Datasets:
Dataset Preview
The full dataset viewer is not available (click to read why). Only showing a preview of the rows.
The dataset generation failed
Error code: DatasetGenerationError
Exception: CastError
Message: Couldn't cast
protocol: string
uid: string
dataset: string
replicate: int64
seed: int64
gold: string
prediction: string
correct: bool
response: string
token_ids: list<item: int64>
child 0, item: int64
generated_tokens: int64
prompt_tokens: int64
finish_reason: string
stop_reason: null
finished_unix: double
rank: int64
execution_id: string
to
{'protocol': Value('string'), 'uid': Value('string'), 'dataset': Value('string'), 'replicate': Value('int64'), 'seed': Value('int64'), 'gold': Value('string'), 'prediction': Value('string'), 'correct': Value('bool'), 'response': Value('string'), 'token_ids': List(Value('int64')), 'generated_tokens': Value('int64'), 'prompt_tokens': Value('int64'), 'finish_reason': Value('string'), 'stop_reason': Value('null'), 'finished_unix': Value('float64'), 'rank': Value('int64')}
because column names don't match
Traceback: Traceback (most recent call last):
File "/usr/local/lib/python3.14/site-packages/datasets/builder.py", line 1835, in _prepare_split_single
for key, table in generator:
^^^^^^^^^
File "/src/services/worker/src/worker/job_runners/config/parquet_and_info.py", line 621, in wrapped
for item in generator(*args, **kwargs):
~~~~~~~~~^^^^^^^^^^^^^^^^^
File "/usr/local/lib/python3.14/site-packages/datasets/packaged_modules/json/json.py", line 361, in _generate_tables
self._cast_table(pa_table, json_field_paths=json_field_paths),
~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/usr/local/lib/python3.14/site-packages/datasets/packaged_modules/json/json.py", line 132, in _cast_table
pa_table = table_cast(pa_table, self.info.features.arrow_schema)
File "/usr/local/lib/python3.14/site-packages/datasets/table.py", line 2381, in table_cast
return cast_table_to_schema(table, schema)
File "/usr/local/lib/python3.14/site-packages/datasets/table.py", line 2309, in cast_table_to_schema
raise CastError(
...<3 lines>...
)
datasets.table.CastError: Couldn't cast
protocol: string
uid: string
dataset: string
replicate: int64
seed: int64
gold: string
prediction: string
correct: bool
response: string
token_ids: list<item: int64>
child 0, item: int64
generated_tokens: int64
prompt_tokens: int64
finish_reason: string
stop_reason: null
finished_unix: double
rank: int64
execution_id: string
to
{'protocol': Value('string'), 'uid': Value('string'), 'dataset': Value('string'), 'replicate': Value('int64'), 'seed': Value('int64'), 'gold': Value('string'), 'prediction': Value('string'), 'correct': Value('bool'), 'response': Value('string'), 'token_ids': List(Value('int64')), 'generated_tokens': Value('int64'), 'prompt_tokens': Value('int64'), 'finish_reason': Value('string'), 'stop_reason': Value('null'), 'finished_unix': Value('float64'), 'rank': Value('int64')}
because column names don't match
The above exception was the direct cause of the following exception:
Traceback (most recent call last):
File "/src/services/worker/src/worker/job_runners/config/parquet_and_info.py", line 1379, in compute_config_parquet_and_info_response
parquet_operations, partial, estimated_dataset_info = stream_convert_to_parquet(
~~~~~~~~~~~~~~~~~~~~~~~~~^
builder, max_dataset_size_bytes=max_dataset_size_bytes
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
)
^
File "/src/services/worker/src/worker/job_runners/config/parquet_and_info.py", line 956, in stream_convert_to_parquet
builder._prepare_split(split_generator=splits_generators[split], file_format="parquet")
~~~~~~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/usr/local/lib/python3.14/site-packages/datasets/builder.py", line 1702, in _prepare_split
for job_id, done, content in self._prepare_split_single(
~~~~~~~~~~~~~~~~~~~~~~~~~~^
gen_kwargs=gen_kwargs, job_id=job_id, **_prepare_split_args
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
):
^
File "/usr/local/lib/python3.14/site-packages/datasets/builder.py", line 1887, in _prepare_split_single
raise DatasetGenerationError("An error occurred while generating the dataset") from e
datasets.exceptions.DatasetGenerationError: An error occurred while generating the datasetNeed help to make the dataset viewer work? Make sure to review how to configure the dataset viewer, and open a discussion for direct support.
