Dataset Viewer
Auto-converted to Parquet Duplicate
example_id
stringlengths
64
64
task_id
stringlengths
6
73
family
stringclasses
5 values
variant
stringclasses
3 values
frame
stringclasses
2 values
template_id
stringclasses
28 values
phrasing_id
stringclasses
6 values
card_id
stringlengths
3
81
messages
listlengths
3
3
label
unknown
label_source
stringclasses
4 values
label_version
stringclasses
2 values
split
stringclasses
1 value
source
stringclasses
7 values
restricted_statement
bool
1 class
source_revision
stringclasses
3 values
2c939fbcc7526be1d3eaddecb0b87225693323afea33b1dc826ff519822ceb3e
BigCodeBench/1
LIB
m1
self
q01
sc01
train-BigCodeBench/1-0
[ { "role": "system", "content": "You are a Python coding agent.\nHow a run works: you read the task and write a complete solution, then the harness runs the task's tests. If a test fails, you see the failure and write a corrected full solution. You get at most 3 attempts. The run ends when the tests pass or ...
{ "task_id": "BigCodeBench/1", "family": "LIB", "split": "train", "card_id": "train-BigCodeBench/1-0", "S": 13, "P": 78, "T_med_s": 18.6809326735, "T_q10_s": 14.9447461388, "T_q90_s": 23.3511658419, "TOK_med": 207, "ATT_med": 1, "PFP": 0.8261506276, "overhead_med_s": 2.7578557505, "n_runs": ...
measured
v2-latency
train
bigcode/bigcodebench
null
null
7e32c244fc5fcb92d075c4ad36fcf9fb46109d43e3a803e5bfea5b5001642e56
BigCodeBench/1
LIB
m1
self
q02
sc05
train-BigCodeBench/1-1
[ { "role": "system", "content": "You are a Python coding agent.\nHow a run works: you read the task and write a complete solution, then the harness runs the task's tests. If a test fails, you see the failure and write a corrected full solution. You get at most 3 attempts. The run ends when the tests pass or ...
{ "task_id": "BigCodeBench/1", "family": "LIB", "split": "train", "card_id": "train-BigCodeBench/1-1", "S": 76, "P": 1600, "T_med_s": 3.847341243, "T_q10_s": 3.0778729944, "T_q90_s": 4.8091765538, "TOK_med": 207, "ATT_med": 1, "PFP": 0.8261506276, "overhead_med_s": 1.1236570325, "n_runs": 8,...
measured
v2-latency
train
bigcode/bigcodebench
null
null
9149d377d82caa422d12557ed404dcc60472d3da81a9eb94afff398abee3f144
BigCodeBench/1
LIB
m1
self
q01
sc03
train-BigCodeBench/1-2
[ { "role": "system", "content": "You are a Python coding agent.\nHow a run works: you read the task and write a complete solution, then the harness runs the task's tests. If a test fails, you see the failure and write a corrected full solution. You get at most 3 attempts. The run ends when the tests pass or ...
{ "task_id": "BigCodeBench/1", "family": "LIB", "split": "train", "card_id": "train-BigCodeBench/1-2", "S": 72, "P": 2100, "T_med_s": 3.9787165563, "T_q10_s": 3.1829732451, "T_q90_s": 4.9733956954, "TOK_med": 207, "ATT_med": 1, "PFP": 0.8261506276, "overhead_med_s": 1.1037165563, "n_runs": 8...
measured
v2-latency
train
bigcode/bigcodebench
null
null
c5f8028bd6e4a61b501be58d32f6cddda7175cf04863fde23710da46ce43b231
BigCodeBench/1
LIB
m1
self
q17
sc04
train-BigCodeBench/1-3
[ { "role": "system", "content": "You are a Python coding agent.\nHow a run works: you read the task and write a complete solution, then the harness runs the task's tests. If a test fails, you see the failure and write a corrected full solution. You get at most 3 attempts. The run ends when the tests pass or ...
{ "task_id": "BigCodeBench/1", "family": "LIB", "split": "train", "card_id": "train-BigCodeBench/1-3", "S": 100, "P": 5700, "T_med_s": 3.1334158044000002, "T_q10_s": 2.5067326435, "T_q90_s": 3.9167697555, "TOK_med": 207, "ATT_med": 1, "PFP": 0.8261506276, "overhead_med_s": 1.0634158044, "n_r...
measured
v2-latency
train
bigcode/bigcodebench
null
null
916b08da78106813f7bffe84e9f43f1e170dd914ab46a8cb52e47cd2df2d9b83
BigCodeBench/1
LIB
m1
self
q15
sc03
train-BigCodeBench/1-4
[ { "role": "system", "content": "You are a Python coding agent.\nHow a run works: you read the task and write a complete solution, then the harness runs the task's tests. If a test fails, you see the failure and write a corrected full solution. You get at most 3 attempts. The run ends when the tests pass or ...
{ "task_id": "BigCodeBench/1", "family": "LIB", "split": "train", "card_id": "train-BigCodeBench/1-4", "S": 21, "P": 2900, "T_med_s": 10.9432567862, "T_q10_s": 8.754605429, "T_q90_s": 13.6790709827, "TOK_med": 207, "ATT_med": 1, "PFP": 0.8261506276, "overhead_med_s": 1.0861139291, "n_runs": ...
measured
v2-latency
train
bigcode/bigcodebench
null
null
9fa50a80e17ea735dd4f9c22ccd503ba34e64bf3ca02734d49f41861c71a2211
BigCodeBench/1
LIB
m1
self
q03
sc04
train-BigCodeBench/1-5
[ { "role": "system", "content": "You are a Python coding agent.\nHow a run works: you read the task and write a complete solution, then the harness runs the task's tests. If a test fails, you see the failure and write a corrected full solution. You get at most 3 attempts. The run ends when the tests pass or ...
{ "task_id": "BigCodeBench/1", "family": "LIB", "split": "train", "card_id": "train-BigCodeBench/1-5", "S": 36, "P": 2200, "T_med_s": 6.8508161234, "T_q10_s": 5.4806528987, "T_q90_s": 8.5635201543, "TOK_med": 207, "ATT_med": 1, "PFP": 0.8261506276, "overhead_med_s": 1.1008161234, "n_runs": 8...
measured
v2-latency
train
bigcode/bigcodebench
null
null
ef4e27e24ab7e55a5a5a24de16855dda5efd14605adb85b77016467bd92c22cd
BigCodeBench/1
LIB
m1
human
h01
sc01
REF
[ { "role": "system", "content": "You are a Python coding agent.\nHow a run works: you read the task and write a complete solution, then the harness runs the task's tests. If a test fails, you see the failure and write a corrected full solution. You get at most 3 attempts. The run ends when the tests pass or ...
{}
base-self-distillation
v1
train
bigcode/bigcodebench
null
null
06c1063cd7d7f913fc248b673c21677aeb8d0ef57d1ca490b304ff9704808932
BigCodeBench/100
LIB
m1
self
q20
sc01
train-BigCodeBench/100-0
[ { "role": "system", "content": "You are a Python coding agent.\nHow a run works: you read the task and write a complete solution, then the harness runs the task's tests. If a test fails, you see the failure and write a corrected full solution. You get at most 3 attempts. The run ends when the tests pass or ...
