Dataset Preview
Duplicate
The full dataset viewer is not available (click to read why). Only showing a preview of the rows.
Job has been terminated due to a temporary spike in resource usage and may be restarted later.
Error code:   JobManagerCrashedError

Need help to make the dataset viewer work? Make sure to review how to configure the dataset viewer, and open a discussion for direct support.

id
string
category
string
task
string
difficulty
int64
messages
list
answer
string
solution
string
verification
string
grids
string
rule
string
generator
string
cvr-0000000
arithmetic
rate_with_idle_time
1
[ { "role": "user", "content": "The machine makes twenty-five tiles a minute while it runs. It was booked for 16 minutes but stood idle for 4 of them. How many tiles did it make?" } ]
300
Running time: 16 − 4 = 12 minutes. 25 × 12 = 300 tiles.
exact integer arithmetic
cael-verified-reasoning-1.1
cvr-0000001
logic
lineup
2
[ { "role": "user", "content": "Quick question: Three people sit in a row: Dara, Kofi, Yusuf. Yusuf is to the left of Dara. Dara is not at either end. Who is in position 3, counting from the left? Please show your working." } ]
Kofi
Trying every order of the 3 people, only one satisfies all the statements: Yusuf, Dara, Kofi. So the answer is Kofi.
exhaustive search over 6 orders; exactly one survives
cael-verified-reasoning-1.1
cvr-0000002
logic
lineup
3
[ { "role": "user", "content": "Four people sit in a row: Diego, Felix, Omar, Oona. Omar is to the left of Oona. Felix is not next to Oona. Felix is not next to Omar. Diego is not at either end. Felix is to the left of Oona. In which position is Diego, counting from the left? I want to be sure I have this rig...
second
Trying every order of the 4 people, only one satisfies all the statements: Felix, Diego, Omar, Oona. So the answer is second.
exhaustive search over 24 orders; exactly one survives
cael-verified-reasoning-1.1
cvr-0000003
arc_grids
tile_2x2
1
[ { "role": "user", "content": "Example 1\nInput:\n0 0 0 0 0 8\n0 8 0 0 8 8\n5 8 0 0 8 0\nOutput:\n0 0 0 0 0 8 0 0 0 0 0 8\n0 8 0 0 8 8 0 8 0 0 8 8\n5 8 0 0 8 0 5 8 0 0 8 0\n0 0 0 0 0 8 0 0 0 0 0 8\n0 8 0 0 8 8 0 8 0 0 8 8\n5 8 0 0 8 0 5 8 0 0 8 0\n\nExample 2\nInput:\n0 0\n8 0\n0 0\n0 0\nOutput:\n0 0 0 0\n8 ...
0 5 0 0 0 0 5 0 0 0 5 5 0 5 0 5 5 0 5 0 0 0 0 8 0 0 0 0 8 0 0 5 0 0 0 0 5 0 0 0 5 5 0 5 0 5 5 0 5 0 0 0 0 8 0 0 0 0 8 0
Rule: Tile the grid twice across and twice down. Applied to the test input it gives the grid above.
rule applied programmatically to every pair
{"train": [{"input": [[0, 0, 0, 0, 0, 8], [0, 8, 0, 0, 8, 8], [5, 8, 0, 0, 8, 0]], "output": [[0, 0, 0, 0, 0, 8, 0, 0, 0, 0, 0, 8], [0, 8, 0, 0, 8, 8, 0, 8, 0, 0, 8, 8], [5, 8, 0, 0, 8, 0, 5, 8, 0, 0, 8, 0], [0, 0, 0, 0, 0, 8, 0, 0, 0, 0, 0, 8], [0, 8, 0, 0, 8, 8, 0, 8, 0, 0, 8, 8], [5, 8, 0, 0, 8, 0, 5, 8, 0, 0, 8, 0]...
tile_2x2
cael-verified-reasoning-1.1
cvr-0000004
arithmetic
shares_owed
3
[ { "role": "user", "content": "Quick question: Six of us split the concert tickets equally. I paid the whole €349.50." }, { "role": "user", "content": "Four of the others have already paid me back. How much do the rest still owe me altogether?" } ]
€58.25
Each share is €349.50 ÷ 6 = €58.25. I owe my own share; 4 have paid, so 1 shares are outstanding. 1 × €58.25 = €58.25.
exact division and multiplication
cael-verified-reasoning-1.1
cvr-0000005
arithmetic
pay_minus_spending
3
[ { "role": "user", "content": "Can you help me with this? Dev earns $29 an hour. They worked 5 hours on Monday and 2.5 hours on Tuesday, then spent $45. How much of that pay is left? I want to be sure I have this right." } ]
$172.50
Hours: 5 + 2.5 = 7.5. Pay: 7.5 × $29 = $217.50. Left: $217.50 − $45 = $172.50.
exact arithmetic in cents
cael-verified-reasoning-1.1
cvr-0000006
conflicting_sources
withdrawn
1
[ { "role": "user", "content": "Report A says the delivery came at 11:30. Report South says the delivery came at 10:00." }, { "role": "user", "content": "Report A has withdrawn its statement. Which source still has a claim on record, and what does it say? I want to be sure I have this right." } ...
Report South: the delivery came at 10:00
Report A withdrew, so only Report South remains: the delivery came at 10:00.
rule-based resolution
cael-verified-reasoning-1.1
cvr-0000007
logic
comparisons
3
[ { "role": "user", "content": "I'm trying to work something out. Ivan is heavier than Eli. Quinn is heavier than Ivan. Hiro is heavier than Quinn. Who is the lightest? Please show your working." } ]
Eli
Only Eli is never on the losing side of a comparison, so Eli is the lightest.
graph of the stated comparisons (sources/sinks)
cael-verified-reasoning-1.1
cvr-0000008
unanswerable
future_count
3
[ { "role": "user", "content": "We are holding a garden party next August. Exactly how many guests will come?" } ]
Cannot be determined
Future attendance depends on choices not yet made; it can be estimated, not known.
future event
cael-verified-reasoning-1.1
cvr-0000009
logic
comparisons
3
[ { "role": "user", "content": "Can you help me with this? Basil is slower than Theo. Kira is faster than Soren. Felix is faster than Kira. Basil is faster than Felix. Who is the slowest? I want to be sure I have this right." } ]
Soren
Only Soren is never on the losing side of a comparison, so Soren is the slowest.
graph of the stated comparisons (sources/sinks)
cael-verified-reasoning-1.1
cvr-0000010
memory
which_of_two
1
[ { "role": "user", "content": "Here is a puzzle. I live in Turku." }, { "role": "user", "content": "The weather is nice today." }, { "role": "user", "content": "I grew up in Lyon." }, { "role": "user", "content": "Where do I live? I want to be sure I have this right." } ...
Turku
You live in Turku and grew up in Lyon.
attribute lookup by relation
cael-verified-reasoning-1.1
cvr-0000011
unanswerable
too_few_clues
3
[ { "role": "user", "content": "I picked a whole number strictly between 40 and 50. It is even. Which number is it?" } ]
Cannot be determined
4 numbers fit: 42, 44, 46, 48.
enumeration of candidates
cael-verified-reasoning-1.1
cvr-0000012
unanswerable
missing_info
3
[ { "role": "user", "content": "The train leaves at 12:24. At what time does it arrive? I want to be sure I have this right." } ]
Cannot be determined
Only the departure time is given; without the journey time or timetable the arrival time is unknown.
missing premise
cael-verified-reasoning-1.1
cvr-0000013
logic
syllogism
1
[ { "role": "user", "content": "I'm trying to work something out. All dogs are mammals. All mammals are animals. Are all dogs animals? Please show your working." } ]
Yes
Every one of the dogs is among the mammals, and every one of those is among the animals.
set inclusion
cael-verified-reasoning-1.1
cvr-0000014
arithmetic
pay_minus_spending
2
