Dataset Viewer
Auto-converted to Parquet Duplicate
new_cards_per_day
int64
5
100
target_retention
stringclasses
3 values
day_7_reviews
int64
10
300
day_30_reviews
int64
20
500
day_90_reviews
int64
25
700
day_180_reviews
int64
25
900
day_365_reviews
int64
30
1k
steady_state_reviews
int64
45
1.6k
lifetime_review_steps
int64
9
16
5
85%
10
20
25
25
30
45
9
5
90%
10
20
25
30
35
55
11
5
95%
15
25
35
45
50
80
16
10
85%
20
40
50
50
60
90
9
10
90%
20
40
50
60
70
110
11
10
95%
30
50
70
90
100
160
16
15
85%
30
60
75
75
90
135
9
15
90%
30
60
75
90
105
165
11
15
95%
45
75
105
135
150
240
16
20
85%
40
80
100
100
120
180
9
20
90%
40
80
100
120
140
220
11
20
95%
60
100
140
180
200
320
16
30
85%
60
120
150
150
180
270
9
30
90%
60
120
150
180
210
330
11
30
95%
90
150
210
270
300
480
16
40
85%
80
160
200
200
240
360
9
40
90%
80
160
200
240
280
440
11
40
95%
120
200
280
360
400
640
16
50
85%
100
200
250
250
300
450
9
50
90%
100
200
250
300
350
550
11
50
95%
150
250
350
450
500
800
16
100
85%
200
400
500
500
600
900
9
100
90%
200
400
500
600
700
1,100
11
100
95%
300
500
700
900
1,000
1,600
16

Flashcard review workload model outputs

Modelled daily flashcard review loads for eight new-card rates (5 to 100 new cards per day) at 85%, 90% and 95% target retention, after 7, 30, 90, 180 and 365 days and at steady state. 24 rows, one CSV (flashcard-workload-model.csv).

Limits

These are outputs of a stated interval ladder, not observed learner data. Lapses are not modelled and real schedulers adapt intervals per card.

Method and source

Licence and citation

CC BY 4.0. Cite as: Fulin Labs, "Flashcard review workload model outputs," 2026, https://doi.org/10.5281/zenodo.23361866

Downloads last month
32