{ "task_id": "BigCodeBench/100", "family": "LIB", "split": "train", "card_id": "train-BigCodeBench/100-0", "S": 24, "P": 240, "T_med_s": 43.8714258548, "T_q10_s": 35.0971406839, "T_q90_s": 54.8392823185, "TOK_med": 815.5, "ATT_med": 3, "PFP": 0.0261506276, "overhead_med_s": 9.8922591882, "n_...
measured
v2-latency
train
bigcode/bigcodebench
null
null
91b9cac0dbe1da6f4b5a38409c1d33681895622406a2f2912a5f03cbe4ff23f0
BigCodeBench/100
LIB
m1
self
q12
sc01
train-BigCodeBench/100-1
[ { "role": "system", "content": "You are a Python coding agent.\nHow a run works: you read the task and write a complete solution, then the harness runs the task's tests. If a test fails, you see the failure and write a corrected full solution. You get at most 3 attempts. The run ends when the tests pass or ...
{ "task_id": "BigCodeBench/100", "family": "LIB", "split": "train", "card_id": "train-BigCodeBench/100-1", "S": 7.2, "P": 37, "T_med_s": 142.2532201491, "T_q10_s": 113.8025761193, "T_q90_s": 177.8165251864, "TOK_med": 815.5, "ATT_med": 3, "PFP": 0.0261506276, "overhead_med_s": 28.9893312602, ...
measured
v2-latency
train
bigcode/bigcodebench
null
null
d5c9936d08a2161c34aae12d8955e0a675cd0b535ba198d3250844efe19e86c7
BigCodeBench/100
LIB
m1
self
q20
sc05
train-BigCodeBench/100-2
[ { "role": "system", "content": "You are a Python coding agent.\nHow a run works: you read the task and write a complete solution, then the harness runs the task's tests. If a test fails, you see the failure and write a corrected full solution. You get at most 3 attempts. The run ends when the tests pass or ...
{ "task_id": "BigCodeBench/100", "family": "LIB", "split": "train", "card_id": "train-BigCodeBench/100-2", "S": 640, "P": 25000, "T_med_s": 7.6662054255, "T_q10_s": 6.1329643404, "T_q90_s": 9.5827567819, "TOK_med": 815.5, "ATT_med": 3, "PFP": 0.0261506276, "overhead_med_s": 6.3919866755, "n_...
measured
v2-latency
train
bigcode/bigcodebench
null
null
00041534f90e2144af0b9c4885bd7dede40d306c326c1328dd3171b4b2cd0b3a
BigCodeBench/100
LIB
m1
self
q15
sc04
train-BigCodeBench/100-3
[ { "role": "system", "content": "You are a Python coding agent.\nHow a run works: you read the task and write a complete solution, then the harness runs the task's tests. If a test fails, you see the failure and write a corrected full solution. You get at most 3 attempts. The run ends when the tests pass or ...
{ "task_id": "BigCodeBench/100", "family": "LIB", "split": "train", "card_id": "train-BigCodeBench/100-3", "S": 22, "P": 2200, "T_med_s": 43.8082422465, "T_q10_s": 35.0465937972, "T_q90_s": 54.7603028082, "TOK_med": 815.5, "ATT_med": 3, "PFP": 0.0261506276, "overhead_med_s": 6.7400604284, "n...
measured
v2-latency
train
bigcode/bigcodebench
null
null
fcd07e1f48e89e28de831cffcd0d3468bba8d6bf58c721771298232068536a2f
BigCodeBench/100
LIB
m1
self
q13
sc01
train-BigCodeBench/100-4
[ { "role": "system", "content": "You are a Python coding agent.\nHow a run works: you read the task and write a complete solution, then the harness runs the task's tests. If a test fails, you see the failure and write a corrected full solution. You get at most 3 attempts. The run ends when the tests pass or ...
{ "task_id": "BigCodeBench/100", "family": "LIB", "split": "train", "card_id": "train-BigCodeBench/100-4", "S": 3.3, "P": 25, "T_med_s": 287.1113500973, "T_q10_s": 229.6890800778, "T_q90_s": 358.8891876216, "TOK_med": 815.5, "ATT_med": 3, "PFP": 0.0261506276, "overhead_med_s": 39.990137976, ...
measured
v2-latency
train
bigcode/bigcodebench
null
null
ad5ab292fd82ca05c084d82dc01374dd5ffb0df9f3c251946b5a260c1013a067
BigCodeBench/100
LIB
m1
self
q06
sc06
train-BigCodeBench/100-5
[ { "role": "system", "content": "You are a Python coding agent.\nHow a run works: you read the task and write a complete solution, then the harness runs the task's tests. If a test fails, you see the failure and write a corrected full solution. You get at most 3 attempts. The run ends when the tests pass or ...
{ "task_id": "BigCodeBench/100", "family": "LIB", "split": "train", "card_id": "train-BigCodeBench/100-5", "S": 1200, "P": 42000, "T_med_s": 7.0770647932, "T_q10_s": 5.6616518345, "T_q90_s": 8.8463309915, "TOK_med": 815.5, "ATT_med": 3, "PFP": 0.0261506276, "overhead_med_s": 6.3974814599, "n...
measured
v2-latency
train
bigcode/bigcodebench
null
null
cb0010648b9f18821a02c99bc398417fc89c9d2fd1cc7d94a7fb9e7ed047cee7
BigCodeBench/100
LIB
m1
human
h01
sc05
REF
[ { "role": "system", "content": "You are a Python coding agent.\nHow a run works: you read the task and write a complete solution, then the harness runs the task's tests. If a test fails, you see the failure and write a corrected full solution. You get at most 3 attempts. The run ends when the tests pass or ...
{}
base-self-distillation
v1
train
bigcode/bigcodebench
null
null
708400f21ef12f482c84cbb772e35759fe9a3828aafcdd031c35401921a2a034
BigCodeBench/1002
LIB
m1
self
q10
sc04
train-BigCodeBench/1002-0
[ { "role": "system", "content": "You are a Python coding agent.\nHow a run works: you read the task and write a complete solution, then the harness runs the task's tests. If a test fails, you see the failure and write a corrected full solution. You get at most 3 attempts. The run ends when the tests pass or ...
{ "task_id": "BigCodeBench/1002", "family": "LIB", "split": "train", "card_id": "train-BigCodeBench/1002-0", "S": 510, "P": 6100, "T_med_s": 2.7863364484, "T_q10_s": 2.2290691586999998, "T_q90_s": 6.4374865713, "TOK_med": 336.5, "ATT_med": 1, "PFP": 0.6261506276000001, "overhead_med_s": 2.1265...
measured
v2-latency
train
bigcode/bigcodebench
null
null
090537ccd922392639fa1ca86895c1cb7ca75cd10de31ad16acb7dff76607122
BigCodeBench/1002
LIB
m1
self
q09
sc06
train-BigCodeBench/1002-1
[ { "role": "system", "content": "You are a Python coding agent.\nHow a run works: you read the task and write a complete solution, then the harness runs the task's tests. If a test fails, you see the failure and write a corrected full solution. You get at most 3 attempts. The run ends when the tests pass or ...
{ "task_id": "BigCodeBench/1002", "family": "LIB", "split": "train", "card_id": "train-BigCodeBench/1002-1", "S": 30, "P": 770, "T_med_s": 13.5690178011, "T_q10_s": 10.8552142409, "T_q90_s": 32.5874021077, "TOK_med": 336.5, "ATT_med": 1, "PFP": 0.6261506276000001, "overhead_med_s": 2.352351134...
measured
v2-latency
train
bigcode/bigcodebench
null
null
87e02e1e04a5b113168d115f3af472f2d89ef1b50ba0beb32734761a2ef5f6f4
BigCodeBench/1002
LIB
m1
self
q01
sc03
train-BigCodeBench/1002-2
[ { "role": "system", "content": "You are a Python coding agent.\nHow a run works: you read the task and write a complete solution, then the harness runs the task's tests. If a test fails, you see the failure and write a corrected full solution. You get at most 3 attempts. The run ends when the tests pass or ...