[ { "role": "user", "content": "I'm trying to work something out. Kira earns £13 an hour. They worked 5 hours on Monday and 6 hours on Tuesday, then spent £75. How much of that pay is left?" } ]
£68
Hours: 5 + 6 = 11. Pay: 11 × £13 = £143. Left: £143 − £75 = £68.
exact arithmetic in cents
cael-verified-reasoning-1.1
cvr-0000015
conflicting_sources
withdrawn
1
[ { "role": "user", "content": "I'm trying to work something out. Report Blue says the gate opened at 08:21. Report North says the gate opened at 08:12." }, { "role": "user", "content": "Report North has withdrawn its statement. Which source still has a claim on record, and what does it say? Expla...
Report Blue: the gate opened at 08:21
Report North withdrew, so only Report Blue remains: the gate opened at 08:21.
rule-based resolution
cael-verified-reasoning-1.1
cvr-0000016
arithmetic
shares_owed
3
[ { "role": "user", "content": "Quick question: Four of us split the concert tickets equally. I paid the whole €234." }, { "role": "user", "content": "One of the others has already paid me back. How much do the rest still owe me altogether?" } ]
€117
Each share is €234 ÷ 4 = €58.50. I owe my own share; 1 has paid, so 2 shares are outstanding. 2 × €58.50 = €117.
exact division and multiplication
cael-verified-reasoning-1.1
cvr-0000017
arc_grids
span_rows
3
[ { "role": "user", "content": "Example 1\nInput:\n0 5 0 7 5\n6 0 6 0 0\n6 0 6 0 0\nOutput:\n0 5 5 7 5\n6 6 6 0 0\n6 6 6 0 0\n\nExample 2\nInput:\n5 5 6 0\n5 0 5 6\n0 0 0 0\nOutput:\n5 5 6 0\n5 5 5 6\n0 0 0 0\n\nExample 3\nInput:\n0 7 0 0 0 0\n0 0 6 0 7 0\n5 0 0 0 0 0\n7 0 0 7 0 0\n0 5 0 0 0 5\n0 7 0 0 0 0\n0...
0 0 0 0 0 0 0 0 7 7 7 7 7 7 0 0 0 0 0 0 0 7 0 0 0 0 0
Rule: In each row, fill the gap between the first and last cell of the same colour with that colour. Applied to the test input it gives the grid above.
rule applied programmatically to every pair
{"train": [{"input": [[0, 5, 0, 7, 5], [6, 0, 6, 0, 0], [6, 0, 6, 0, 0]], "output": [[0, 5, 5, 7, 5], [6, 6, 6, 0, 0], [6, 6, 6, 0, 0]]}, {"input": [[5, 5, 6, 0], [5, 0, 5, 6], [0, 0, 0, 0]], "output": [[5, 5, 6, 0], [5, 5, 5, 6], [0, 0, 0, 0]]}, {"input": [[0, 7, 0, 0, 0, 0], [0, 0, 6, 0, 7, 0], [5, 0, 0, 0, 0, 0], [7...
span_rows
cael-verified-reasoning-1.1
cvr-0000018
dates_times
weekday_after
3
[ { "role": "user", "content": "The notice went up on a Thursday. The change takes effect exactly five weeks and two days later. Which weekday is that? Explain briefly." } ]
Saturday
5 weeks bring you back to Thursday; 2 more days give Saturday.
arithmetic mod 7
cael-verified-reasoning-1.1
cvr-0000019
arc_grids
border
2
[ { "role": "user", "content": "Example 1\nInput:\n3 0 0 0 0 2 0 2\n0 0 0 0 2 0 0 2\n0 0 0 0 0 0 9 3\n0 2 2 3 0 0 9 9\n9 0 0 0 3 9 3 2\n2 0 0 0 0 0 9 3\n2 9 0 0 0 0 0 9\nOutput:\n5 5 5 5 5 5 5 5\n5 0 0 0 2 0 0 5\n5 0 0 0 0 0 9 5\n5 2 2 3 0 0 9 5\n5 0 0 0 3 9 3 5\n5 0 0 0 0 0 9 5\n5 5 5 5 5 5 5 5\n\nExample 2\...
5 5 5 5 5 5 0 0 0 5 5 0 0 0 5 5 0 0 0 5 5 0 0 0 5 5 0 0 0 5 5 5 5 5 5
Rule: Draw a one-cell frame of colour 5 around the edge of the grid. Applied to the test input it gives the grid above.
rule applied programmatically to every pair
{"train": [{"input": [[3, 0, 0, 0, 0, 2, 0, 2], [0, 0, 0, 0, 2, 0, 0, 2], [0, 0, 0, 0, 0, 0, 9, 3], [0, 2, 2, 3, 0, 0, 9, 9], [9, 0, 0, 0, 3, 9, 3, 2], [2, 0, 0, 0, 0, 0, 9, 3], [2, 9, 0, 0, 0, 0, 0, 9]], "output": [[5, 5, 5, 5, 5, 5, 5, 5], [5, 0, 0, 0, 2, 0, 0, 5], [5, 0, 0, 0, 0, 0, 9, 5], [5, 2, 2, 3, 0, 0, 9, 5], ...
border
cael-verified-reasoning-1.1
cvr-0000020
memory
never_said
1
[ { "role": "user", "content": "Here is a puzzle. I bought a striped apron at the market." }, { "role": "user", "content": "It came wrapped in brown paper." }, { "role": "user", "content": "How much did I tell you I paid for it? I want to be sure I have this right." } ]
You did not say
No price was mentioned anywhere in the conversation.
absence check over the conversation
cael-verified-reasoning-1.1
cvr-0000021
memory
latest_value
1
[ { "role": "user", "content": "My seat is 6A." }, { "role": "user", "content": "The weather is nice today." }, { "role": "user", "content": "They changed my seat to 24D this week." }, { "role": "user", "content": "I started a new book." }, { "role": "user", "co...
24D
The seat was 6A and was later changed to 24D; the latest value is 24D.
latest-write wins
cael-verified-reasoning-1.1
cvr-0000022
arithmetic
recipe_scaling
1
[ { "role": "user", "content": "A recipe for 4 people needs 240 g of butter. How much butter for 8 people? Explain briefly." } ]
480 g
240 g ÷ 4 = 60 g per person. × 8 = 480 g.
exact proportional scaling
cael-verified-reasoning-1.1
cvr-0000023
dates_times
days_between
1
[ { "role": "user", "content": "I'm trying to work something out. How many days are there from June 26, 2020 to September 22, 2020? I want to be sure I have this right." } ]
88 days
Counting from 26 June 2020 up to 22 September 2020 (not counting the first day) gives 88 days.
Python datetime
cael-verified-reasoning-1.1
cvr-0000024
composite
trip_budget
5
[ { "role": "user", "content": "Can you help me with this? I will travel from 3 June 2017 to 29 August 2017, both days included, and budget $7.50 a day. I have saved $410. Is that enough, and by how much am I over or short?" } ]
No, $250 short
3 June 2017 to 29 August 2017 inclusive is 88 days. 88 × $7.50 = $660. Saved $410: $250 short.
inclusive calendar count then exact money arithmetic
cael-verified-reasoning-1.1
cvr-0000025
composite
rental_due
4
[ { "role": "user", "content": "Quick question: I rent a van from 21 December 2029 to 14 January 2030 at $177.50 per night. I already paid a $125 deposit that counts towards the total. How much is still due?" } ]
$4,135
21 December 2029 to 14 January 2030 is 24 nights. 24 × $177.50 = $4,260. $4,260 − $125 = $4,135.
calendar arithmetic then exact money arithmetic
cael-verified-reasoning-1.1
cvr-0000026
arc_grids
gravity_down
2
[ { "role": "user", "content": "Example 1\nInput:\n9 0 0 0\n9 0 0 0\n0 0 0 0\n9 9 9 9\n0 9 9 0\nOutput:\n0 0 0 0\n0 0 0 0\n9 0 0 0\n9 9 9 0\n9 9 9 9\n\nExample 2\nInput:\n0 9 0 9 9 0 9\n0 0 0 9 0 0 9\n0 9 0 9 0 0 0\n0 0 0 0 0 0 0\n9 0 0 9 0 0 0\n0 0 0 0 9 0 0\n9 9 9 0 9 0 0\n9 9 0 0 0 0 9\n0 0 9 0 0 0 9\n0 9 ...
0 0 0 0 0 0 0 9 0 9 9 9
Rule: Every coloured cell falls to the bottom of its column, keeping order. Applied to the test input it gives the grid above.
rule applied programmatically to every pair
{"train": [{"input": [[9, 0, 0, 0], [9, 0, 0, 0], [0, 0, 0, 0], [9, 9, 9, 9], [0, 9, 9, 0]], "output": [[0, 0, 0, 0], [0, 0, 0, 0], [9, 0, 0, 0], [9, 9, 9, 0], [9, 9, 9, 9]]}, {"input": [[0, 9, 0, 9, 9, 0, 9], [0, 0, 0, 9, 0, 0, 9], [0, 9, 0, 9, 0, 0, 0], [0, 0, 0, 0, 0, 0, 0], [9, 0, 0, 9, 0, 0, 0], [0, 0, 0, 0, 9, 0,...
gravity_down
cael-verified-reasoning-1.1
cvr-0000027
composite
trip_budget
5
