AI Paper Atlas

Research map / Vision-language and multimodal / Multimodal LLMs

LiveK12Bench: Have Large Multimodal Models Truly Conquered High School-level Examinations?

Xiaohan Wang, Mingze Yin, Yilin Zhao, sinbadliu sinbadliu, Dian Li

ECCV 2026 · Vision-language and multimodal › Multimodal LLMs · approach: Multimodal LLM benchmarks · 3 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
switch from parsed text/image input to full exam-page image inputClaude-opus-4.6 · Complex Layout Setaccuracy: 87.2 -> 55.0Hurt
switch from parsed input to full exam-page image inputGemini-3-pro · Complex Layout Setaccuracy: 88.3 -> 60.0Hurt
switch from parsed input to full exam-page image inputGPT-4o · Complex Layout Setaccuracy: 35.0 -> 5.5Hurt
switch from parsed input to full exam-page image inputGPT series · Complex Layout Setaccuracy and OCS: average drop of 33.7% in Acc and 44.9 points in OCSHurt
evaluate process quality in addition to final answerGPT-5 · Rigorous Process SetOCS vs PES: 83.0 -> 56.0Hurt
evaluate process quality in addition to final answerGemini-3-pro · Rigorous Process SetOCS vs PES: 90.9 -> 76.9Hurt
compare long-horizon subset with overall datasetall models · Long-horizon Reasoning SetARL: -8 to -15Hurt
restrict generation budget to 10% of default lengthmost models · Long-horizon Reasoning Setaccuracy: approximately half of unconstrained accuracyHurt
use multi-LLM arbitration instead of a single GPT-5 judgeprocess-error evaluator · Rigorous Process Set sampleCohen's kappa: 0.82 -> 0.87Improved
use multi-LLM arbitration instead of a single GPT-5 judgeprocess-error evaluator · Rigorous Process Set sampleaccuracy: 82.0 -> 89.6Improved
compare subjects: Chemistry/Biology versus Mathematicsall models · LiveK12Benchaccuracy: all models perform worse in Chemistry and Biology compared to Mathematics and PhysicsHurt

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy