Pegasus Assess

Every Exam. Written,Watched, Graded.

Get Pegasus Assess for your Institution
assess.youruniversity.whitebird.ai

CS-405 · Midterm · draft 2

generated from Lectures 1–7 · 50 marks

AI drafted
Q1

Define the learning rate and its role in convergence.

Recall2m
Q2

Why does momentum dampen oscillation near minima?

Apply5m
Q3

Critique early stopping vs. L2 for the given loss curve.

Analyse8m
Q4

Design an experiment to catch overfitting before epoch 20.

Create10m
+ 8 more questions · Bloom's spread balanced

Set the blueprint. AI writes the paper.

Question generation from your own syllabus and lecture material — balanced across Bloom's levels, units and difficulty. Faculty approve, never start from blank.

Question Generation

The syllabus goes in, and a balanced paper comes out — drawn from your own lectures, graded by Bloom's level, yours to approve.

Midterm · draft

12 questions · awaiting approval

Bloom's balanced

Define the learning rate.

Recall

Why does momentum dampen oscillation?

Apply

Critique early stopping vs. L2.

Analyse
Paper Variants

One paper becomes three, shuffled sets matched for difficulty — no two neighbours ever hold the same sequence.

Midterm · approved

50 marks

12 questions · Bloom's spread locked

Set A

50m

Set B

50m

Set C

50m

shuffled order · same difficulty · no two neighbours match

Blueprint Coverage

Marks, units, difficulty — the blueprint you set is the paper you get, with the coverage map to prove nothing was skipped.

Marks

50 · 12 questions

Difficulty mix

40 / 40 / 20

Coverage

all 4 units · no gaps

the blueprint

Coverage · marks per unit

balanced

U1

U2

U3

U4

Recall

4m4m2m2m

Apply

4m6m6m4m

Analyse

2m4m4m2m

Watched by AI. Graded in hours.

Integrity signals instead of hours of video, auto-grading with rubric confidence, and results that land in the ERP gradebook without re-entry.

AI Proctoring

Signals, not surveillance, face match, second devices and tab switches — flagged for a human to review, never auto-ruled.

integrity · flagged for review

Seat R-41 · remote

2 signals

second face · phone in frame · clip saved

a human reviews · AI never rules alone

Integrity signals · CS-405 midterm

240 students · 97% clear

Live

Face match

240 / 240 verified

Tab switches

3 warned · auto-noted

Second device

seat R-41 · 2 sightings

Auto-Grading

720 scripts in hours, MCQs instant, descriptive answers rubric-scored — only low-confidence pages reach faculty.

scripts_hallA.zip

scripts_hallB.zip

remote_240.zip

Grading · CS-405 midterm

720 scripts · done in 3h 40m

97% auto

MCQ · 40 marks

240 scripts · instant

Graded

Short answers

rubric-scored · confidence 0.94

Graded

Long-form · Q12

18 low-confidence → faculty

Review
Integrity Review

Three moments to review, not three hours of video — every flag timestamped with its clip, escalation one click away.

Session · seat R-41 · 2h 58m

recorded · encrypted

tab switch

gaze away 14s

second face

3 moments to review · not 3 hours of video

1 escalated
ERP Sync

Publish once, and marks land in the Atlas gradebook — SGPA recomputed, transcripts current, zero re-entry.

Pegasus Assess · results

CS-405 · Midterm

720 scriptsgraded
ModerationHOD approved

published once

Atlas ERP · gradebook

zero re-entry

Internal marks updated

SGPA recomputed

1,240 transcripts current

Portal notified

AI

Question Generation

Papers drafted from your own material across Bloom's taxonomy — faculty approve.

AI

Integrity Signals

Face match, device and gaze signals, clip-stamped for fast human review.

AI

Rubric Scoring

Descriptive answers scored against your rubric with a confidence figure per script.

AI

Item Analytics

Per-question difficulty and discrimination after every exam — the bank improves itself.

Set the paper. AI does the rest.

Blueprint to question paper, proctoring to published marks — the whole assessment lifecycle, watched and graded in hours.

Talk to us