{ "task_id": "BigCodeBench/1002", "family": "LIB", "split": "train", "card_id": "train-BigCodeBench/1002-2", "S": 68, "P": 14000, "T_med_s": 7.0566532735, "T_q10_s": 5.6453226188, "T_q90_s": 16.7620524642, "TOK_med": 336.5, "ATT_med": 1, "PFP": 0.6261506276000001, "overhead_med_s": 2.108123861...
measured
v2-latency
train
bigcode/bigcodebench
null
null
cb34dbe3208317e0448e38dd5b3fd589813e216a5fd7292d981e9cb1f29e7986
BigCodeBench/1002
LIB
m1
self
q04
sc02
train-BigCodeBench/1002-3
[ { "role": "system", "content": "You are a Python coding agent.\nHow a run works: you read the task and write a complete solution, then the harness runs the task's tests. If a test fails, you see the failure and write a corrected full solution. You get at most 3 attempts. The run ends when the tests pass or ...
{ "task_id": "BigCodeBench/1002", "family": "LIB", "split": "train", "card_id": "train-BigCodeBench/1002-3", "S": 110, "P": 15000, "T_med_s": 5.1662671518, "T_q10_s": 4.1330137214, "T_q90_s": 12.1888932765, "TOK_med": 336.5, "ATT_med": 1, "PFP": 0.6261506276000001, "overhead_med_s": 2.10717624...
measured
v2-latency
train
bigcode/bigcodebench
null
null
bc7f31c89efa4eaec20f3d6177552a71ac61b229584abb013946876ec58fabef
BigCodeBench/1002
LIB
m1
self
q12
sc05
train-BigCodeBench/1002-4
[ { "role": "system", "content": "You are a Python coding agent.\nHow a run works: you read the task and write a complete solution, then the harness runs the task's tests. If a test fails, you see the failure and write a corrected full solution. You get at most 3 attempts. The run ends when the tests pass or ...
{ "task_id": "BigCodeBench/1002", "family": "LIB", "split": "train", "card_id": "train-BigCodeBench/1002-4", "S": 6.1, "P": 41, "T_med_s": 62.1115025388, "T_q10_s": 49.6892020311, "T_q90_s": 151.363815577, "TOK_med": 336.5, "ATT_med": 1, "PFP": 0.6261506276000001, "overhead_med_s": 6.947568112...
measured
v2-latency
train
bigcode/bigcodebench
null
null
74c3ee9c2bd71ddab116368752939c580204336a333da0ea8d4b3ae2b95f0526
BigCodeBench/1002
LIB
m1
self
q17
sc01
train-BigCodeBench/1002-5
[ { "role": "system", "content": "You are a Python coding agent.\nHow a run works: you read the task and write a complete solution, then the harness runs the task's tests. If a test fails, you see the failure and write a corrected full solution. You get at most 3 attempts. The run ends when the tests pass or ...
{ "task_id": "BigCodeBench/1002", "family": "LIB", "split": "train", "card_id": "train-BigCodeBench/1002-5", "S": 61, "P": 6800, "T_med_s": 7.6395677245, "T_q10_s": 6.1116541796, "T_q90_s": 18.1765592971, "TOK_med": 336.5, "ATT_med": 1, "PFP": 0.6261506276000001, "overhead_med_s": 2.1231742819...
measured
v2-latency
train
bigcode/bigcodebench
null
null
cd0a79a3bccfad2a04564f2c25760535472c5534ce7f2cec53a44e5857bd591d
BigCodeBench/1002
LIB
m1
human
h06
sc02
REF
[ { "role": "system", "content": "You are a Python coding agent.\nHow a run works: you read the task and write a complete solution, then the harness runs the task's tests. If a test fails, you see the failure and write a corrected full solution. You get at most 3 attempts. The run ends when the tests pass or ...
{}
base-self-distillation
v1
train
bigcode/bigcodebench
null
null
5f0d510b60685170c76e67f5ceec7ecb455e7ae7018a0f269ee53020a2fc06e8
BigCodeBench/1003
LIB
m1
self
q06
sc06
train-BigCodeBench/1003-0
[ { "role": "system", "content": "You are a Python coding agent.\nHow a run works: you read the task and write a complete solution, then the harness runs the task's tests. If a test fails, you see the failure and write a corrected full solution. You get at most 3 attempts. The run ends when the tests pass or ...
{ "task_id": "BigCodeBench/1003", "family": "LIB", "split": "train", "card_id": "train-BigCodeBench/1003-0", "S": 4.8, "P": 32, "T_med_s": 263.4776124788, "T_q10_s": 210.7820899831, "T_q90_s": 329.3470155986, "TOK_med": 1106.5, "ATT_med": 3, "PFP": 0.0261506276, "overhead_med_s": 32.9567791455...
measured
v2-latency
train
bigcode/bigcodebench
null
null
5064496118d8a9014eea4c27251f06d615fcd7a7a33d56b4657f4e118f1789d2
BigCodeBench/1003
LIB
m1
self
q06
sc02
train-BigCodeBench/1003-1
[ { "role": "system", "content": "You are a Python coding agent.\nHow a run works: you read the task and write a complete solution, then the harness runs the task's tests. If a test fails, you see the failure and write a corrected full solution. You get at most 3 attempts. The run ends when the tests pass or ...
{ "task_id": "BigCodeBench/1003", "family": "LIB", "split": "train", "card_id": "train-BigCodeBench/1003-1", "S": 2200, "P": 33000, "T_med_s": 4.2769151226, "T_q10_s": 3.421532098, "T_q90_s": 5.3461439032, "TOK_med": 1106.5, "ATT_med": 3, "PFP": 0.0261506276, "overhead_med_s": 3.7739605771, ...
measured
v2-latency
train
bigcode/bigcodebench
null
null
f25fd5dadae8b266bb6744a00700d6d41f32dd9e0b3c53512328b3082343a3a1
BigCodeBench/1003
LIB
m1
self
q08
sc05
train-BigCodeBench/1003-2
[ { "role": "system", "content": "You are a Python coding agent.\nHow a run works: you read the task and write a complete solution, then the harness runs the task's tests. If a test fails, you see the failure and write a corrected full solution. You get at most 3 attempts. The run ends when the tests pass or ...
{ "task_id": "BigCodeBench/1003", "family": "LIB", "split": "train", "card_id": "train-BigCodeBench/1003-2", "S": 1300, "P": 28000, "T_med_s": 4.6225758488, "T_q10_s": 3.698060679, "T_q90_s": 5.778219811, "TOK_med": 1106.5, "ATT_med": 3, "PFP": 0.0261506276, "overhead_med_s": 3.7714220026, "...
measured
v2-latency
train
bigcode/bigcodebench
null
null
ee220dfbfcffa76e8c5d143b013a06076f86afc7d306bd05d1f3565c3bd31d78
BigCodeBench/1003
LIB
m1
self
q17
sc04
train-BigCodeBench/1003-3
[ { "role": "system", "content": "You are a Python coding agent.\nHow a run works: you read the task and write a complete solution, then the harness runs the task's tests. If a test fails, you see the failure and write a corrected full solution. You get at most 3 attempts. The run ends when the tests pass or ...
{ "task_id": "BigCodeBench/1003", "family": "LIB", "split": "train", "card_id": "train-BigCodeBench/1003-3", "S": 33, "P": 3000, "T_med_s": 37.5799988425, "T_q10_s": 30.063999074, "T_q90_s": 46.9749985531, "TOK_med": 1106.5, "ATT_med": 3, "PFP": 0.0261506276, "overhead_med_s": 4.0496958122, ...
measured
v2-latency
train
bigcode/bigcodebench
null
null
de89eddeee0defb8b84f02581b62787f767bf256dff7435d60c051dfdf3e9f33
BigCodeBench/1003
LIB
m1
self
q20
sc06
train-BigCodeBench/1003-4
[ { "role": "system", "content": "You are a Python coding agent.\nHow a run works: you read the task and write a complete solution, then the harness runs the task's tests. If a test fails, you see the failure and write a corrected full solution. You get at most 3 attempts. The run ends when the tests pass or ...