[ { "role": "user", "content": "Can you help me with this? I will travel from 9 October 2031 to 21 December 2031, both days included, and budget £42.50 a day. I have saved £530. Is that enough, and by how much am I over or short?" } ]
No, £2,615 short
9 October 2031 to 21 December 2031 inclusive is 74 days. 74 × £42.50 = £3,145. Saved £530: £2,615 short.
inclusive calendar count then exact money arithmetic
cael-verified-reasoning-1.1
cvr-0000028
logic
lineup
2
[ { "role": "user", "content": "Here is a puzzle. Three people finished a race: Emil, Leila, Zoe. Zoe is first. Leila is not at either end. In which position is Zoe, counting from first place? Just give me the answer." } ]
first
Trying every order of the 3 people, only one satisfies all the statements: Zoe, Leila, Emil. So the answer is first.
exhaustive search over 6 orders; exactly one survives
cael-verified-reasoning-1.1
cvr-0000029
arithmetic
packs
3
[ { "role": "user", "content": "I'm trying to work something out. Shells are only sold in boxes of 12, at €6.50 a box. I need 107 shells. What is the least I can spend? Please show your working." } ]
€58.50
107 ÷ 12 = 8.92, so 9 boxes are needed (8 would give only 96). 9 × €6.50 = €58.50.
ceiling division, then multiplication
cael-verified-reasoning-1.1
cvr-0000030
dates_times
duration_minutes
1
[ { "role": "user", "content": "Here is a puzzle. My recording starts at 9:41 am and stops at 12:10 pm. How many minutes long is it? Please show your working." } ]
149 minutes
From 9:41 am to 12:10 pm is 2 h 29 min = 149 minutes.
modular clock arithmetic
cael-verified-reasoning-1.1
cvr-0000031
logic
comparisons
3
[ { "role": "user", "content": "I'm trying to work something out. Basil is slower than Gita. Basil is faster than Paolo. Rosa is faster than Sam. Sam is faster than Gita. Who is the slowest?" } ]
Paolo
Only Paolo is never on the losing side of a comparison, so Paolo is the slowest.
graph of the stated comparisons (sources/sinks)
cael-verified-reasoning-1.1
cvr-0000032
dates_times
weekday_of_date
1
[ { "role": "user", "content": "I'm trying to work something out. What day of the week was 25 February 2086? Just give me the answer." } ]
Monday
In the Gregorian calendar 25 February 2086 falls on a Monday.
Python datetime (proleptic Gregorian)
cael-verified-reasoning-1.1
cvr-0000033
logic
lineup
2
[ { "role": "user", "content": "Can you help me with this? Three people stand in a queue: Chen, Mateo, Soren. Chen is in front of Soren. Soren is directly behind Chen. Mateo is in front of Soren. In which position is Mateo, counting from the front? Explain briefly." } ]
first
Trying every order of the 3 people, only one satisfies all the statements: Mateo, Chen, Soren. So the answer is first.
exhaustive search over 6 orders; exactly one survives
cael-verified-reasoning-1.1
cvr-0000034
arc_grids
flip_v
3
[ { "role": "user", "content": "Example 1\nInput:\n0 2 0 0 0 0 0\n0 4 0 2 0 0 0\n0 0 2 0 0 2 4\n4 4 4 0 0 0 0\n0 0 0 0 0 2 0\n0 0 0 0 2 0 0\n4 0 2 0 0 0 0\n0 0 0 0 0 0 0\n0 4 2 0 4 0 0\n2 0 0 4 0 0 0\nOutput:\n2 0 0 4 0 0 0\n0 4 2 0 4 0 0\n0 0 0 0 0 0 0\n4 0 2 0 0 0 0\n0 0 0 0 2 0 0\n0 0 0 0 0 2 0\n4 4 4 0 0 ...
4 4 0 0 0 0 0 0 4 0 2 2 2 4 0 0 0 0 4 0 0 2 0 2 2 0 2 4 0 2 2 2 0 4 0 0 0 4 0 0 4 0 0 0 0 2 0 4 2 4 0 0 4 4 0 2 0 0 0 0 0 0 2
Rule: Mirror the grid top to bottom. Applied to the test input it gives the grid above.
rule applied programmatically to every pair
{"train": [{"input": [[0, 2, 0, 0, 0, 0, 0], [0, 4, 0, 2, 0, 0, 0], [0, 0, 2, 0, 0, 2, 4], [4, 4, 4, 0, 0, 0, 0], [0, 0, 0, 0, 0, 2, 0], [0, 0, 0, 0, 2, 0, 0], [4, 0, 2, 0, 0, 0, 0], [0, 0, 0, 0, 0, 0, 0], [0, 4, 2, 0, 4, 0, 0], [2, 0, 0, 4, 0, 0, 0]], "output": [[2, 0, 0, 4, 0, 0, 0], [0, 4, 2, 0, 4, 0, 0], [0, 0, 0, ...
flip_v
cael-verified-reasoning-1.1
cvr-0000035
memory
moved_time
1
[ { "role": "user", "content": "Quick question: The appointment is at 6:00 pm." }, { "role": "user", "content": "The weather is nice today." }, { "role": "user", "content": "The appointment got moved forward by 120 minutes." }, { "role": "user", "content": "When is the appo...
4:00 pm
6:00 pm − 120 min = 4:00 pm.
clock arithmetic on the latest state
cael-verified-reasoning-1.1
cvr-0000036
logic
syllogism
1
[ { "role": "user", "content": "All violins are string instruments. This one is not one of the string instruments. Could this one be one of the violins?" } ]
No
Every one of the violins is among the string instruments; this one is not among the string instruments, so it cannot be one of the violins.
contrapositive
cael-verified-reasoning-1.1
cvr-0000037
arithmetic
change
3
[ { "role": "user", "content": "My basket had 2 magazines at €3.50 each, two tickets at €4.99 each and two salads at €6 each. If I hand over €50, what change do I get back? I want to be sure I have this right." } ]
€21.02
2 × €3.50 = €7 2 × €4.99 = €9.98 2 × €6 = €12 Total €28.98. Change: €50 − €28.98 = €21.02.
exact integer arithmetic in cents
cael-verified-reasoning-1.1
cvr-0000038
memory
latest_value
1
[ { "role": "user", "content": "Quick question: My seat is 27D." }, { "role": "user", "content": "The coffee here is good." }, { "role": "user", "content": "They changed my seat to 17E this week." }, { "role": "user", "content": "Work was busy today." }, { "role": "...
17E
The seat was 27D and was later changed to 17E; the latest value is 17E.
latest-write wins
cael-verified-reasoning-1.1
cvr-0000039
dates_times
time_zone
3
[ { "role": "user", "content": "In a story, Ridge clocks run 8 hours behind Vale clocks. An event starts Thursday at 20:05 in Ridge. What day and time is that in Vale? I want to be sure I have this right." } ]
Friday 04:05
Vale is 8 hours ahead of Ridge, so add 8 hours: Friday 04:05.
modular arithmetic over a week
cael-verified-reasoning-1.1
cvr-0000040
dates_times
work_minus_break
1
[ { "role": "user", "content": "My shift runs from 08:13 to 15:27 with a 20-minute break. How long do I actually work? Just give me the answer." } ]
6 h 54 min
08:13 to 15:27 is 7 h 14 min; minus 20 min = 6 h 54 min.
clock arithmetic
cael-verified-reasoning-1.1
cvr-0000041
dates_times
age_on_date
3
[ { "role": "user", "content": "Uma was born on 18 December 1988. How old were they on 26 May 2022?" } ]
33
2022 − 1988 = 34; the birthday had not yet come that year, so the age is 33.
calendar comparison
cael-verified-reasoning-1.1
cvr-0000042
logic
syllogism
1
[ { "role": "user", "content": "I'm trying to work something out. All squares are rectangles. This tile is not one of the rectangles. Could this tile be one of the squares? I want to be sure I have this right." } ]
No
Every one of the squares is among the rectangles; this tile is not among the rectangles, so it cannot be one of the squares.
contrapositive
cael-verified-reasoning-1.1
cvr-0000043
logic
lineup
3
