K12-Bench - Neighbor Recommendation: leaderboard

The Neighbor Recommendation 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: pick the taxonomic and related neighbours of a concept. Exact match in percent (every option must be chosen correctly); higher is better.

Metric: Neighbor Recommendation 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: taxonomic and related neighbours of a concept (4,358 items); higher is better. Source: arxiv.org. Saturation forecast: Around January 2028. 10 models tracked. Updated 7 Oct 2026.

Top models

#ModelScoreOverall rank
1Gemini 3 Flash33.4#175
2Gemini 2.5 Flash15.4#463
3GLM-4.7 Flash15.2#1349
4Gemma 4 31B (IT)15#350
5Qwen 3 32B14.6#972
6Ministral-3-14B-Instruct14.5#1139
7GPT-5.213.1#185
8GPT-5 Mini12.5#364
9GPT-4o10.1#751
10Llama 3 8B Instruct3.8#4852

No result here: #1 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.

Interactive version: theaggregate.ai/benchmark?slug=k12-bench-neighbor-recommendation · How It Works · Data refreshed daily, snapshot 2026-11-21.