{ "task_id": "BigCodeBench/1003", "family": "LIB", "split": "train", "card_id": "train-BigCodeBench/1003-4", "S": 64, "P": 4000, "T_med_s": 21.2608416455, "T_q10_s": 17.0086733164, "T_q90_s": 26.5760520569, "TOK_med": 1106.5, "ATT_med": 3, "PFP": 0.0261506276, "overhead_med_s": 3.9717791455, ...
measured
v2-latency
train
bigcode/bigcodebench
null
null
21c4bc4beb97648a6f6b4937165d58bdf8152528ce8681e3fa1e0c0045e2252b
BigCodeBench/1003
LIB
m1
self
q05
sc04
train-BigCodeBench/1003-5
[ { "role": "system", "content": "You are a Python coding agent.\nHow a run works: you read the task and write a complete solution, then the harness runs the task's tests. If a test fails, you see the failure and write a corrected full solution. You get at most 3 attempts. The run ends when the tests pass or ...
{ "task_id": "BigCodeBench/1003", "family": "LIB", "split": "train", "card_id": "train-BigCodeBench/1003-5", "S": 360, "P": 41000, "T_med_s": 6.8344451347, "T_q10_s": 5.4675561077, "T_q90_s": 8.5430564183, "TOK_med": 1106.5, "ATT_med": 3, "PFP": 0.0261506276, "overhead_med_s": 3.7608340236, ...
measured
v2-latency
train
bigcode/bigcodebench
null
null
0be6f5ce09b8b418181e89fe8553797fa3bb895ca2070fdbfa6682c671ae9551
BigCodeBench/1003
LIB
m1
human
h08
sc04
REF
[ { "role": "system", "content": "You are a Python coding agent.\nHow a run works: you read the task and write a complete solution, then the harness runs the task's tests. If a test fails, you see the failure and write a corrected full solution. You get at most 3 attempts. The run ends when the tests pass or ...
{}
base-self-distillation
v1
train
bigcode/bigcodebench
null
null
d1397c8872613cd586fb8e4197d0c81763333d86370b9a3a4aaa779b5ff6496f
BigCodeBench/1006
LIB
m1
self
q05
sc03
train-BigCodeBench/1006-0
[ { "role": "system", "content": "You are a Python coding agent.\nHow a run works: you read the task and write a complete solution, then the harness runs the task's tests. If a test fails, you see the failure and write a corrected full solution. You get at most 3 attempts. The run ends when the tests pass or ...
{ "task_id": "BigCodeBench/1006", "family": "LIB", "split": "train", "card_id": "train-BigCodeBench/1006-0", "S": 330, "P": 29000, "T_med_s": 6.9139686822, "T_q10_s": 5.5311749458, "T_q90_s": 8.6424608527, "TOK_med": 1045, "ATT_med": 3, "PFP": 0.0261506276, "overhead_med_s": 3.7473020155, "n...
measured
v2-latency
train
bigcode/bigcodebench
null
null
91596ed36d24ed3421b4a75bb8890d9c2be55d01bf347b154979386018cd3fbb
BigCodeBench/1006
LIB
m1
self
q09
sc01
train-BigCodeBench/1006-1
[ { "role": "system", "content": "You are a Python coding agent.\nHow a run works: you read the task and write a complete solution, then the harness runs the task's tests. If a test fails, you see the failure and write a corrected full solution. You get at most 3 attempts. The run ends when the tests pass or ...
{ "task_id": "BigCodeBench/1006", "family": "LIB", "split": "train", "card_id": "train-BigCodeBench/1006-1", "S": 2000, "P": 40000, "T_med_s": 4.26015805, "T_q10_s": 3.40812644, "T_q90_s": 5.3251975625, "TOK_med": 1045, "ATT_med": 3, "PFP": 0.0261506276, "overhead_med_s": 3.7376580500000003, ...
measured
v2-latency
train
bigcode/bigcodebench
null
null
5867d50c0ee7e87bd09cde86401a56b21fa9e0683fd4ca7e739b0705b5cfc519
BigCodeBench/1006
LIB
m1
self
q16
sc04
train-BigCodeBench/1006-2
[ { "role": "system", "content": "You are a Python coding agent.\nHow a run works: you read the task and write a complete solution, then the harness runs the task's tests. If a test fails, you see the failure and write a corrected full solution. You get at most 3 attempts. The run ends when the tests pass or ...
{ "task_id": "BigCodeBench/1006", "family": "LIB", "split": "train", "card_id": "train-BigCodeBench/1006-2", "S": 110, "P": 3400, "T_med_s": 13.5115605644, "T_q10_s": 10.8092484515, "T_q90_s": 16.8894507054, "TOK_med": 1045, "ATT_med": 3, "PFP": 0.0261506276, "overhead_med_s": 4.0115605644, ...
measured
v2-latency
train
bigcode/bigcodebench
null
null
6fca0189389cc3b0ff7d6076f5c15b5c9fd5e14b33324a864ece1c07072f9aef
BigCodeBench/1006
LIB
m1
self
q10
sc01
train-BigCodeBench/1006-3
[ { "role": "system", "content": "You are a Python coding agent.\nHow a run works: you read the task and write a complete solution, then the harness runs the task's tests. If a test fails, you see the failure and write a corrected full solution. You get at most 3 attempts. The run ends when the tests pass or ...
{ "task_id": "BigCodeBench/1006", "family": "LIB", "split": "train", "card_id": "train-BigCodeBench/1006-3", "S": 530, "P": 6200, "T_med_s": 5.8479634213, "T_q10_s": 4.678370737, "T_q90_s": 7.3099542766, "TOK_med": 1045, "ATT_med": 3, "PFP": 0.0261506276, "overhead_med_s": 3.8762653081, "n_r...
measured
v2-latency
train
bigcode/bigcodebench
null
null
30fcdd11c5d7ae81b4bad5e19beb4829143e2979dee9860622db58efe26de5b8
BigCodeBench/1006
LIB
m1
self
q13
sc04
train-BigCodeBench/1006-4
[ { "role": "system", "content": "You are a Python coding agent.\nHow a run works: you read the task and write a complete solution, then the harness runs the task's tests. If a test fails, you see the failure and write a corrected full solution. You get at most 3 attempts. The run ends when the tests pass or ...
{ "task_id": "BigCodeBench/1006", "family": "LIB", "split": "train", "card_id": "train-BigCodeBench/1006-4", "S": 48, "P": 1700, "T_med_s": 26.08209978, "T_q10_s": 20.865679824, "T_q90_s": 32.602624725, "TOK_med": 1045, "ATT_med": 3, "PFP": 0.0261506276, "overhead_med_s": 4.3112664467, "n_ru...
measured
v2-latency
train
bigcode/bigcodebench
null
null
06a83097d1a7d5d7540af415dd2c73a5ba9fb0c1ea9e3c90441200f5db69beb9
BigCodeBench/1006
LIB
m1
self
q04
sc02
train-BigCodeBench/1006-5
[ { "role": "system", "content": "You are a Python coding agent.\nHow a run works: you read the task and write a complete solution, then the harness runs the task's tests. If a test fails, you see the failure and write a corrected full solution. You get at most 3 attempts. The run ends when the tests pass or ...