[ { "role": "user", "content": "Can you help me with this? Four people sit in a row: Ben, Paolo, Pia, Sam. Ben is not next to Pia. Paolo is to the left of Pia. Pia is on the right end. Pia is directly to the right of Sam. Ben is not at either end. Who is in position 3, counting from the left? Explain briefly....
Sam
Trying every order of the 4 people, only one satisfies all the statements: Paolo, Ben, Sam, Pia. So the answer is Sam.
exhaustive search over 24 orders; exactly one survives
cael-verified-reasoning-1.1
cvr-0000044
arithmetic
rate_with_idle_time
1
[ { "role": "user", "content": "The printer makes 5 labels a minute while it runs. It was booked for 67 minutes but stood idle for 21 of them. How many labels did it make? I want to be sure I have this right." } ]
230
Running time: 67 − 21 = 46 minutes. 5 × 46 = 230 labels.
exact integer arithmetic
cael-verified-reasoning-1.1
cvr-0000045
logic
lineup
4
[ { "role": "user", "content": "Five people finished a race: Esme, Fay, Nils, Soren, Tara. Soren is not next to Tara. Fay finished before Soren. Tara finished before Esme. Fay finished before Nils. Nils finished before Soren. Fay is not next to Nils. Esme is not next to Fay. Fay is not next to Soren. Fay fini...
Soren
Trying every order of the 5 people, only one satisfies all the statements: Fay, Tara, Esme, Nils, Soren. So the answer is Soren.
exhaustive search over 120 orders; exactly one survives
cael-verified-reasoning-1.1
cvr-0000046
unanswerable
emptied
3
[ { "role": "user", "content": "The drawer held exactly three bottles. I took out all three and put nothing back. What colour is the bottle still in the drawer?" } ]
There is none
All 3 were removed, so no bottle is left.
false premise
cael-verified-reasoning-1.1
cvr-0000047
conflicting_sources
correction
1
[ { "role": "user", "content": "Note Blue says the delivery came at 11:30. Note 1 says the delivery came at 14:00." }, { "role": "user", "content": "Note Blue has now corrected its entry to 16:15. Do the sources agree now?" } ]
No
After the correction Note Blue says 16:15 and Note 1 says 14:00; they still disagree.
rule-based resolution
cael-verified-reasoning-1.1
cvr-0000048
logic
lineup
5
[ { "role": "user", "content": "Help me check this: Six people sit in a row: Diego, Nils, Nora, Pia, Rafa, Vik. Diego is not next to Pia. Pia is not next to Rafa. Nora is not next to Pia. Nora is directly to the right of Rafa. Rafa is directly to the right of Vik. Vik is to the left of Pia. Diego is not next ...
fourth
Trying every order of the 6 people, only one satisfies all the statements: Diego, Vik, Rafa, Nora, Nils, Pia. So the answer is fourth.
exhaustive search over 720 orders; exactly one survives
cael-verified-reasoning-1.1
cvr-0000049
composite
shift_pay
4
[ { "role": "user", "content": "Quick question: I worked from 10:55 to 16:55 with an unpaid 30-minute break. I am paid £16.80 per hour. How much did I earn?" } ]
£92.40
10:55 to 16:55 is 360 minutes; minus 30 = 330 minutes. 330 minutes × £16.80 per 60 minutes = £92.40.
clock arithmetic then exact money arithmetic
cael-verified-reasoning-1.1
cvr-0000050
dates_times
days_between
1
[ { "role": "user", "content": "Can you help me with this? How many days are there from December 2, 2005 to June 23, 2006?" } ]
203 days
Counting from 2 December 2005 up to 23 June 2006 (not counting the first day) gives 203 days.
Python datetime
cael-verified-reasoning-1.1
cvr-0000051
dates_times
work_minus_break
1
[ { "role": "user", "content": "My shift runs from 07:54 to 14:30 with a 45-minute break. How long do I actually work? Just give me the answer." } ]
5 h 51 min
07:54 to 14:30 is 6 h 36 min; minus 45 min = 5 h 51 min.
clock arithmetic
cael-verified-reasoning-1.1
cvr-0000052
composite
rental_due
4
[ { "role": "user", "content": "Here is a puzzle. I rent a flat from 27 July 2031 to 1 August 2031 at $165 per night. I already paid a $100 deposit that counts towards the total. How much is still due? Just give me the answer." } ]
$725
27 July 2031 to 1 August 2031 is 5 nights. 5 × $165 = $825. $825 − $100 = $725.
calendar arithmetic then exact money arithmetic
cael-verified-reasoning-1.1
cvr-0000053
logic
lineup
5
[ { "role": "user", "content": "Help me check this: Six people finished a race: Amara, Hugo, Luca, Maya, Nina, Wei. Nina finished before Amara. Maya is first. Nina finished before Luca. Luca finished before Amara. Nina finished before Hugo. Luca is not next to Wei. Hugo finished right after Luca. Maya finishe...
sixth
Trying every order of the 6 people, only one satisfies all the statements: Maya, Nina, Luca, Hugo, Wei, Amara. So the answer is sixth.
exhaustive search over 720 orders; exactly one survives
cael-verified-reasoning-1.1
cvr-0000054
logic
conditional_ponens
2
[ { "role": "user", "content": "If the alarm goes off, the dog barks. The alarm went off. Did the dog bark?" } ]
Yes
Rule: if the alarm goes off, then the dog barks. Known: the alarm went off. Modus ponens: the condition holds, so the result follows.
propositional logic (truth-table)
cael-verified-reasoning-1.1
cvr-0000055
arithmetic
change
3
[ { "role": "user", "content": "Can you help me with this? My basket had 2 juices at €4.99 each and 2 pens at €4.50 each. If I hand over €20, what change do I get back?" } ]
€1.02
2 × €4.99 = €9.98 2 × €4.50 = €9 Total €18.98. Change: €20 − €18.98 = €1.02.
exact integer arithmetic in cents
cael-verified-reasoning-1.1
cvr-0000056
arithmetic
pay_minus_spending
2
[ { "role": "user", "content": "Quick question: Wei earns $29 an hour. They worked 5 hours on Monday and 4 hours on Tuesday, then spent $30. How much of that pay is left? I want to be sure I have this right." } ]
$231
Hours: 5 + 4 = 9. Pay: 9 × $29 = $261. Left: $261 − $30 = $231.
exact arithmetic in cents
cael-verified-reasoning-1.1
cvr-0000057
arithmetic
rate_with_idle_time
1
[ { "role": "user", "content": "Here is a puzzle. The press makes 30 tiles a minute while it runs. It was booked for 41 minutes but stood idle for 12 of them. How many tiles did it make?" } ]
870
Running time: 41 − 12 = 29 minutes. 30 × 29 = 870 tiles.
exact integer arithmetic
cael-verified-reasoning-1.1
cvr-0000058
arithmetic
discount_then_tax
3
[ { "role": "user", "content": "The lamp was $120 before a 10% discount. I also pay 12% sales tax on the discounted price. What do I pay in total?" } ]
$120.96
10% off $120 leaves $108. Adding 12% sales tax gives $120.96 (rounded to the nearest cent).
exact decimal arithmetic, rounded half up to cents
cael-verified-reasoning-1.1
cvr-0000059
memory
never_said
1
[ { "role": "user", "content": "Can you help me with this? I bought a wool hat at the market." }, { "role": "user", "content": "It came wrapped in brown paper." }, { "role": "user", "content": "How much did I tell you I paid for it? Explain briefly." } ]
You did not say