protocol string | uid string | dataset string | replicate int64 | seed int64 | gold string | prediction string | correct bool | response string | token_ids list | generated_tokens int64 | prompt_tokens int64 | finish_reason string | stop_reason null | finished_unix float64 | rank int64 |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
looprl-math-pass8-16k-v1 | aime24/60 | aime24 | 4 | 1,761,157,396 | 204 | 204 | true | Okay, let's see. So Aya walks 9 kilometers each morning, and then stops at a coffee shop for t minutes. The problem gives two scenarios: one where she walks at s km/h taking 4 hours total (including coffee time), and another where she walks s+2 km/h taking 2 hours and 24 minutes total (again, including coffee time). We... | [
39122,
28,
1303,
506,
963,
30,
1644,
330,
5488,
16028,
216,
41,
12091,
971,
5738,
28,
284,
965,
12326,
418,
253,
6728,
6434,
327,
252,
3487,
30,
378,
1732,
3894,
827,
8537,
42,
582,
837,
1041,
16028,
418,
262,
6381,
31,
88,
2498,
... | 1,906 | 175 | stop | null | 1,793,608,337.180429 | 0 |
looprl-math-pass8-16k-v1 | aime24/60 | aime24 | 5 | 1,620,673,690 | 204 | 204 | true | Okay, let's see. So the problem is about Aya's walking speed and the time she spends walking plus the time in the coffee shop. We need to find out how long her walk takes when she walks at s + 0.5 km/h, including the t minutes in the coffee shop. Hmm.
First, let's parse the information given. When she walks at speed s... | [
39122,
28,
1303,
506,
963,
30,
1644,
260,
1732,
314,
563,
330,
5488,
506,
6139,
3547,
284,
260,
655,
1041,
21768,
6139,
8055,
260,
655,
281,
260,
6728,
6434,
30,
1046,
737,
288,
1042,
578,
638,
986,
874,
4549,
2933,
645,
1041,
16028... | 2,134 | 175 | stop | null | 1,793,608,339.382391 | 0 |
looprl-math-pass8-16k-v1 | aime24/60 | aime24 | 1 | 2,034,484,604 | 204 | 204 | true | Okay, let's try to solve this problem step by step. So, Aya goes for a 9-kilometer walk each morning, and then she stops at a coffee shop for t minutes. The time taken includes both her walking and the coffee shop stop.
First, when she walks at a constant speed of s kilometers per hour, the entire time (walking plus ... | [
39122,
28,
1303,
506,
1576,
288,
5482,
451,
1732,
1833,
411,
1833,
30,
1644,
28,
330,
5488,
3935,
327,
253,
216,
41,
29,
31385,
10798,
4549,
971,
5738,
28,
284,
965,
1041,
12326,
418,
253,
6728,
6434,
327,
252,
3487,
30,
378,
655,
... | 2,533 | 175 | stop | null | 1,793,608,343.213004 | 0 |
looprl-math-pass8-16k-v1 | aime24/60 | aime24 | 2 | 60,009,469 | 204 | 204 | true | Okay, let's see. So Aya walks 9 kilometers every morning. When she walks at a constant speed of s km/h, the total time including the coffee shop stop is 4 hours. Then, when she walks faster at s + 2 km/h, the total time is 2 hours and 24 minutes, again including the same t minutes in the coffee shop. Now, we need to fi... | [
39122,
28,
1303,
506,
963,
30,
1644,
330,
5488,
16028,
216,
41,
12091,
897,
5738,
30,
1550,
1041,
16028,
418,
253,
3836,
3547,
282,
262,
6381,
31,
88,
28,
260,
2719,
655,
1285,
260,
6728,
6434,
2853,
314,
216,
36,
2737,
30,
3875,
... | 2,576 | 175 | stop | null | 1,793,608,343.645381 | 0 |
looprl-math-pass8-16k-v1 | aime24/60 | aime24 | 7 | 759,576,576 | 204 | 204 | true | Okay, let me try to figure out this problem step by step. So, Aya goes for a 9-kilometer walk every morning and then stops at a coffee shop. The problem gives me two different scenarios where her walking speed changes, and in both cases, the total time includes the time she spends in the coffee shop, which is 't' minut... | [