{ "task_id": "BigCodeBench/1006", "family": "LIB", "split": "train", "card_id": "train-BigCodeBench/1006-5", "S": 18, "P": 150, "T_med_s": 68.5607435709, "T_q10_s": 54.8485948567, "T_q90_s": 85.7009294636, "TOK_med": 1045, "ATT_med": 3, "PFP": 0.0261506276, "overhead_med_s": 10.5051880153, "...
measured
v2-latency
train
bigcode/bigcodebench
null
null
02efbd7f9139c795c450fad3d3bfdf47cb76c9132d638472e0841316f63cf4e7
BigCodeBench/1006
LIB
m1
human
h07
sc01
REF
[ { "role": "system", "content": "You are a Python coding agent.\nHow a run works: you read the task and write a complete solution, then the harness runs the task's tests. If a test fails, you see the failure and write a corrected full solution. You get at most 3 attempts. The run ends when the tests pass or ...
{}
base-self-distillation
v1
train
bigcode/bigcodebench
null
null
ab1eb9330662ddd6c76b5da75791f0fe80c554a314b0fa5f69c685ae34b64e2d
BigCodeBench/1008
LIB
m1
self
q19
sc06
train-BigCodeBench/1008-0
[ { "role": "system", "content": "You are a Python coding agent.\nHow a run works: you read the task and write a complete solution, then the harness runs the task's tests. If a test fails, you see the failure and write a corrected full solution. You get at most 3 attempts. The run ends when the tests pass or ...
{ "task_id": "BigCodeBench/1008", "family": "LIB", "split": "train", "card_id": "train-BigCodeBench/1008-0", "S": 620, "P": 9900, "T_med_s": 5.5440040399, "T_q10_s": 4.4352032319, "T_q90_s": 6.9300050499, "TOK_med": 602.5, "ATT_med": 3, "PFP": 0.0261506276, "overhead_med_s": 4.5722298463, "n...
measured
v2-latency
train
bigcode/bigcodebench
null
null
81a7e8c05b21a7780a896f02c5dd218d23a1e2fb5ccca5b3cd7b79495df66c56
BigCodeBench/1008
LIB
m1
self
q16
sc04
train-BigCodeBench/1008-1
[ { "role": "system", "content": "You are a Python coding agent.\nHow a run works: you read the task and write a complete solution, then the harness runs the task's tests. If a test fails, you see the failure and write a corrected full solution. You get at most 3 attempts. The run ends when the tests pass or ...
{ "task_id": "BigCodeBench/1008", "family": "LIB", "split": "train", "card_id": "train-BigCodeBench/1008-1", "S": 69, "P": 5900, "T_med_s": 13.3806076554, "T_q10_s": 10.7044861243, "T_q90_s": 17.2525985313, "TOK_med": 602.5, "ATT_med": 3, "PFP": 0.0261506276, "overhead_med_s": 4.6487235974, ...
measured
v2-latency
train
bigcode/bigcodebench
null
null
e237c9e35233544ac4db3236b4b50949756e4646914f201a1802611f4100a932
BigCodeBench/1008
LIB
m1
self
q14
sc05
train-BigCodeBench/1008-2
[ { "role": "system", "content": "You are a Python coding agent.\nHow a run works: you read the task and write a complete solution, then the harness runs the task's tests. If a test fails, you see the failure and write a corrected full solution. You get at most 3 attempts. The run ends when the tests pass or ...
{ "task_id": "BigCodeBench/1008", "family": "LIB", "split": "train", "card_id": "train-BigCodeBench/1008-2", "S": 440, "P": 8500, "T_med_s": 5.96013151, "T_q10_s": 4.768105208, "T_q90_s": 7.4501643875, "TOK_med": 602.5, "ATT_med": 3, "PFP": 0.0261506276, "overhead_med_s": 4.5908133282, "n_ru...
measured
v2-latency
train
bigcode/bigcodebench
null
null
5f39376e288dd3e9832d6a72b402c82cb03396ec238438908f192594da40bd8d
BigCodeBench/1008
LIB
m1
self
q12
sc06
train-BigCodeBench/1008-3
[ { "role": "system", "content": "You are a Python coding agent.\nHow a run works: you read the task and write a complete solution, then the harness runs the task's tests. If a test fails, you see the failure and write a corrected full solution. You get at most 3 attempts. The run ends when the tests pass or ...
{ "task_id": "BigCodeBench/1008", "family": "LIB", "split": "train", "card_id": "train-BigCodeBench/1008-3", "S": 1100, "P": 41000, "T_med_s": 5.0235111024, "T_q10_s": 4.0188088819, "T_q90_s": 6.279388878, "TOK_med": 602.5, "ATT_med": 3, "PFP": 0.0261506276, "overhead_med_s": 4.4757838297, "...
measured
v2-latency
train
bigcode/bigcodebench
null
null
80aeafcb0f2b732e04bfdf66d52d79f465f27778f730849d4716b614ad6c0aea
BigCodeBench/1008
LIB
m1
self
q02
sc04
train-BigCodeBench/1008-4
[ { "role": "system", "content": "You are a Python coding agent.\nHow a run works: you read the task and write a complete solution, then the harness runs the task's tests. If a test fails, you see the failure and write a corrected full solution. You get at most 3 attempts. The run ends when the tests pass or ...
{ "task_id": "BigCodeBench/1008", "family": "LIB", "split": "train", "card_id": "train-BigCodeBench/1008-4", "S": 69, "P": 3600, "T_med_s": 13.5015633993, "T_q10_s": 10.8012507194, "T_q90_s": 17.3699266857, "TOK_med": 602.5, "ATT_med": 3, "PFP": 0.0261506276, "overhead_med_s": 4.7696793413, ...
measured
v2-latency
train
bigcode/bigcodebench
null
null
9414b924fbe1e2116c7be1bf93fc9bd2f39f4c689845af51a86ee2d9c5ff321a
BigCodeBench/1008
LIB
m1
self
q02
sc03
train-BigCodeBench/1008-5
[ { "role": "system", "content": "You are a Python coding agent.\nHow a run works: you read the task and write a complete solution, then the harness runs the task's tests. If a test fails, you see the failure and write a corrected full solution. You get at most 3 attempts. The run ends when the tests pass or ...
{ "task_id": "BigCodeBench/1008", "family": "LIB", "split": "train", "card_id": "train-BigCodeBench/1008-5", "S": 920, "P": 28000, "T_med_s": 5.1409002549, "T_q10_s": 4.1127202039, "T_q90_s": 6.4261253186000005, "TOK_med": 602.5, "ATT_med": 3, "PFP": 0.0261506276, "overhead_med_s": 4.486008950...
measured
v2-latency
train
bigcode/bigcodebench
null
null
8f95e6ed6dd7c043afa749a8ff11d914b400ea610c95037464ae083fa48cc354
BigCodeBench/1008
LIB
m1
human
h05
sc03
REF
[ { "role": "system", "content": "You are a Python coding agent.\nHow a run works: you read the task and write a complete solution, then the harness runs the task's tests. If a test fails, you see the failure and write a corrected full solution. You get at most 3 attempts. The run ends when the tests pass or ...
{}
base-self-distillation
v1
train
bigcode/bigcodebench
null
null
3a56182688d0fbb3d22732aa2bbfa8d89ae79239ca7fad3d350d3bfcc14b4184
BigCodeBench/1013
LIB
m1
self
q20
sc03
train-BigCodeBench/1013-0
[ { "role": "system", "content": "You are a Python coding agent.\nHow a run works: you read the task and write a complete solution, then the harness runs the task's tests. If a test fails, you see the failure and write a corrected full solution. You get at most 3 attempts. The run ends when the tests pass or ...