No price was mentioned anywhere in the conversation.
absence check over the conversation
cael-verified-reasoning-1.1
cvr-0000060
arc_grids
flip_h
2
[ { "role": "user", "content": "Example 1\nInput:\n3 7 7\n3 0 2\n0 3 2\n2 3 2\n2 7 3\nOutput:\n7 7 3\n2 0 3\n2 3 0\n2 3 2\n3 7 2\n\nExample 2\nInput:\n7 0 3 3 0 0 2 0 0\n7 3 0 0 0 7 7 2 0\n0 0 2 7 0 7 0 7 3\n2 7 0 2 7 0 0 0 3\n0 0 0 7 0 0 7 3 0\n3 3 0 2 0 3 7 0 0\n0 0 0 0 0 0 2 0 7\nOutput:\n0 0 2 0 0 3 3 0 7...
0 0 0 7 0 0 0 0 7 2 0 0 3 0 2 3 3 0 2 0 3 0 7 0 0 0 0 2 0 2
Rule: Mirror the grid left to right. Applied to the test input it gives the grid above.
rule applied programmatically to every pair
{"train": [{"input": [[3, 7, 7], [3, 0, 2], [0, 3, 2], [2, 3, 2], [2, 7, 3]], "output": [[7, 7, 3], [2, 0, 3], [2, 3, 0], [2, 3, 2], [3, 7, 2]]}, {"input": [[7, 0, 3, 3, 0, 0, 2, 0, 0], [7, 3, 0, 0, 0, 7, 7, 2, 0], [0, 0, 2, 7, 0, 7, 0, 7, 3], [2, 7, 0, 2, 7, 0, 0, 0, 3], [0, 0, 0, 7, 0, 0, 7, 3, 0], [3, 3, 0, 2, 0, 3,...
flip_h
cael-verified-reasoning-1.1
cvr-0000061
arithmetic
shares_owed
3
[ { "role": "user", "content": "Six of us split the dinner bill equally. I paid the whole £193.50." }, { "role": "user", "content": "How much do the rest still owe me altogether?" } ]
£161.25
Each share is £193.50 ÷ 6 = £32.25. I owe my own share; 0 have paid, so 5 shares are outstanding. 5 × £32.25 = £161.25.
exact division and multiplication
cael-verified-reasoning-1.1
cvr-0000062
conflicting_sources
no_priority
3
[ { "role": "user", "content": "Here is a puzzle. Note A says the visitor count was 125. Note 1 says the visitor count was 120. Neither source is more reliable than the other. What do these sources establish? Please show your working." } ]
Cannot be determined
The two sources disagree (125 vs 120) and neither has priority, so the answer stays open.
rule-based resolution
cael-verified-reasoning-1.1
cvr-0000063
logic
lineup
5
[ { "role": "user", "content": "Can you help me with this? Six people stand in a queue: Carmen, Cleo, Esme, Greta, Omar, Soren. Esme is not next to Omar. Cleo is not next to Soren. Carmen is not next to Esme. Esme is in front of Omar. Carmen is not at either end. Omar is last. Greta is not at either end. Cleo...
second
Trying every order of the 6 people, only one satisfies all the statements: Soren, Esme, Greta, Carmen, Cleo, Omar. So the answer is second.
exhaustive search over 720 orders; exactly one survives
cael-verified-reasoning-1.1
cvr-0000064
logic
lineup
5
[ { "role": "user", "content": "Here is a puzzle. Six people sit in a row: Iris, Kira, Nils, Sam, Viktor, Yusuf. Viktor is not next to Yusuf. Nils is to the left of Sam. Kira is to the left of Nils. Kira is not next to Nils. Kira is not at either end. Iris is directly to the right of Nils. Viktor is directly ...
Yusuf
Trying every order of the 6 people, only one satisfies all the statements: Yusuf, Kira, Viktor, Nils, Iris, Sam. So the answer is Yusuf.
exhaustive search over 720 orders; exactly one survives
cael-verified-reasoning-1.1
cvr-0000065
arithmetic
pay_minus_spending
2
[ { "role": "user", "content": "Can you help me with this? Priya earns £19 an hour. They worked 6 hours on Monday and 2 hours on Tuesday, then spent £35. How much of that pay is left? Explain briefly." } ]
£117
Hours: 6 + 2 = 8. Pay: 8 × £19 = £152. Left: £152 − £35 = £117.
exact arithmetic in cents
cael-verified-reasoning-1.1
cvr-0000066
arc_grids
stamp_cross
2
[ { "role": "user", "content": "Example 1\nInput:\n0 0 7 0\n0 7 0 2\n0 0 2 0\nOutput:\n0 1 7 1\n1 7 1 2\n0 1 2 1\n\nExample 2\nInput:\n0 0 0 0 0 0 0 0\n0 0 0 7 0 0 0 0\n0 0 0 0 0 0 0 0\nOutput:\n0 0 0 1 0 0 0 0\n0 0 1 7 1 0 0 0\n0 0 0 1 0 0 0 0\n\nExample 3\nInput:\n0 0 0 0 0 0 0 0 0 0\n0 0 0 0 0 0 0 0 2 0\n0...
0 1 7 1 2 1 0 1 0
Rule: Every coloured cell gets colour 1 on its empty up, down, left and right neighbours. Applied to the test input it gives the grid above.
rule applied programmatically to every pair
{"train": [{"input": [[0, 0, 7, 0], [0, 7, 0, 2], [0, 0, 2, 0]], "output": [[0, 1, 7, 1], [1, 7, 1, 2], [0, 1, 2, 1]]}, {"input": [[0, 0, 0, 0, 0, 0, 0, 0], [0, 0, 0, 7, 0, 0, 0, 0], [0, 0, 0, 0, 0, 0, 0, 0]], "output": [[0, 0, 0, 1, 0, 0, 0, 0], [0, 0, 1, 7, 1, 0, 0, 0], [0, 0, 0, 1, 0, 0, 0, 0]]}, {"input": [[0, 0, 0...
stamp_cross
cael-verified-reasoning-1.1
cvr-0000067
conflicting_sources
newest
1
[ { "role": "user", "content": "Help me check this: The website updated in 2018 says the gate opened at 07:55. The leaflet printed in 2015 says the gate opened at 08:21. Which is the current information?" } ]
07:55
The website (2018) is newer than the leaflet (2015), so its value 07:55 is current.
recency rule
cael-verified-reasoning-1.1
cvr-0000068
logic
lineup
2
[ { "role": "user", "content": "Three people finished a race: Luca, Milo, Ugo. Luca finished before Milo. Luca is not next to Ugo. In which position is Ugo, counting from first place? Please show your working." } ]
third
Trying every order of the 3 people, only one satisfies all the statements: Luca, Milo, Ugo. So the answer is third.
exhaustive search over 6 orders; exactly one survives
cael-verified-reasoning-1.1
cvr-0000069
logic
lineup
2
[ { "role": "user", "content": "Quick question: Three people stand in a queue: Omar, Rosa, Tara. Omar is not next to Tara. Omar is first. Who is in position 3, counting from the front? Explain briefly." } ]
Tara
Trying every order of the 3 people, only one satisfies all the statements: Omar, Rosa, Tara. So the answer is Tara.
exhaustive search over 6 orders; exactly one survives
cael-verified-reasoning-1.1
cvr-0000070
composite
rental_due
4
[ { "role": "user", "content": "Quick question: I rent a flat from 21 December 2018 to 13 January 2019 at £120 per night. I already paid a £50 deposit that counts towards the total. How much is still due? Just give me the answer." } ]
£2,710
21 December 2018 to 13 January 2019 is 23 nights. 23 × £120 = £2,760. £2,760 − £50 = £2,710.
calendar arithmetic then exact money arithmetic
cael-verified-reasoning-1.1
cvr-0000071
composite
trip_budget
5
[ { "role": "user", "content": "Can you help me with this? I will travel from 28 November 2027 to 1 February 2028, both days included, and budget €36.50 a day. I have saved €340. Is that enough, and by how much am I over or short?" } ]
No, €2,069 short
28 November 2027 to 1 February 2028 inclusive is 66 days. 66 × €36.50 = €2,409. Saved €340: €2,069 short.
inclusive calendar count then exact money arithmetic
cael-verified-reasoning-1.1
cvr-0000072
logic
conditional_tollens
2
[ { "role": "user", "content": "Quick question: If the shop is open, the lights are on. The lights are not on. Is the shop open?" } ]
No
Rule: if the shop is open, then the lights are on. Known: the lights are not on. Modus tollens: the result did not happen, so the condition cannot have held.
