metadata
configs:
- config_name: gsm8k
data_files:
- split: train
path: data/gsm8k.jsonl
- config_name: math500
data_files:
- split: train
path: data/math500.jsonl
- config_name: rat
data_files:
- split: train
path: data/rat.jsonl
- config_name: simpleqa_verified
data_files:
- split: train
path: data/simpleqa_verified.jsonl
gsm8k
10000 rows (10000 successful, 0 failed). finish_reason counts: {'stop': 10000}. Model: Qwen/Qwen3-8B (revision b968826d9c46dd6066d109eabc6255188de91218), thinking on.
Sampling (as sent to the server)
{
"temperature": 0.9,
"top_p": 0.95,
"top_k": 20,
"min_p": 0.0,
"max_tokens": 32768,
"seed": "not sent"
}
Runs
git sha(s): 114fa47685b1a775f7d06964bf084369e7a5a561; slurm job(s): 21173750. Full run manifest: manifests/gsm8k.json.
Dataset provenance
{
"dataset": "gsm8k",
"repo": "openai/gsm8k",
"file": "gsm8k_test.jsonl",
"revision": "740312add88f781978c0658806c59bc2815b9866",
"sha256": "752adc99f23132275139f1f7be57126b08735a55daf0c56859ee6df70abafbac",
"rows_total": 1319,
"instruction": null
}
Columns
| Column | Description |
|---|---|
| dataset | dataset name (also the config name of this subset) |
| instance_id | instance id within the dataset |
| sample_idx | index of this sample for the instance |
| prompt_sha1 | sha1 of the exact prompt string sent to the server |
| prompt_tokens | prompt length in tokens |
| content | raw model output, untouched (thinking and inline) |
| finish_reason | vLLM finish_reason: stop or length |
| usage | vLLM usage block: prompt_tokens, completion_tokens, total_tokens |
| success | true if the request returned a generation |
| error | error string for a failed request, else null |
| attempts | requests made for this row (1 + retries) |
| latency_s | wall seconds for this request, from client send to response |
| ts | unix time the row was written |
simpleqa_verified
10000 rows (10000 successful, 0 failed). finish_reason counts: {'stop': 9992, 'length': 8}. Model: Qwen/Qwen3-8B (revision b968826d9c46dd6066d109eabc6255188de91218), thinking on.
Sampling (as sent to the server)
{
"temperature": 0.9,
"top_p": 0.95,
"top_k": 20,
"min_p": 0.0,
"max_tokens": 32768,
"seed": "not sent"
}
Runs
git sha(s): 114fa47685b1a775f7d06964bf084369e7a5a561; slurm job(s): 21173751. Full run manifest: manifests/simpleqa_verified.json.
Dataset provenance
{
"dataset": "simpleqa_verified",
"repo": "google/simpleqa-verified",
"file": "simpleqa_verified.csv",
"revision": "0dc97e0d28d8233463e005cdc4475cc2a13ba2dc",
"sha256": "b5db21155444763543fe31b67e7cf28ce2bb225742a5b889421c5f182e2f92f5",
"rows_total": 1000,
"instruction": null
}
Columns
| Column | Description |
|---|---|
| dataset | dataset name (also the config name of this subset) |
| instance_id | instance id within the dataset |
| sample_idx | index of this sample for the instance |
| prompt_sha1 | sha1 of the exact prompt string sent to the server |
| prompt_tokens | prompt length in tokens |
| content | raw model output, untouched (thinking and inline) |
| finish_reason | vLLM finish_reason: stop or length |
| usage | vLLM usage block: prompt_tokens, completion_tokens, total_tokens |
| success | true if the request returned a generation |
| error | error string for a failed request, else null |
| attempts | requests made for this row (1 + retries) |
| latency_s | wall seconds for this request, from client send to response |
| ts | unix time the row was written |
rat
21200 rows (21200 successful, 0 failed). finish_reason counts: {'stop': 21200}. Model: Qwen/Qwen3-8B (revision b968826d9c46dd6066d109eabc6255188de91218), thinking on.
Sampling (as sent to the server)
{
"temperature": 0.9,
"top_p": 0.95,
"top_k": 20,
"min_p": 0.0,
"max_tokens": 32768,
"seed": "not sent"
}
Runs
git sha(s): 40360c26c912c61f9cf27a3b69bb34f1c2ff8715; slurm job(s): 21173752. Full run manifest: manifests/rat.json.
Dataset provenance
{
"dataset": "rat",
"rows_total": 212,
"instruction": "Respond with only the single word answer, nothing else.",
"source": {
"xlsx": "remote_association_test.xlsx",
"xlsx_sha256": "24e6ec274e6901294f82182878fa87f2c5a1b8a5c8948c5d769fb30c2a3bb36d",
"solutions_module": "remote_associations_test.py",
"solutions_module_sha256": "904562f242f46962eb6a21c8c922a698e12b4d3708d6989a73bdb8791da4aade",
"solution_entries_in_module": 312
},
"duplicate_cue_sets": [
[
"birthday",
"light",
"stick"
]
]
}
Columns
| Column | Description |
|---|---|
| dataset | dataset name (also the config name of this subset) |
| instance_id | instance id within the dataset |
| sample_idx | index of this sample for the instance |
| prompt_sha1 | sha1 of the exact prompt string sent to the server |
| prompt_tokens | prompt length in tokens |
| content | raw model output, untouched (thinking and inline) |
| finish_reason | vLLM finish_reason: stop or length |
| usage | vLLM usage block: prompt_tokens, completion_tokens, total_tokens |
| success | true if the request returned a generation |
| error | error string for a failed request, else null |
| attempts | requests made for this row (1 + retries) |
| latency_s | wall seconds for this request, from client send to response |
| ts | unix time the row was written |
math500
10000 rows (9950 successful, 50 failed). finish_reason counts: {'stop': 9947, None: 50, 'length': 3}. Model: Qwen/Qwen3-8B (revision b968826d9c46dd6066d109eabc6255188de91218), thinking on.
Sampling (as sent to the server)
{
"temperature": 0.9,
"top_p": 0.95,
"top_k": 20,
"min_p": 0.0,
"max_tokens": 32768,
"seed": "not sent",
"stop": []
}
Runs
git sha(s): 87dd0ea592e6ef65a8bb3e5fa11fa92985654613 (dirty); slurm job(s): 22032741. Full run manifest: manifests/math500.json.
Dataset provenance
{
"dataset": "math500",
"repo": "HuggingFaceH4/MATH-500",
"file": "math500_test.jsonl",
"revision": "6e4ed1a2a79af7d8630a6b768ec859cb5af4d3be",
"sha256": "35dc41080a3680858b27fa7e0533d2d547825316fc5dafe5d316f4ccc5a06132",
"rows_total": 500,
"instruction": null
}
Columns
| Column | Description |
|---|---|
| dataset | dataset name (also the config name of this subset) |
| instance_id | instance id within the dataset |
| sample_idx | index of this sample for the instance |
| prompt_sha1 | sha1 of the exact prompt string sent to the server |
| prompt_tokens | prompt length in tokens |
| content | raw model output, untouched (thinking and inline) |
| finish_reason | vLLM finish_reason: stop or length |
| usage | vLLM usage block: prompt_tokens, completion_tokens, total_tokens |
| success | true if the request returned a generation |
| error | error string for a failed request, else null |
| attempts | requests made for this row (1 + retries) |
| latency_s | wall seconds for this request, from client send to response |
| ts | unix time the row was written |