{ "task_id": "BigCodeBench/1013", "family": "LIB", "split": "train", "card_id": "train-BigCodeBench/1013-0", "S": 400, "P": 54000, "T_med_s": 1.98892987, "T_q10_s": 1.5911438960000002, "T_q90_s": 2.4861623375, "TOK_med": 261, "ATT_med": 1, "PFP": 0.8261506276, "overhead_med_s": 1.33642987, "...
measured
v2-latency
train
bigcode/bigcodebench
null
null
d2d09aa6a1ff1a3c68405b694a7a8fbef8d3008fb0a668a64fdce0d718a08fce
BigCodeBench/1013
LIB
m1
self
q03
sc06
train-BigCodeBench/1013-1
[ { "role": "system", "content": "You are a Python coding agent.\nHow a run works: you read the task and write a complete solution, then the harness runs the task's tests. If a test fails, you see the failure and write a corrected full solution. You get at most 3 attempts. The run ends when the tests pass or ...
{ "task_id": "BigCodeBench/1013", "family": "LIB", "split": "train", "card_id": "train-BigCodeBench/1013-1", "S": 450, "P": 7500, "T_med_s": 1.9322983886, "T_q10_s": 1.5458387108, "T_q90_s": 2.4153729857, "TOK_med": 261, "ATT_med": 1, "PFP": 0.8261506276, "overhead_med_s": 1.3522983886, "n_r...
measured
v2-latency
train
bigcode/bigcodebench
null
null
d0f951d7019351cd3b6e86feb0981bb2bc777c9ab2107291ec6929ca4edf26f3
BigCodeBench/1013
LIB
m1
self
q15
sc01
train-BigCodeBench/1013-2
[ { "role": "system", "content": "You are a Python coding agent.\nHow a run works: you read the task and write a complete solution, then the harness runs the task's tests. If a test fails, you see the failure and write a corrected full solution. You get at most 3 attempts. The run ends when the tests pass or ...
{ "task_id": "BigCodeBench/1013", "family": "LIB", "split": "train", "card_id": "train-BigCodeBench/1013-2", "S": 1600, "P": 39000, "T_med_s": 1.5014681535, "T_q10_s": 1.2011745228, "T_q90_s": 1.8768351919000001, "TOK_med": 261, "ATT_med": 1, "PFP": 0.8261506276, "overhead_med_s": 1.3383431534...
measured
v2-latency
train
bigcode/bigcodebench
null
null
6d3f7602e16c77eef0e85c8424ba688bfac2e521bc19a03491ee055eebf3506d
BigCodeBench/1013
LIB
m1
self
q08
sc06
train-BigCodeBench/1013-3
[ { "role": "system", "content": "You are a Python coding agent.\nHow a run works: you read the task and write a complete solution, then the harness runs the task's tests. If a test fails, you see the failure and write a corrected full solution. You get at most 3 attempts. The run ends when the tests pass or ...
{ "task_id": "BigCodeBench/1013", "family": "LIB", "split": "train", "card_id": "train-BigCodeBench/1013-3", "S": 4.7, "P": 50, "T_med_s": 59.5827578866, "T_q10_s": 47.6662063093, "T_q90_s": 85.6308912913, "TOK_med": 261, "ATT_med": 1, "PFP": 0.8261506276, "overhead_med_s": 4.050842993, "n_r...
measured
v2-latency
train
bigcode/bigcodebench
null
null
04d4459e89b9b8620d4e3981ac2d12d5e4fc664783b3ec067a75dc2211fb4051
BigCodeBench/1013
LIB
m1
self
q09
sc06
train-BigCodeBench/1013-4
[ { "role": "system", "content": "You are a Python coding agent.\nHow a run works: you read the task and write a complete solution, then the harness runs the task's tests. If a test fails, you see the failure and write a corrected full solution. You get at most 3 attempts. The run ends when the tests pass or ...
{ "task_id": "BigCodeBench/1013", "family": "LIB", "split": "train", "card_id": "train-BigCodeBench/1013-4", "S": 99, "P": 8600, "T_med_s": 3.9823861969000003, "T_q10_s": 3.1859089575, "T_q90_s": 5.2253472062, "TOK_med": 261, "ATT_med": 1, "PFP": 0.8261506276, "overhead_med_s": 1.3460225605, ...
measured
v2-latency
train
bigcode/bigcodebench
null
null
1b68f9ac19963206c5db624946e023220089766e9d60d8c98619193cc12d65a8
BigCodeBench/1013
LIB
m1
self
q07
sc02
train-BigCodeBench/1013-5
[ { "role": "system", "content": "You are a Python coding agent.\nHow a run works: you read the task and write a complete solution, then the harness runs the task's tests. If a test fails, you see the failure and write a corrected full solution. You get at most 3 attempts. The run ends when the tests pass or ...
{ "task_id": "BigCodeBench/1013", "family": "LIB", "split": "train", "card_id": "train-BigCodeBench/1013-5", "S": 670, "P": 7400, "T_med_s": 1.7424623167000002, "T_q10_s": 1.3939698533, "T_q90_s": 2.1780778958, "TOK_med": 261, "ATT_med": 1, "PFP": 0.8261506276, "overhead_med_s": 1.3529100779, ...
measured
v2-latency
train
bigcode/bigcodebench
null
null
af4469e20ebc0538af1ea23fe894d2513fa060cde953b36c5711eebffae18f32
BigCodeBench/1013
LIB
m1
human
h04
sc06
REF
[ { "role": "system", "content": "You are a Python coding agent.\nHow a run works: you read the task and write a complete solution, then the harness runs the task's tests. If a test fails, you see the failure and write a corrected full solution. You get at most 3 attempts. The run ends when the tests pass or ...
{}
base-self-distillation
v1
train
bigcode/bigcodebench
null
null
72d8837e39cc1b59eea0e912e5b2bddaeec97b1c1f93d22ee8d7721ffc30417f
BigCodeBench/1015
LIB
m1
self
q15
sc03
train-BigCodeBench/1015-0
[ { "role": "system", "content": "You are a Python coding agent.\nHow a run works: you read the task and write a complete solution, then the harness runs the task's tests. If a test fails, you see the failure and write a corrected full solution. You get at most 3 attempts. The run ends when the tests pass or ...
{ "task_id": "BigCodeBench/1015", "family": "LIB", "split": "train", "card_id": "train-BigCodeBench/1015-0", "S": 37, "P": 4500, "T_med_s": 49.9605738744, "T_q10_s": 39.9684590995, "T_q90_s": 62.450717343, "TOK_med": 1682.5, "ATT_med": 3, "PFP": 0.0261506276, "overhead_med_s": 4.48760090140000...
measured
v2-latency
train
bigcode/bigcodebench
null
null
87b9c08e765b7369610a4b9279073b1a39eccfdd76ee498a028760d3f641f961
BigCodeBench/1015
LIB
m1
self
q13
sc05
train-BigCodeBench/1015-1
[ { "role": "system", "content": "You are a Python coding agent.\nHow a run works: you read the task and write a complete solution, then the harness runs the task's tests. If a test fails, you see the failure and write a corrected full solution. You get at most 3 attempts. The run ends when the tests pass or ...
{ "task_id": "BigCodeBench/1015", "family": "LIB", "split": "train", "card_id": "train-BigCodeBench/1015-1", "S": 48, "P": 1000, "T_med_s": 40.2939178983, "T_q10_s": 32.2351343187, "T_q90_s": 50.3673973729, "TOK_med": 1682.5, "ATT_med": 3, "PFP": 0.0261506276, "overhead_med_s": 5.241834565, ...
measured
v2-latency
train
bigcode/bigcodebench
null
null
0f8fd3fd85b85c76aea491aa5f1b58aac24ae12ed3147dba369783d99a01c888
BigCodeBench/1015
LIB
m1
self
q06
sc05
train-BigCodeBench/1015-2
[ { "role": "system", "content": "You are a Python coding agent.\nHow a run works: you read the task and write a complete solution, then the harness runs the task's tests. If a test fails, you see the failure and write a corrected full solution. You get at most 3 attempts. The run ends when the tests pass or ...