propositional logic (truth-table)
cael-verified-reasoning-1.1
cvr-0000073
logic
comparisons
3
[ { "role": "user", "content": "Here is a puzzle. Jonas is older than Wei. Jonas is younger than Basil. Who is the oldest? Just give me the answer." } ]
Basil
Only Basil is never on the losing side of a comparison, so Basil is the oldest.
graph of the stated comparisons (sources/sinks)
cael-verified-reasoning-1.1
cvr-0000074
memory
moved_time
1
[ { "role": "user", "content": "I'm trying to work something out. The appointment is at 11:00 am." }, { "role": "user", "content": "My cousin is visiting next week." }, { "role": "user", "content": "The appointment got moved forward by 30 minutes." }, { "role": "user", "con...
10:30 am
11:00 am − 30 min = 10:30 am.
clock arithmetic on the latest state
cael-verified-reasoning-1.1
cvr-0000075
dates_times
date_plus_period
3
[ { "role": "user", "content": "Quick question: The warranty begins on 31 July 2012 and lasts 1 year. On what date does it end? Just give me the answer." } ]
31 July 2013
31 July 2012 + 1 years = 31 July 2013.
calendar arithmetic with month-end clamping
cael-verified-reasoning-1.1
cvr-0000076
unanswerable
emptied
3
[ { "role": "user", "content": "Can you help me with this? The drawer held exactly four stickers. I took out all four and put nothing back. What colour is the sticker still in the drawer?" } ]
There is none
All 4 were removed, so no sticker is left.
false premise
cael-verified-reasoning-1.1
cvr-0000077
logic
lineup
4
[ { "role": "user", "content": "Quick question: Five people stand in a queue: Ada, Kira, Omar, Oona, Rosa. Oona is not at either end. Ada is in front of Omar. Ada is directly behind Oona. Omar is last. Oona is in front of Omar. Kira is not next to Rosa. Omar is not next to Oona. Omar is directly behind Kira. ...
third
Trying every order of the 5 people, only one satisfies all the statements: Rosa, Oona, Ada, Kira, Omar. So the answer is third.
exhaustive search over 120 orders; exactly one survives
cael-verified-reasoning-1.1
cvr-0000078
logic
lineup
3
[ { "role": "user", "content": "Can you help me with this? Four people finished a race: Felix, Iris, Omar, Rosa. Rosa is not at either end. Omar finished before Iris. Iris is not at either end. Felix finished right after Rosa. In which position is Omar, counting from first place?" } ]
first
Trying every order of the 4 people, only one satisfies all the statements: Omar, Iris, Rosa, Felix. So the answer is first.
exhaustive search over 24 orders; exactly one survives
cael-verified-reasoning-1.1
cvr-0000079
arc_grids
border
3
[ { "role": "user", "content": "Example 1\nInput:\n0 0 0 0 0\n0 0 0 0 0\n1 0 0 1 0\n0 0 0 0 0\n0 0 0 1 0\n0 0 0 0 1\nOutput:\n9 9 9 9 9\n9 0 0 0 9\n9 0 0 1 9\n9 0 0 0 9\n9 0 0 1 9\n9 9 9 9 9\n\nExample 2\nInput:\n0 0 1 0 0 0\n0 1 0 0 0 0\n0 0 1 0 0 0\n1 0 1 0 1 0\n0 0 1 0 0 0\n0 1 0 0 0 0\n0 0 0 0 0 0\n0 1 0 ...
9 9 9 9 9 9 9 9 1 0 0 0 0 9 9 0 0 0 0 1 9 9 1 0 0 0 1 9 9 0 0 0 0 0 9 9 0 0 0 0 0 9 9 0 0 1 0 1 9 9 1 0 0 0 1 9 9 9 9 9 9 9 9
Rule: Draw a one-cell frame of colour 9 around the edge of the grid. Applied to the test input it gives the grid above.
rule applied programmatically to every pair
{"train": [{"input": [[0, 0, 0, 0, 0], [0, 0, 0, 0, 0], [1, 0, 0, 1, 0], [0, 0, 0, 0, 0], [0, 0, 0, 1, 0], [0, 0, 0, 0, 1]], "output": [[9, 9, 9, 9, 9], [9, 0, 0, 0, 9], [9, 0, 0, 1, 9], [9, 0, 0, 0, 9], [9, 0, 0, 1, 9], [9, 9, 9, 9, 9]]}, {"input": [[0, 0, 1, 0, 0, 0], [0, 1, 0, 0, 0, 0], [0, 0, 1, 0, 0, 0], [1, 0, 1,...
border
cael-verified-reasoning-1.1
cvr-0000080
dates_times
work_minus_break
1
[ { "role": "user", "content": "My shift runs from 09:04 to 15:54 with a 45-minute break. How long do I actually work?" } ]
6 h 5 min
09:04 to 15:54 is 6 h 50 min; minus 45 min = 6 h 5 min.
clock arithmetic
cael-verified-reasoning-1.1
cvr-0000081
dates_times
weekday_of_date
1
[ { "role": "user", "content": "What day of the week was 12 July 2076?" } ]
Sunday
In the Gregorian calendar 12 July 2076 falls on a Sunday.
Python datetime (proleptic Gregorian)
cael-verified-reasoning-1.1
cvr-0000082
dates_times
days_between
1
[ { "role": "user", "content": "I'm trying to work something out. How many days are there from February 6, 2020 to May 15, 2020?" } ]
99 days
Counting from 6 February 2020 up to 15 May 2020 (not counting the first day) gives 99 days.
Python datetime
cael-verified-reasoning-1.1
cvr-0000083
arithmetic
change
3
[ { "role": "user", "content": "I'm trying to work something out. My basket had 1 coffee at €4.50 each, 3 sandwiches at €2.75 each and 3 magazines at €1.20 each. If I hand over €20, what change do I get back? Please show your working." } ]
€3.65
1 × €4.50 = €4.50 3 × €2.75 = €8.25 3 × €1.20 = €3.60 Total €16.35. Change: €20 − €16.35 = €3.65.
exact integer arithmetic in cents
cael-verified-reasoning-1.1
cvr-0000084
conflicting_sources
no_priority
3
[ { "role": "user", "content": "Quick question: Report South says the delivery came at 14:00. Report Blue says the delivery came at 16:15. Neither source is more reliable than the other. What do these sources establish? Just give me the answer." } ]
Cannot be determined
The two sources disagree (14:00 vs 16:15) and neither has priority, so the answer stays open.
rule-based resolution
cael-verified-reasoning-1.1
cvr-0000085
conflicting_sources
no_priority
3
[ { "role": "user", "content": "Report B says the delivery came at 10:00. Report 2 says the delivery came at 16:15. Neither source is more reliable than the other. What do these sources establish?" } ]
Cannot be determined
The two sources disagree (10:00 vs 16:15) and neither has priority, so the answer stays open.
rule-based resolution
cael-verified-reasoning-1.1
cvr-0000086
logic
lineup
2
[ { "role": "user", "content": "Three people sit in a row: Aiko, Bruno, Jonas. Jonas is on the left end. Jonas is to the left of Aiko. Jonas is to the left of Bruno. Bruno is not at either end. Who is in position 1, counting from the left? Please show your working." } ]
Jonas
Trying every order of the 3 people, only one satisfies all the statements: Jonas, Bruno, Aiko. So the answer is Jonas.
exhaustive search over 6 orders; exactly one survives
cael-verified-reasoning-1.1
cvr-0000088
logic
lineup
5
[ { "role": "user", "content": "Can you help me with this? Six people sit in a row: Amara, Chen, Greta, Ivan, Jade, Lina. Lina is on the left end. Chen is to the left of Greta. Amara is not next to Greta. Amara is not next to Chen. Greta is to the left of Amara. Lina is to the left of Jade. Chen is directly t...
second
Trying every order of the 6 people, only one satisfies all the statements: Lina, Jade, Chen, Greta, Ivan, Amara. So the answer is second.
exhaustive search over 720 orders; exactly one survives
cael-verified-reasoning-1.1
cvr-0000089
composite
shift_pay
4
