K12-KGraph: A Curriculum-Aligned Knowledge Graph for Benchmarking and Training Educational LLMs

Hugging Face Daily Papers Papers
knowledge-graph education benchmark llm k12 curriculum multimodal

Summary

Introduces K12-KGraph, a curriculum-aligned knowledge graph from Chinese textbooks, along with a benchmark (K12-Bench) and training data (K12-Train) for evaluating and improving LLMs' curriculum cognition in K-12 education.

Large language models are increasingly used in K-12 education, but existing benchmarks mainly test exam question answering rather than understanding how curriculum knowledge is structured and visually presented. We call this capability curriculum cognition. It covers prerequisite chains, concept taxonomies, experiment-concept links, pedagogical sequencing, and visual grounding. We introduce K12-KGraph, a curriculum-aligned knowledge graph extracted from official People's Education Press textbooks in mathematics, physics, chemistry, and biology across primary, middle, and high school. It contains nine node types and fourteen relation types covering curriculum structure and visual grounding. From this graph, we derive K12-Bench, a 23,640-question multi-select benchmark with five task families: Ground, Prereq, Neighbor, Evidence, and Locate. We also build K12-Train, a graph-guided supervised fine-tuning corpus of 7,335 samples, including 2,267 text-only QA pairs and 5,068 multimodal VQA pairs. On K12-Bench, Gemini-3-Flash achieves only 57 percent exact match and Gemma-4-31B-IT reaches 46 percent, with Prereq and Neighbor being the hardest tasks. Our training experiments show that domain-specific supervision can reduce this gap. Under a matched 2,300-sample budget, K12-Train-Text consistently outperforms equally sized subsets of eight mainstream instruction-tuning corpora on GaokaoBench and EduEval. For vision-language models, K12-Train-Full achieves the best overall results on Gaokao-MM, MDK12-medium, and K12Vista among all compared training configurations, despite using fewer samples than the full DataFlow and WizardLM baselines. It also surpasses both text-only and multimodal-only variants, showing that textual and visual supervision are complementary. We release the graph, benchmark, training data, and complete construction pipeline.
Original Article
View Cached Full Text

Cached at: 07/24/26, 09:00 AM

Paper page - K12-KGraph: A Curriculum-Aligned Knowledge Graph for Benchmarking and Training Educational LLMs

Source: https://huggingface.co/papers/2605.09635

Abstract

LargelanguagemodelsareincreasinglyusedinK-12education,butexistingbenchmarksmainlytestexamquestionansweringratherthanunderstandinghowcurriculumknowledgeisstructuredandvisuallypresented.Wecallthiscapabilitycurriculumcognition.Itcoversprerequisitechains,concepttaxonomies,experiment-conceptlinks,pedagogicalsequencing,andvisualgrounding.WeintroduceK12-KGraph,acurriculum-alignedknowledgegraphextractedfromofficialPeople’sEducationPresstextbooksinmathematics,physics,chemistry,andbiologyacrossprimary,middle,andhighschool.Itcontainsninenodetypesandfourteenrelationtypescoveringcurriculumstructureandvisualgrounding.Fromthisgraph,wederiveK12-Bench,a23,640-questionmulti-selectbenchmarkwithfivetaskfamilies:Ground,Prereq,Neighbor,Evidence,andLocate.WealsobuildK12-Train,agraph-guidedsupervisedfine-tuningcorpusof7,335samples,including2,267text-onlyQApairsand5,068multimodalVQApairs.OnK12-Bench,Gemini-3-Flashachievesonly57percentexactmatchandGemma-4-31B-ITreaches46percent,withPrereqandNeighborbeingthehardesttasks.Ourtrainingexperimentsshowthatdomain-specificsupervisioncanreducethisgap.Underamatched2,300-samplebudget,K12-Train-Textconsistentlyoutperformsequallysizedsubsetsofeightmainstreaminstruction-tuningcorporaonGaokaoBenchandEduEval.Forvision-languagemodels,K12-Train-FullachievesthebestoverallresultsonGaokao-MM,MDK12-medium,andK12Vistaamongallcomparedtrainingconfigurations,despiteusingfewersamplesthanthefullDataFlowandWizardLMbaselines.Italsosurpassesbothtext-onlyandmultimodal-onlyvariants,showingthattextualandvisualsupervisionarecomplementary.Wereleasethegraph,benchmark,trainingdata,andcompleteconstructionpipeline.

View arXiv pageView PDFProject pageGitHub38Add to collection

Get this paper in your agent:

hf papers read 2605\.09635

Don’t have the latest CLI?curl \-LsSf https://hf\.co/cli/install\.sh \| bash

Models citing this paper0

No model linking this paper

Cite arxiv.org/abs/2605.09635 in a model README.md to link it from this page.

Datasets citing this paper0

No dataset linking this paper

Cite arxiv.org/abs/2605.09635 in a dataset README.md to link it from this page.

Spaces citing this paper0

No Space linking this paper

Cite arxiv.org/abs/2605.09635 in a Space README.md to link it from this page.

Collections including this paper0

No Collection including this paper

Add this paper to acollectionto link it from this page.

Similar Articles

BLINKG: A Benchmark for LLM-Integrated Knowledge Graph Generation

arXiv cs.AI

BLINKG is a benchmark designed to evaluate the mapping capabilities of Large Language Models (LLMs) in constructing Knowledge Graphs from heterogeneous data sources. It provides a standardized framework to assess how effectively LLMs establish correspondences between data schemas and ontology concepts.

GTBench: A Curriculum-Grounded Benchmark for Evaluating LLMs as Mathematical Research Assistants in Graph Theory

arXiv cs.AI

The paper introduces GTBench, a curriculum-grounded benchmark for evaluating LLMs as mathematical research assistants in graph theory, containing 63 problems across three difficulty levels. It evaluates five frontier models and finds that performance degrades with difficulty, with GPT-5 achieving near-perfect results on basic problems but only 82% on graduate-level proofs.