custom-config accuracy · 23,624 rubric evals · all time since Jun 30, 2026
Excludes Visuals & Code on Paper — beta item types, tracked per-type below
Q2 guide · manual evals: 85.00% overall (incl. beta types)
88.13%
16,251 rubric evals · custom · all time
Q2 guide · 86.77%
91.55%
7,373 rubric evals · custom · all time
Q2 guide · 86.39%
99.63%
2,188 rubric evals · custom · all time
Q2 guide · 85.15% · beta — not counted in Overall
100.00%
250 rubric evals · custom · all time
Q2 guide · 66.04% · beta — not counted in Overall
98.95%
4,166,944 answers · all time
Objective · answer-level · all items
92.92%
728,329 answers · all time
Objective · answer-level · all items
78.72%
37,137 answers · all time
Objective · answer-level · all items
82.19%
20,937 answers · all time
Objective · answer-level · all items
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), from teachers with at least 3 activities of 10+ papers. 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 Nov 5, 3:09 AM
Cron-computed strictness accuracy per day. Cells under 30 rubric evaluations are muted (low volume). Most recent day is highlighted.
| Day | Problem Solvingavg 88.13% | Essayavg 91.55% | Visualsavg 99.63% | Code on Paperavg 100.00% |
|---|---|---|---|---|
| 2026-11-05 | 100.00% 80 evals | — | — | — |
| 2026-11-04 | 100.00% 91 evals | 100.00% 8 evals | — | — |
| 2026-11-03 | 100.00% 5 evals | 100.00% 17 evals | — | — |
| 2026-11-02 | 99.49% 198 evals | 100.00% 4 evals | — | — |
| 2026-11-01 | 100.00% 4 evals | — | — | — |
| 2026-10-31 | — | 100.00% 3 evals | — | — |
| 2026-10-30 | 93.02% 43 evals | 100.00% 65 evals | — | — |
| 2026-10-29 | — | 90.17% 1,078 evals | — | — |
| 2026-10-28 | 100.00% 147 evals | 100.00% 58 evals | 100.00% 4 evals | — |
| 2026-10-27 | — | 98.99% 594 evals | — | — |
| 2026-10-26 | 100.00% 16 evals | 100.00% 10 evals | — | — |
| 2026-10-23 | 81.55% 710 evals | — | — | — |
| 2026-10-22 | 100.00% 32 evals | 87.23% 94 evals | — | — |
| 2026-10-21 | 80.97% 1,750 evals | 90.91% 44 evals | — | — |
| 2026-10-20 | 88.84% 1,147 evals | 93.18% 44 evals | — | — |
| 2026-10-19 | 84.08% 1,225 evals | 99.17% 121 evals | — | 100.00% 250 evals |
| 2026-10-17 | — | 95.83% 24 evals | — | — |
| 2026-10-16 | 97.24% 543 evals | 58.06% 31 evals | 95.45% 110 evals | — |
| 2026-10-15 | 100.00% 22 evals | 91.41% 163 evals | 50.00% 4 evals | — |
| 2026-10-14 | — | 83.19% 226 evals | 100.00% 4 evals | — |
| 2026-10-13 | 80.95% 252 evals | 100.00% 4 evals | 100.00% 166 evals | — |
| 2026-10-12 | — | 66.67% 12 evals | 100.00% 30 evals | — |
| 2026-10-09 | 99.50% 201 evals | 100.00% 10 evals | 100.00% 12 evals | — |
| 2026-10-08 | 100.00% 9 evals | 100.00% 5 evals | — | — |
| 2026-10-07 | 88.89% 9 evals | 100.00% 70 evals | — | — |
| 2026-10-06 | — | 100.00% 132 evals | — | — |
| 2026-10-05 | 100.00% 32 evals | — | 100.00% 6 evals | — |
| 2026-10-04 | 0.00% 5 evals | 100.00% 12 evals | — | — |
| 2026-10-03 | 100.00% 30 evals | — | — | — |
| 2026-10-02 | 54.86% 720 evals | — | — | — |
| 2026-10-01 | 70.89% 1,364 evals | 100.00% 4 evals | — | — |