[ { "role": "user", "content": "Quick question: I worked from 09:05 to 14:40 with an unpaid 45-minute break. I am paid £24.60 per hour. How much did I earn? Please show your working." } ]
£118.90
09:05 to 14:40 is 335 minutes; minus 45 = 290 minutes. 290 minutes × £24.60 per 60 minutes = £118.90.
clock arithmetic then exact money arithmetic
cael-verified-reasoning-1.1
cvr-0000090
dates_times
days_between
1
[ { "role": "user", "content": "How many days are there from 20 August 2024 to 19 December 2024? Please show your working." } ]
121 days
Counting from 20 August 2024 up to 19 December 2024 (not counting the first day) gives 121 days.
Python datetime
cael-verified-reasoning-1.1
cvr-0000091
logic
lineup
4
[ { "role": "user", "content": "Here is a puzzle. Five people stand in a queue: Amara, Dara, Ines, Rosa, Soren. Amara is in front of Soren. Amara is not next to Ines. Amara is not at either end. Rosa is in front of Amara. Dara is not at either end. Dara is directly behind Ines. Dara is not next to Rosa. In wh...
third
Trying every order of the 5 people, only one satisfies all the statements: Rosa, Ines, Dara, Amara, Soren. So the answer is third.
exhaustive search over 120 orders; exactly one survives
cael-verified-reasoning-1.1
cvr-0000092
logic
lineup
5
[ { "role": "user", "content": "Quick question: Six people finished a race: Farah, Kai, Lina, Rosa, Ugo, Vera. Ugo finished right after Vera. Farah finished before Ugo. Farah is not next to Rosa. Farah is first. Farah finished before Lina. Vera finished before Rosa. Lina finished right after Ugo. Farah finish...
Lina
Trying every order of the 6 people, only one satisfies all the statements: Farah, Vera, Ugo, Lina, Kai, Rosa. So the answer is Lina.
exhaustive search over 720 orders; exactly one survives
cael-verified-reasoning-1.1
cvr-0000093
arc_grids
flip_v
3
[ { "role": "user", "content": "Example 1\nInput:\n5 5 0 0 5 5 0\n5 0 5 0 5 0 0\n0 0 5 0 0 5 0\n0 0 0 5 0 5 5\n0 5 5 5 5 0 0\n5 5 5 5 0 5 5\n5 0 5 5 5 5 5\n0 5 0 0 0 5 0\nOutput:\n0 5 0 0 0 5 0\n5 0 5 5 5 5 5\n5 5 5 5 0 5 5\n0 5 5 5 5 0 0\n0 0 0 5 0 5 5\n0 0 5 0 0 5 0\n5 0 5 0 5 0 0\n5 5 0 0 5 5 0\n\nExample ...
5 5 5 5 5 0 0 0 5 5 5 0 0 5 0 5 0 5 0 5 5 0 5 5 5 0 5 0 5 0 0 5 5 0 0 5 5 0 0 0 5 5 0 5 0 5 0 5 5 5 0 5 5 0 0 0 0 5 5 5 0 5 5 0 0 0 5 0 5 5 5 5 5 0 5 5 0 5 5 5 5 0 5 5 0 5 5 5 5 5 0 0 0 5 0 0 0 5 0 5
Rule: Mirror the grid top to bottom. Applied to the test input it gives the grid above.
rule applied programmatically to every pair
{"train": [{"input": [[5, 5, 0, 0, 5, 5, 0], [5, 0, 5, 0, 5, 0, 0], [0, 0, 5, 0, 0, 5, 0], [0, 0, 0, 5, 0, 5, 5], [0, 5, 5, 5, 5, 0, 0], [5, 5, 5, 5, 0, 5, 5], [5, 0, 5, 5, 5, 5, 5], [0, 5, 0, 0, 0, 5, 0]], "output": [[0, 5, 0, 0, 0, 5, 0], [5, 0, 5, 5, 5, 5, 5], [5, 5, 5, 5, 0, 5, 5], [0, 5, 5, 5, 5, 0, 0], [0, 0, 0, ...
flip_v
cael-verified-reasoning-1.1
cvr-0000094
arc_grids
span_rows
3
[ { "role": "user", "content": "Example 1\nInput:\n0 0 8 0 0 8 8 0 8\n0 0 0 8 0 0 8 0 8\n0 0 8 0 0 0 0 0 0\nOutput:\n0 0 8 8 8 8 8 8 8\n0 0 0 8 8 8 8 8 8\n0 0 8 0 0 0 0 0 0\n\nExample 2\nInput:\n8 8 0 0 0 0 0 0\n0 0 0 8 0 8 0 0\n0 0 8 0 0 0 0 8\nOutput:\n8 8 0 0 0 0 0 0\n0 0 0 8 8 8 0 0\n0 0 8 8 8 8 8 8\n\nEx...
0 0 8 8 8 8 8 0 0 0 0 8 8 8 8 0 0 8 8 0 0 0 0 0 0 0 0 0 0 0 0 0
Rule: In each row, fill the gap between the first and last cell of the same colour with that colour. Applied to the test input it gives the grid above.
rule applied programmatically to every pair
{"train": [{"input": [[0, 0, 8, 0, 0, 8, 8, 0, 8], [0, 0, 0, 8, 0, 0, 8, 0, 8], [0, 0, 8, 0, 0, 0, 0, 0, 0]], "output": [[0, 0, 8, 8, 8, 8, 8, 8, 8], [0, 0, 0, 8, 8, 8, 8, 8, 8], [0, 0, 8, 0, 0, 0, 0, 0, 0]]}, {"input": [[8, 8, 0, 0, 0, 0, 0, 0], [0, 0, 0, 8, 0, 8, 0, 0], [0, 0, 8, 0, 0, 0, 0, 8]], "output": [[8, 8, 0,...
span_rows
cael-verified-reasoning-1.1
cvr-0000095
logic
comparisons
3
[ { "role": "user", "content": "Help me check this: Nils is older than Chen. Mateo is younger than Oona. Ivan is older than Nils. Mateo is older than Ivan. Who is the oldest? Just give me the answer." } ]
Oona
Only Oona is never on the losing side of a comparison, so Oona is the oldest.
graph of the stated comparisons (sources/sinks)
cael-verified-reasoning-1.1
cvr-0000096
composite
shift_pay
4
[ { "role": "user", "content": "Help me check this: I worked from 10:25 to 20:10 with an unpaid 45-minute break. I am paid £13.80 per hour. How much did I earn?" } ]
£124.20
10:25 to 20:10 is 585 minutes; minus 45 = 540 minutes. 540 minutes × £13.80 per 60 minutes = £124.20.
clock arithmetic then exact money arithmetic
cael-verified-reasoning-1.1
cvr-0000097
arc_grids
scale_2
2
[ { "role": "user", "content": "Example 1\nInput:\n0 2 0\n0 2 0\n0 0 2\n0 0 0\nOutput:\n0 0 2 2 0 0\n0 0 2 2 0 0\n0 0 2 2 0 0\n0 0 2 2 0 0\n0 0 0 0 2 2\n0 0 0 0 2 2\n0 0 0 0 0 0\n0 0 0 0 0 0\n\nExample 2\nInput:\n0 2 0 7\n7 7 0 2\n0 7 0 0\nOutput:\n0 0 2 2 0 0 7 7\n0 0 2 2 0 0 7 7\n7 7 7 7 0 0 2 2\n7 7 7 7 0 ...
0 0 7 7 0 0 2 2 0 0 0 0 7 7 0 0 2 2 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 7 7 0 0 0 0 0 0 0 0 7 7 2 2 0 0 2 2 0 0 7 7 2 2 0 0 2 2 0 0 7 7 0 0 0 0 0 0 0 0 7 7 0 0 0 0 0 0 0 0 7 7
Rule: Scale the grid up by 2: each cell becomes a 2×2 block. Applied to the test input it gives the grid above.
rule applied programmatically to every pair
{"train": [{"input": [[0, 2, 0], [0, 2, 0], [0, 0, 2], [0, 0, 0]], "output": [[0, 0, 2, 2, 0, 0], [0, 0, 2, 2, 0, 0], [0, 0, 2, 2, 0, 0], [0, 0, 2, 2, 0, 0], [0, 0, 0, 0, 2, 2], [0, 0, 0, 0, 2, 2], [0, 0, 0, 0, 0, 0], [0, 0, 0, 0, 0, 0]]}, {"input": [[0, 2, 0, 7], [7, 7, 0, 2], [0, 7, 0, 0]], "output": [[0, 0, 2, 2, 0,...
scale_2
cael-verified-reasoning-1.1
cvr-0000098
dates_times
date_plus_period
3
[ { "role": "user", "content": "My membership starts on 9 November 2011 and lasts 3 years. On what date does it end?" } ]
9 November 2014
9 November 2011 + 3 years = 9 November 2014.
calendar arithmetic with month-end clamping
cael-verified-reasoning-1.1
cvr-0000099
composite
shift_pay
4
[ { "role": "user", "content": "Can you help me with this? I worked from 06:35 to 15:20 with an unpaid 60-minute break. I am paid €17.40 per hour. How much did I earn?" } ]
€134.85
06:35 to 15:20 is 525 minutes; minus 60 = 465 minutes. 465 minutes × €17.40 per 60 minutes = €134.85.
clock arithmetic then exact money arithmetic
cael-verified-reasoning-1.1
cvr-0000100
arithmetic
shares_owed
3
[ { "role": "user", "content": "Here is a puzzle. Six of us split the cabin rental equally. I paid the whole £354." }, { "role": "user", "content": "One of the others has already paid me back. How much do the rest still owe me altogether? Explain briefly." } ]
£236
Each share is £354 ÷ 6 = £59. I owe my own share; 1 has paid, so 4 shares are outstanding. 4 × £59 = £236.
exact division and multiplication
cael-verified-reasoning-1.1
End of preview.