39122,
28,
1303,
549,
1576,
288,
4110,
578,
451,
1732,
1833,
411,
1833,
30,
1644,
28,
330,
5488,
3935,
327,
253,
216,
41,
29,
31385,
10798,
4549,
897,
5738,
284,
965,
12326,
418,
253,
6728,
6434,
30,
378,
1732,
3894,
549,
827,
896,
... | 2,799 | 175 | stop | null | 1,793,608,345.723643 | 0 |
looprl-math-pass8-16k-v1 | aime24/60 | aime24 | 0 | 1,435,705,626 | 204 | 204 | true | Okay, let me try to figure out this problem. So, Aya walks 9 kilometers each morning, and then stops at a coffee shop for t minutes. The problem gives two different scenarios where her walking speed changes, and we need to find the time it takes her when she walks at s + 1/2 km/h, including the coffee shop stop. Hmm, l... | [
39122,
28,
1303,
549,
1576,
288,
4110,
578,
451,
1732,
30,
1644,
28,
330,
5488,
16028,
216,
41,
12091,
971,
5738,
28,
284,
965,
12326,
418,
253,
6728,
6434,
327,
252,
3487,
30,
378,
1732,
3894,
827,
896,
8537,
837,
874,
6139,
3547,
... | 3,000 | 175 | stop | null | 1,793,608,347.554172 | 0 |
looprl-math-pass8-16k-v1 | aime24/60 | aime24 | 6 | 466,326,022 | 204 | 204 | true | Okay, let's try to solve this problem step by step. So, Aya goes for a 9-kilometer walk every morning, and she stops at a coffee shop afterwards. The problem gives us two different scenarios where her walking speed changes, and we need to find out how long her walk takes when she walks at s + 1/2 km/h, including the co... | [
39122,
28,
1303,
506,
1576,
288,
5482,
451,
1732,
1833,
411,
1833,
30,
1644,
28,
330,
5488,
3935,
327,
253,
216,
41,
29,
31385,
10798,
4549,
897,
5738,
28,
284,
1041,
12326,
418,
253,
6728,
6434,
15195,
30,
378,
1732,
3894,
468,
827... | 3,384 | 175 | stop | null | 1,793,608,351.016972 | 0 |
looprl-math-pass8-16k-v1 | hmmt_feb25/1 | hmmt_feb25 | 0 | 823,194,946 | 103 | 103 | true | Okay, let's see. The problem is to compute the sum of the positive divisors of 9! that have a units digit of 1. Hmm, so first I need to find all the divisors of 9! that end with 1, and then add them up. But 9! is a big number, so maybe there's a smarter way than listing all divisors.
First, let's factorize 9! to under... | [
39122,
28,
1303,
506,
963,
30,
378,
1732,
314,
288,
12258,
260,
2028,
282,
260,
2731,
3044,
40520,
282,
216,
41,
17,
338,
457,
253,
4784,
11403,
282,
216,
33,
30,
41827,
28,
588,
808,
339,
737,
288,
1042,
511,
260,
3044,
40520,
28... | 2,938 | 67 | stop | null | 1,793,608,364.825813 | 0 |
looprl-math-pass8-16k-v1 | hmmt_feb25/1 | hmmt_feb25 | 1 | 582,798,224 | 103 | 103 | true | Okay, so I need to find the sum of all positive divisors of 9! that end with the digit 1. Hmm, let me start by recalling that 9! is 362880. But maybe it's better to factorize 9! into its prime factors first. That might make it easier to handle the divisors.
Alright, 9! = 9 × 8 × 7 × 6 × 5 × 4 × 3 × 2 × 1. Let me break... | [
39122,
28,
588,
339,
737,
288,
1042,
260,
2028,
282,
511,
2731,
3044,
40520,
282,
216,
41,
17,
338,
1112,
351,
260,
11403,
216,
33,
30,
41827,
28,
1303,
549,
1120,
411,
39029,
338,
216,
41,
17,
314,
216,
35,
38,
34,
40,
40,
32,
... | 2,966 | 67 | stop | null | 1,793,608,367.376252 | 0 |
looprl-math-pass8-16k-v1 | hmmt_feb25/1 | hmmt_feb25 | 6 | 318,720,521 | 103 | 103 | true | Okay, so I need to find the sum of all positive divisors of 9! that end with the digit 1. Hmm, let's start by figuring out what 9! is. 9 factorial is 9 × 8 × 7 × ... × 1. Let me compute that first.
