Nice CUET-Guru: An Adaptive CUET Practice & Mock-Test App
Nice CUET-Guru is an adaptive practice and mock-test app for
India's CUET (Common University Entrance Test), built in Flutter.
Every question is chosen by a Computerized Adaptive Testing (CAT)
engine built on a 2-parameter IRT model: student ability (theta)
is recomputed from the full response history after every answer
using Expected A Posteriori estimation, while a parallel Bayesian
Knowledge Tracing layer maintains per-topic mastery and drives the
strength/weakness analytics and remediation flow.
-
Six Adaptive Test Modes – Interactive, Subject, CUET Mock,
Weakness, Speed (Rapid), and PYQ sessions, all driven by the
same CAT engine.
-
IRT-Based Ability Estimation – Theta is re-estimated via EAP
over an 81-point quadrature grid, not a simple right/wrong
heuristic.
-
Fisher-Information Question Selection – Picks the most
informative question for a student's current ability, with
topic cooldown for variety.
-
Bayesian Knowledge Tracing – Maintains per-topic mastery
probability, updated instantly after each response.
-
CUET-Format Mock Tests – Section timers, NTA-style question
palette, mark-for-review, and +5 / -1 / 0 marking.
-
Exposure Cooldown – 14/30/60-day cooldowns by mode so the
same question never reappears too soon.
-
Home Dashboard – Ability ring, weekly accuracy chart,
strength/weakness panel, subject progress, and recent tests.
-
Subject Mastery Bands – Mastered, Proficient, Developing,
Weak, and Not Started, computed from live BKT probabilities.
-
Atomic Passage Groups – Reading-comprehension passages are
scheduled at fixed slots, with ability updated once per
group, never mid-passage.
-
Rich Content Rendering – LaTeX/MathJax rendering for math
and science questions, with a pure-Flutter fallback.
-
PYQ Bank – Previous-year CUET questions with their own
60-day exposure cooldown.
-
Session Review & Analytics – Post-test breakdown by
question outcome, per-topic accuracy, and total time.