CAEL Verified Reasoning

2,000,000 reasoning problems whose answers are computed, not guessed, plus a 1,427-item hard evaluation split and the generator and verifier that made them.

Every example is built by a small program that sets up the situation, computes the answer exactly and writes the working. No neural network wrote any of it, so every answer is correct by construction, and every answer can be checked automatically with the included verify.py.

The generators come from the exact solvers of CAEL, a reasoning system built without neural networks.

What it is for

  • RL with verifiable rewards. verify.check(model_reply, row) returns True/False, so it can be used directly as a reward function. It reads the final answer and accepts different formats of the same answer.

  • Calibration and refusal training. About 2% of the problems (roughly 37,000) have the correct answer "Cannot be determined", or say that a premise is contradictory or was never stated. These come from:

    • broken comparison chains;
    • invalid if-then inferences (affirming the consequent, denying the antecedent);
    • missing journey times;
    • hidden random outcomes;
    • too few clues.

    Models should learn to say so instead of guessing.

  • Evaluation. The hard split is fixed and balanced across 8 areas, and none of its problems appear in train or test.

  • Curricula. Every row has a difficulty from 1 to 5.

Contents

Area Rows Examples
Multi-step arithmetic 557,513 change, discounts then tax, pack sizes, shares owed, rates with idle time, unit prices, corrected tallies
Dates and times 478,409 days between dates, weekday of a date, date plus months, time zones, durations, ages
ARC-style grids 261,445 14 grid transformations with 3 demonstrations and a test (generated, not ARC-AGI tasks)
Composite (2–3 areas chained) 259,936 shift times into pay, rental dates into money due, flight plus time zone, trip budgets
Logic and constraints 250,398 line-ups by exhaustive search, if-then (valid and invalid forms), comparison chains, syllogisms
Conflicting or changing information 117,280 disagreeing reports, priority rules, corrections, withdrawals, first-hand vs hearsay, newest source
Memory over a chat 67,934 multi-turn chats with distractors: latest value, moved times, set changes, declined offers
Unanswerable or trick 7,085 missing information, hidden outcomes, contradictions, too few clues, future counts

Splits:

Split Rows
train 1,980,000
test 20,000
hard 1,427

hard uses a different seed, keeps only difficulty ≥ 3, is about 200 per area (the small "unanswerable" area contributes only 27, because nearly all of its distinct problems are already in train), and is deduplicated against train and test.

Fields

Field Meaning
id Stable identifier
category, task Area and problem type (for example logic / lineup)
difficulty 1 (one step) to 5 (several steps, or a large search)
messages The conversation, as a list of {role, content} user turns. Some problems span several turns with distractors.
answer The exact answer
solution Step-by-step working
verification How the answer was computed, for example "exhaustive search over 720 orders; exactly one survives"
grids For ARC-style rows: JSON with the train pairs and the test input and output
rule For ARC-style rows: the transformation name
generator Generator version

Example

{
 "category": "logic",
 "task": "lineup",
 "difficulty": 4,
 "messages": [
  {
   "role": "user",
   "content": "Five people stand in a queue: Dara, Felix, Hugo, Ines, Nina. Dara is directly behind Ines. Hugo is directly behind Dara. Ines is not next to Nina. Nina is last. Hugo is not next to Nina. Who is in position 2, counting from the front? Just give me the answer."
  }
 ],
 "answer": "Dara",
 "solution": "Trying every order of the 5 people, only one satisfies all the statements: Ines, Dara, Hugo, Felix, Nina. So the answer is Dara.",
 "verification": "exhaustive search over 120 orders; exactly one survives"
}

Use

from datasets import load_dataset
ds = load_dataset("vynly/cael-verified-reasoning")

from verify import check            # verify.py is in this repository
check("Working... Answer: fourth", ds["train"][0])   # -> True / False

Generate as much as you want:

python generate.py --out mydata --n 5000000 --seed 123

The output is deterministic for a given seed. A new seed gives fresh problems, which is useful for contamination-free evaluation.

How it was checked

  • Verifier against the gold answers: every gold answer passes verify.check.
  • Verifier against wrong answers: wrong answers swapped in from other rows of the same task are rejected. The only exceptions are answers that are genuinely equivalent, such as the same amount with a different currency symbol, or a list in a different order.
  • Independent cross-check: CAEL's own solvers re-answered a random sample. They answered 251 of 800 sampled test problems (they stay silent on wording they do not cover) and agreed on 232. Every one of the 19 disagreements was traced to a reading error in CAEL or a verifier phrasing case (since fixed); none was a wrong dataset answer.
  • Exact duplicates removed: duplicates are judged on the problem text, before the extra wording is added.

Limitations

  • Templated language. Each problem type has a limited set of sentence patterns (with varied names, items, numbers, lead-ins and closing requests). It teaches reasoning patterns, not broad language; mix it with natural data.
  • Moderate difficulty. Mostly grade-school arithmetic, calendar and logic puzzles, plus composite multi-step problems. There is no proof-style mathematics and no code.
  • ARC-style grids are generated. They are not ARC-AGI tasks and contain none of them, and they use 14 common transformations. They are useful for training and probing program induction, not as an ARC-AGI score.
  • English only.
  • Uneven area sizes. Some areas (memory, unanswerable) have fewer distinct problems, because their templates allow fewer combinations. They were capped rather than padded with near-duplicates.

Licence and citation

CC BY 4.0. You may use it commercially, with attribution:

@misc{cael_verified_reasoning_2026,
  title  = {CAEL Verified Reasoning: computed-answer reasoning problems with a verifier},
  author = {Vovala14 (CAEL)},
  year   = {2026},
  url    = {https://huggingface.co/datasets/vynly/cael-verified-reasoning}
}

Made with the exact solvers of CAEL (non-neural reasoning, source-available). If you use it, I'd like to hear how it went.

Downloads last month
30