{ "task_id": "BigCodeBench/1015", "family": "LIB", "split": "train", "card_id": "train-BigCodeBench/1015-2", "S": 2100, "P": 60000, "T_med_s": 5.0979549311, "T_q10_s": 4.0783639449, "T_q90_s": 6.3724436639, "TOK_med": 1682.5, "ATT_med": 3, "PFP": 0.0261506276, "overhead_med_s": 4.2967644549, ...
measured
v2-latency
train
bigcode/bigcodebench
null
null
0646823fa67a49f9f00e45ae26bfc4f680232f964ef3a397f9b90c91f79538be
BigCodeBench/1015
LIB
m1
self
q05
sc03
train-BigCodeBench/1015-3
[ { "role": "system", "content": "You are a Python coding agent.\nHow a run works: you read the task and write a complete solution, then the harness runs the task's tests. If a test fails, you see the failure and write a corrected full solution. You get at most 3 attempts. The run ends when the tests pass or ...
{ "task_id": "BigCodeBench/1015", "family": "LIB", "split": "train", "card_id": "train-BigCodeBench/1015-3", "S": 140, "P": 11000, "T_med_s": 16.374604509, "T_q10_s": 13.0996836072, "T_q90_s": 20.4682556362, "TOK_med": 1682.5, "ATT_med": 3, "PFP": 0.0261506276, "overhead_med_s": 4.356747366100...
measured
v2-latency
train
bigcode/bigcodebench
null
null
44f602ee353de04f453751a7b23364d82251e41b856f95468fb81cf1de1065a4
BigCodeBench/1015
LIB
m1
self
q05
sc01
train-BigCodeBench/1015-4
[ { "role": "system", "content": "You are a Python coding agent.\nHow a run works: you read the task and write a complete solution, then the harness runs the task's tests. If a test fails, you see the failure and write a corrected full solution. You get at most 3 attempts. The run ends when the tests pass or ...
{ "task_id": "BigCodeBench/1015", "family": "LIB", "split": "train", "card_id": "train-BigCodeBench/1015-4", "S": 330, "P": 54000, "T_med_s": 9.3830950092, "T_q10_s": 7.5064760074, "T_q90_s": 11.7288687615, "TOK_med": 1682.5, "ATT_med": 3, "PFP": 0.0261506276, "overhead_med_s": 4.2846101607, ...
measured
v2-latency
train
bigcode/bigcodebench
null
null
52bd96456a9378995a4354ad115f8128f52d4a0bf08aa004ba03c8225113d8bb
BigCodeBench/1015
LIB
m1
self
q02
sc06
train-BigCodeBench/1015-5
[ { "role": "system", "content": "You are a Python coding agent.\nHow a run works: you read the task and write a complete solution, then the harness runs the task's tests. If a test fails, you see the failure and write a corrected full solution. You get at most 3 attempts. The run ends when the tests pass or ...
{ "task_id": "BigCodeBench/1015", "family": "LIB", "split": "train", "card_id": "train-BigCodeBench/1015-5", "S": 400, "P": 5800, "T_med_s": 8.6345496766, "T_q10_s": 6.9076397413, "T_q90_s": 10.7931870957, "TOK_med": 1682.5, "ATT_med": 3, "PFP": 0.0261506276, "overhead_med_s": 4.4282996766, ...
measured
v2-latency
train
bigcode/bigcodebench
null
null
4b886b9bc3b5675a0527a0b18e663a990dbb431c4ad377cd84f39c42208d4f6d
BigCodeBench/1015
LIB
m1
human
h03
sc01
REF
[ { "role": "system", "content": "You are a Python coding agent.\nHow a run works: you read the task and write a complete solution, then the harness runs the task's tests. If a test fails, you see the failure and write a corrected full solution. You get at most 3 attempts. The run ends when the tests pass or ...
{}
base-self-distillation
v1
train
bigcode/bigcodebench
null
null
31f0eb8685813ad1504adcaf4d9f5abc43a30af4f4d124e9fd8cf1e83f3d654e
BigCodeBench/1022
LIB
m1
self
q14
sc04
train-BigCodeBench/1022-0
[ { "role": "system", "content": "You are a Python coding agent.\nHow a run works: you read the task and write a complete solution, then the harness runs the task's tests. If a test fails, you see the failure and write a corrected full solution. You get at most 3 attempts. The run ends when the tests pass or ...
{ "task_id": "BigCodeBench/1022", "family": "LIB", "split": "train", "card_id": "train-BigCodeBench/1022-0", "S": 58, "P": 2800, "T_med_s": 23.2035723206, "T_q10_s": 18.5628578565, "T_q90_s": 29.0044654007, "TOK_med": 1118.5, "ATT_med": 3, "PFP": 0.0261506276, "overhead_med_s": 3.9190895619, ...
measured
v2-latency
train
bigcode/bigcodebench
null
null
5f82b7d8ff607759e640de21d762616d0943d4220046e42281cd40439f856805
BigCodeBench/1022
LIB
m1
self
q14
sc04
train-BigCodeBench/1022-1
[ { "role": "system", "content": "You are a Python coding agent.\nHow a run works: you read the task and write a complete solution, then the harness runs the task's tests. If a test fails, you see the failure and write a corrected full solution. You get at most 3 attempts. The run ends when the tests pass or ...
{ "task_id": "BigCodeBench/1022", "family": "LIB", "split": "train", "card_id": "train-BigCodeBench/1022-1", "S": 84, "P": 4300, "T_med_s": 17.142311485, "T_q10_s": 13.713849188, "T_q90_s": 21.4278893563, "TOK_med": 1118.5, "ATT_med": 3, "PFP": 0.0261506276, "overhead_med_s": 3.8268352946, "...
measured
v2-latency
train
bigcode/bigcodebench
null
null
6b5d26392793199a9202f6859bed8e1d1af1cea37f55f9f6fb1cd7006d859e31
BigCodeBench/1022
LIB
m1
self
q15
sc05
train-BigCodeBench/1022-2
[ { "role": "system", "content": "You are a Python coding agent.\nHow a run works: you read the task and write a complete solution, then the harness runs the task's tests. If a test fails, you see the failure and write a corrected full solution. You get at most 3 attempts. The run ends when the tests pass or ...
{ "task_id": "BigCodeBench/1022", "family": "LIB", "split": "train", "card_id": "train-BigCodeBench/1022-2", "S": 15, "P": 110, "T_med_s": 85.9710947356, "T_q10_s": 68.7768757885, "T_q90_s": 107.4638684195, "TOK_med": 1118.5, "ATT_med": 3, "PFP": 0.0261506276, "overhead_med_s": 11.404428069, ...
measured
v2-latency
train
bigcode/bigcodebench
null
null
4105ca23c62ee28ef080a98fc27ef365a870e8acd2724f08a9da452b0a30d709
BigCodeBench/1022
LIB
m1
self
q20
sc01
train-BigCodeBench/1022-3
[ { "role": "system", "content": "You are a Python coding agent.\nHow a run works: you read the task and write a complete solution, then the harness runs the task's tests. If a test fails, you see the failure and write a corrected full solution. You get at most 3 attempts. The run ends when the tests pass or ...
