GradeChumEngineeringPresentQ3 · 2026
04
Goal #04 · AI Accuracy
90.85%

custom-config accuracy · 984 rubric evals · last 14 days

Q2 guide · manual evals: 85.00% overall

Gap to 95% target
4.15%
from daily strictness tracking
0%95% target100%
Problem Solving
4.16% gap

90.84%

982 rubric evals · custom · 14d

Q2 guide · 86.77%

Essay

100.00%

2 rubric evals · custom · 14d · low volume

Q2 guide · 86.39%

Visuals

No custom-config volume in the last 14 days

Q2 guide · 85.15%

Code on Paper

No custom-config volume in the last 14 days

Q2 guide · 66.04%

Multiple Choice
Target met

99.71%

63,182 answers · 14d

Objective · answer-level · all items

Identification
4.57% gap

90.43%

11,205 answers · 14d

Objective · answer-level · all items

Enumeration
2.96% gap

92.04%

754 answers · 14d

Objective · answer-level · all items

Multiple Answer
3.10% gap

91.90%

210 answers · 14d

Objective · answer-level · all items

Strictness Config Accuracy (daily)

Share of teacher-reviewed AI rubric grading that was not corrected, on items using strictness configs (reviewed = result viewed by the teacher, or answer explicitly corrected). Custom (teacher-made) configs are the primary series; system configs are shown for comparison. Dashed lines show the Q2 guide. Days with fewer than 30 rubric evaluations are hidden.

Last computed Jul 10, 3:06 AM

Daily History

Cron-computed strictness accuracy per day. Cells under 30 rubric evaluations are muted (low volume). Most recent day is highlighted.

DayProblem Solvingavg 90.83%Essayavg 100.00%VisualsCode on Paper
2026-07-03
89.80%
765 evals
2026-07-02
100.00%
16 evals
2026-07-01
100.00%
2 evals
2026-06-30
94.03%
201 evals

Manual Evaluation Runs

Curated eval-run results. Currently showing the Q2 guide — values below are reference only until a Q3 eval run is added.

Overall
10.00% gap

Combined accuracy across all item types

85.00%
85.0% currenttarget 95%
Problem Solving
8.23% gap

Math, science & logical reasoning

86.77%

34,610 evaluations

86.8% currenttarget 95%
Essay
8.61% gap

Written response evaluation quality

86.39%

16,757 evaluations

86.4% currenttarget 95%
Visuals
9.85% gap

Diagram & image-based questions

85.15%

653 evaluations

85.2% currenttarget 95%
Code on Paper
28.96% gap

Handwritten code interpretation

66.04%

4,455 evaluations

66.0% currenttarget 95%
Add Accuracy Snapshot

Record a new evaluation run. Enter the eval count and accuracy % for each item type.

Date the evaluation run was completed

Grand total across all item types

Combined accuracy across all types

Per Item Type

Problem Solving
Essay
Visuals
Code on Paper
Snapshot History

Changes compared to the previous snapshot (+ = improved, − = declined). Most recent row is highlighted.

DateEvalsOverallProb. SolvingEssayVisualsCode on Paper
2026-06-29original
56,475
85.00%
86.77%
86.39%
85.15%
66.04%