Search papers, labs, and topics across Lattice.
This paper introduces GRACE, a parameter-efficient framework that enhances educational visual question answering (VQA) by leveraging structured pedagogical metadata and multimodal evidence. By employing lightweight language and vision adapters alongside evidence-aware option calibration, GRACE achieves a notable accuracy improvement on ScienceQA, elevating the baseline from 90.5% to 93.1%. The findings demonstrate that incorporating educational context significantly enhances model performance, highlighting the importance of structured state cues in multimodal adaptation.
Grounded reasoning in educational VQA can boost accuracy by over 2.5% when leveraging structured pedagogical cues.
Educational visual question answering, or VQA, requires models to solve curriculum-oriented multiple-choice questions using both language and visual evidence. Compared with conventional open-ended VQA, educational examples often include structured assessment metadata, diagrams or image contexts, and semantically close answer options, creating strong opportunities for question-option shortcuts. We develop and evaluate a parameter-efficient adaptation framework for a frozen multimodal large language model in this setting. We introduce GRACE, Grounded Reasoning via Adapter Composition and Evidence-Aware Calibration, a framework that uses the pedagogical state of each question to specialize lightweight language and vision adaptation. The state combines inference-visible subject, grouped skill, grade, visual-context, question-intent, and option-structure cues. GRACE uses factor-specific prompts and lightweight visual adapters, then applies evidence-aware option calibration to score all candidates under a shared multimodal context. On ScienceQA, GRACE improves a shared-adapter baseline from 90.5 percent to 93.1 percent overall accuracy and from 88.7 percent to 91.2 percent on image-context questions. Removing pedagogical composition, option calibration, or the visual adapter reduces overall accuracy by 1.4, 1.0, and 1.5 points, respectively. These controlled results show that structured educational state is an effective routing signal for parameter-efficient multimodal adaptation.