9! = 9 × 8 × 7 × 6 × 5 × 4 × 3 × 2 × 1. Let me compute step by step:
9 × 8 = 72
72 × 7 = 504
504 × 6 = 3024
3024 × 5... | [
39122,
28,
588,
339,
737,
288,
1042,
260,
2028,
282,
511,
2731,
3044,
40520,
282,
216,
41,
17,
338,
1112,
351,
260,
11403,
216,
33,
30,
41827,
28,
1303,
506,
1120,
411,
23154,
578,
732,
216,
41,
17,
314,
30,
216,
41,
4186,
409,
... | 3,204 | 67 | stop | null | 1,793,608,382.76982 | 0 |
End of preview.
Ouro-1.4B Thinking: completed base, GRPO and RLTT evaluation results
Public evaluation artifacts: frozen benchmark questions/prompts and source revisions, all generated responses and token IDs, predictions/correctness, per-request seeds, native grader and sampling code, final summaries, worker exit evidence and logs. The checkpoint model repositories retain their existing visibility.
pass@8 is the fraction of questions with at least one correct answer in eight samples. Sampling: four recurrent loops, BF16 inference, maximum 16384 new tokens, temperature 0.6, top_p 0.95, top_k disabled, engine seed 42. Request seeds are SHA256(protocol|uid|replicate|42)[:16] & 0x7fffffff. The protocol is looprl-math-pass8-16k-v1. Last complete boxed answer is graded by exact normalization or math-verify 0.8.0; the native grader is included.
Base and GRPO have 1637 questions and 13096 responses each across nine suites. RLTT has 962 questions and 7696 responses across the eight requested suites. RLTT was not evaluated on OlympiadBench. All committed records were checked against frozen data, request seeds, unique replicate IDs 0-7 and native summaries. All evaluation workers and coordinators exited 0.
Base resumed 6947 previously committed samples. Its earlier execution used concurrency8/memory0.60/token-batch4096; the resumed execution used concurrency16/memory0.85/token-batch32768. Both original protocols are included. The original RLTT training driver has no recorded native exit code after platform termination. Evaluation was explicitly authorized after full checkpoint and exact CPU export verification. This does not change its unknown training exit code.
Checkpoint repositories
Final pass@8
| Dataset | Base | GRPO step140 | RLTT step140 |
|---|---|---|---|
| aime24 | 63.33% (19/30) | 73.33% (22/30) | 76.67% (23/30) |
| aime25 | 60.00% (18/30) | 73.33% (22/30) | 70.00% (21/30) |
| aime26 | 70.00% (21/30) | 70.00% (21/30) | 70.00% (21/30) |
| math500 | 97.40% (487/500) | 97.80% (489/500) | 98.00% (490/500) |
| amc23 | 92.50% (37/40) | 95.00% (38/40) | 95.00% (38/40) |
| olympiadbench | 74.07% (500/675) | 76.30% (515/675) | Not evaluated |
| minerva_math | 54.41% (148/272) | 55.88% (152/272) | 54.04% (147/272) |
| hmmt_feb25 | 40.00% (12/30) | 46.67% (14/30) | 36.67% (11/30) |
| hmmt_nov25 | 56.67% (17/30) | 66.67% (20/30) | 60.00% (18/30) |
Artifact directories
- base: summary.json, validation.json, responses/, data/, code/, logs/ and worker evidence.
- grpo_step140: summary.json, validation.json, responses/, data/, code/, logs/ and worker evidence.
- rltt_step140: summary.json, validation.json, responses/, data/, code/, logs/ and worker evidence.
comparison.json contains exact metrics and completion checks. release-manifest.json and SHA256SUMS bind the published files. No model weights, training data, training audit traces, runtime caches or credentials are included.
- Downloads last month
- 91