathematics, physics, chemistry and biology textbooks; zero-shot, answer-only, temperature 0: map an exercise to the concepts and skills it tests, or a concept or skill to its exercises (3,850 items); higher is better. Source: arxiv.org. Saturation forecast: Around February 2027. 10 models tracked. Updated 7 Oct 2026.

Top models

#ModelScoreOverall rank
1Gemini 3 Flash63.4#180
2Gemini 2.5 Flash58#466
3GPT-5.250.8#190
4Gemma 4 31B (IT)50.6#354
5Qwen 3 32B46.7#983
6Ministral-3-14B-Instruct43#1151
7GLM-4.7 Flash36.4#1349
8GPT-5 Mini30.4#373
9GPT-4o30.3#754
10Llama 3 8B Instruct6.2#4858

No result here: #2 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.

Interactive version: theaggregate.ai/benchmark?slug=k12-bench-knowledge-grounding · How It Works · Data refreshed daily, snapshot 2026-11-22.