| 2026-09-30 | 98.16% 272 evals | 100.00% 3 evals | — | — |
| 2026-09-29 | 99.51% 206 evals | 100.00% 47 evals | — | — |
| 2026-09-28 | — | 100.00% 105 evals | — | — |
| 2026-09-25 | 100.00% 130 evals | 100.00% 85 evals | — | — |
| 2026-09-24 | 100.00% 7 evals | 99.49% 195 evals | — | — |
| 2026-09-23 | 100.00% 66 evals | 99.51% 203 evals | — | — |
| 2026-09-22 | 100.00% 6 evals | 97.81% 137 evals | — | — |
| 2026-09-21 | 100.00% 22 evals | 75.00% 4 evals | — | — |
| 2026-09-20 | 100.00% 110 evals | — | — | — |
| 2026-09-19 | 100.00% 44 evals | — | — | — |
| 2026-09-18 | 98.31% 830 evals | 100.00% 48 evals | 100.00% 78 evals | — |
| 2026-09-17 | 94.86% 486 evals | 90.41% 490 evals | — | — |
| 2026-09-16 | 98.75% 883 evals | 79.37% 727 evals | — | — |
| 2026-09-15 | 99.42% 691 evals | 89.36% 47 evals | — | — |
| 2026-09-14 | 97.70% 87 evals | 82.08% 703 evals | — | — |
| 2026-09-11 | 100.00% 114 evals | 94.23% 52 evals | — | — |
| 2026-09-10 | 99.65% 285 evals | — | 99.94% 1,740 evals | — |
| 2026-09-08 | 98.77% 570 evals | 100.00% 159 evals | — | — |
| 2026-09-07 | — | 85.47% 461 evals | — | — |
| 2026-09-06 | 96.33% 109 evals | — | — | — |
| 2026-09-05 | 100.00% 52 evals | — | — | — |
| 2026-09-04 | 100.00% 13 evals | — | — | — |
| 2026-09-03 | 100.00% 13 evals | 100.00% 4 evals | — | — |
| 2026-09-02 | 100.00% 9 evals | 83.33% 36 evals | — | — |
| 2026-09-01 | 100.00% 42 evals | 100.00% 9 evals | — | — |
| 2026-08-29 | — | 100.00% 24 evals | — | — |
| 2026-08-28 | 97.54% 854 evals | 100.00% 173 evals | — | — |
| 2026-08-27 | — | 100.00% 45 evals | — | — |
| 2026-08-25 | 92.31% 13 evals | — | — | — |
| 2026-08-24 | — | 100.00% 75 evals | — | — |
| 2026-08-19 | 100.00% 24 evals | — | — | — |
| 2026-08-18 | 96.00% 50 evals | 100.00% 54 evals | — | — |
| 2026-08-17 | — | 100.00% 12 evals | — | — |
| 2026-08-14 | — | 100.00% 45 evals | — | — |
| 2026-08-13 | — | 100.00% 114 evals | — | — |
| 2026-08-11 | — | 97.33% 225 evals | — | — |
| 2026-08-05 | — | 97.81% 228 evals | 100.00% 34 evals | — |
| 2026-08-04 | — | 100.00% 6 evals | — | — |
| 2026-08-03 | — | 100.00% 24 evals | — | — |
| 2026-07-25 | 100.00% 10 evals | — | — | — |
| 2026-07-24 | 72.67% 172 evals | — | — | — |
| 2026-07-23 | 44.17% 206 evals | — | — | — |
| 2026-07-20 | 100.00% 210 evals | — | — | — |
| 2026-07-17 | 100.00% 32 evals | — | — | — |
| 2026-07-03 | 90.46% 765 evals | — | — | — |
| 2026-07-02 | 96.88% 32 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.
Combined accuracy across all item types
Math, science & logical reasoning
34,610 evaluations
Written response evaluation quality
16,757 evaluations
Diagram & image-based questions
653 evaluations
Handwritten code interpretation
4,455 evaluations
Record a new evaluation run. Enter the eval count and accuracy % for each item type.
Changes compared to the previous snapshot (+ = improved, − = declined). Most recent row is highlighted.
| Date | Evals | Overall | Prob. Solving | Essay | Visuals | Code on Paper |
|---|---|---|---|---|---|---|
2026-06-29original | 56,475 | 85.00% — | 86.77% — | 86.39% — | 85.15% — | 66.04% — |