Skip to content

LearnOpt framework: exam preparation as constrained optimization over an LLM-built exam knowledge graph with BKT mastery updates

Design · All designs
Evidence · no claims cited

Description

LearnOpt reframes exam preparation as a constrained optimization problem: it ingests historical question papers, LLM-tags each question with chapter, topic, difficulty and skill via a chain-of-evidence prompt, builds a knowledge graph in FalkorDB with prerequisite edges, and "produces a ranked, time-bounded study plan specifying which concept subgroups to study, in what sequence, and with what expected score gain." Mastery is initialized from self-report and updated via standard BKT; the objective is solved with Google OR-Tools CP-SAT and re-run after each study session. Its substantive extension over the Embibe PAJ precedent is a skill-weighted objective coupling the exam's empirical cognitive-operation distribution into the reward signal.

Design Implications

Context

Requirements

  • Historical question papers in PDF format, an official syllabus, and a student mastery vector as inputs; full question text including options for reliable skill tagging

Constraints

  • Validated on MCQ exams; transfer to subjective or coding assessments requires further work
  • Mastery vectors initialized from self-report introduce subjective bias; optimization evaluation is simulation-based with no measured student outcomes

Target Learners

  • NEET and JEE aspirants preparing for standardized examinations

Learning Goals

  • Efficient allocation of limited preparation time across topics and cognitive skills to maximize expected exam score

Claims

Examples

-

Key Sources

  • Joy Bose, Om Thomas. (2025). LearnOpt: Recovering the Latent Cognitive Structure of Standardized Examinations via Knowledge Graphs and Constrained Optimization. https://github.com/learnopt/learnopt
Page metadata
Field Value
type design
id learnopt-framework-theory
title LearnOpt framework: exam preparation as constrained optimization over an LLM-built exam knowledge graph with BKT mastery updates
description LearnOpt reframes exam preparation as a constrained optimization problem: it ingests historical question papers, LLM-tags each question with chapter, topic, difficulty and skill via a chain-of-evidence prompt, builds...
status review
generated by process:settle-candidates on 2026-10-11
trust unverified — no one has confirmed this page's content
sources 1 — joy-bose-2025

Frontmatter is the page's machine-readable record — see the Schema & Guide.