{ "task_id": "BigCodeBench/1022", "family": "LIB", "split": "train", "card_id": "train-BigCodeBench/1022-3", "S": 140, "P": 41000, "T_med_s": 11.6618895715, "T_q10_s": 9.3295116572, "T_q90_s": 14.5773619644, "TOK_med": 1118.5, "ATT_med": 3, "PFP": 0.0261506276, "overhead_med_s": 3.6726038572, ...
measured
v2-latency
train
bigcode/bigcodebench
null
null
0c5f403e558cba86702fe8efc4a86197ff15244cf83addc66e59fa9a8403783a
BigCodeBench/1022
LIB
m1
self
q01
sc03
train-BigCodeBench/1022-4
[ { "role": "system", "content": "You are a Python coding agent.\nHow a run works: you read the task and write a complete solution, then the harness runs the task's tests. If a test fails, you see the failure and write a corrected full solution. You get at most 3 attempts. The run ends when the tests pass or ...
{ "task_id": "BigCodeBench/1022", "family": "LIB", "split": "train", "card_id": "train-BigCodeBench/1022-4", "S": 1500, "P": 31000, "T_med_s": 4.4227126019, "T_q10_s": 3.5381700815, "T_q90_s": 5.5283907524, "TOK_med": 1118.5, "ATT_med": 3, "PFP": 0.0261506276, "overhead_med_s": 3.6770459352, ...
measured
v2-latency
train
bigcode/bigcodebench
null
null
fd8bead212cf12dea7df10df3f327285e57fc8b9ae95c2c935ca966ef3a30eb1
BigCodeBench/1022
LIB
m1
self
q09
sc03
train-BigCodeBench/1022-5
[ { "role": "system", "content": "You are a Python coding agent.\nHow a run works: you read the task and write a complete solution, then the harness runs the task's tests. If a test fails, you see the failure and write a corrected full solution. You get at most 3 attempts. The run ends when the tests pass or ...
{ "task_id": "BigCodeBench/1022", "family": "LIB", "split": "train", "card_id": "train-BigCodeBench/1022-5", "S": 850, "P": 28000, "T_med_s": 4.9975981406, "T_q10_s": 3.9980785124000002, "T_q90_s": 6.2469976757, "TOK_med": 1118.5, "ATT_med": 3, "PFP": 0.0261506276, "overhead_med_s": 3.68171578...
measured
v2-latency
train
bigcode/bigcodebench
null
null
1ffa5f6a6a55738495ebeda71beff0654bd1bef21088360dea7800978631b8c1
BigCodeBench/1022
LIB
m1
human
h07
sc01
REF
[ { "role": "system", "content": "You are a Python coding agent.\nHow a run works: you read the task and write a complete solution, then the harness runs the task's tests. If a test fails, you see the failure and write a corrected full solution. You get at most 3 attempts. The run ends when the tests pass or ...
{}
base-self-distillation
v1
train
bigcode/bigcodebench
null
null
0b895615d81c3fe52ac9331bfcb4e9d86ce673465bb17609ef38f59a3f2cf563
BigCodeBench/1035
LIB
m1
self
q05
sc02
train-BigCodeBench/1035-0
[ { "role": "system", "content": "You are a Python coding agent.\nHow a run works: you read the task and write a complete solution, then the harness runs the task's tests. If a test fails, you see the failure and write a corrected full solution. You get at most 3 attempts. The run ends when the tests pass or ...
{ "task_id": "BigCodeBench/1035", "family": "LIB", "split": "train", "card_id": "train-BigCodeBench/1035-0", "S": 16, "P": 110, "T_med_s": 72.9257477378, "T_q10_s": 58.3405981902, "T_q90_s": 107.5843296844, "TOK_med": 942.5, "ATT_med": 3, "PFP": 0.0261506276, "overhead_med_s": 14.0194977378, ...
measured
v2-latency
train
bigcode/bigcodebench
null
null
718da2806b8300819ebcba1bbe4da6894fe628430098ae43958cad1a16251524
BigCodeBench/1035
LIB
m1
self
q15
sc05
train-BigCodeBench/1035-1
[ { "role": "system", "content": "You are a Python coding agent.\nHow a run works: you read the task and write a complete solution, then the harness runs the task's tests. If a test fails, you see the failure and write a corrected full solution. You get at most 3 attempts. The run ends when the tests pass or ...
{ "task_id": "BigCodeBench/1035", "family": "LIB", "split": "train", "card_id": "train-BigCodeBench/1035-1", "S": 80, "P": 13000, "T_med_s": 18.1343526329, "T_q10_s": 14.5074821063, "T_q90_s": 25.3977464676, "TOK_med": 942.5, "ATT_med": 3, "PFP": 0.0261506276, "overhead_med_s": 6.3531026329, ...
measured
v2-latency
train
bigcode/bigcodebench
null
null
c62248ef9b827ed7823ac4c0881ea151a6b40df2676d9c9c9c30b1603dcf0a85
BigCodeBench/1035
LIB
m1
self
q17
sc06
train-BigCodeBench/1035-2
[ { "role": "system", "content": "You are a Python coding agent.\nHow a run works: you read the task and write a complete solution, then the harness runs the task's tests. If a test fails, you see the failure and write a corrected full solution. You get at most 3 attempts. The run ends when the tests pass or ...
{ "task_id": "BigCodeBench/1035", "family": "LIB", "split": "train", "card_id": "train-BigCodeBench/1035-2", "S": 1300, "P": 41000, "T_med_s": 7.5987377286, "T_q10_s": 6.0789901829, "T_q90_s": 9.4984221608, "TOK_med": 942.5, "ATT_med": 3, "PFP": 0.0261506276, "overhead_med_s": 6.8737377286, ...
measured
v2-latency
train
bigcode/bigcodebench
null
null
9e94085af2e839460ecf47e726ca9984630ea4c47f4c2e47ae84576e11c5be4d
BigCodeBench/1035
LIB
m1
self
q10
sc02
train-BigCodeBench/1035-3
[ { "role": "system", "content": "You are a Python coding agent.\nHow a run works: you read the task and write a complete solution, then the harness runs the task's tests. If a test fails, you see the failure and write a corrected full solution. You get at most 3 attempts. The run ends when the tests pass or ...
{ "task_id": "BigCodeBench/1035", "family": "LIB", "split": "train", "card_id": "train-BigCodeBench/1035-3", "S": 530, "P": 11000, "T_med_s": 8.6212841905, "T_q10_s": 6.8970273524, "T_q90_s": 10.7766052381, "TOK_med": 942.5, "ATT_med": 3, "PFP": 0.0261506276, "overhead_med_s": 6.8429823037, ...
measured
v2-latency
train
bigcode/bigcodebench
null
null
4ed86a4e732da3b4aee49f73cc39456baaadddbdb1d4a42f1fcd249860596791
BigCodeBench/1035
LIB
m1
self
q09
sc03
train-BigCodeBench/1035-4
[ { "role": "system", "content": "You are a Python coding agent.\nHow a run works: you read the task and write a complete solution, then the harness runs the task's tests. If a test fails, you see the failure and write a corrected full solution. You get at most 3 attempts. The run ends when the tests pass or ...
{ "task_id": "BigCodeBench/1035", "family": "LIB", "split": "train", "card_id": "train-BigCodeBench/1035-4", "S": 2.8, "P": 21, "T_med_s": 383.3948224131, "T_q10_s": 306.7158579305, "T_q90_s": 578.1455959182, "TOK_med": 942.5, "ATT_med": 3, "PFP": 0.0261506276, "overhead_med_s": 46.787679556, ...
measured
v2-latency
train
bigcode/bigcodebench
null
null
f77f6c3fac78ae7353dca5ffc444c6903f288bb292c04e799f900f74a5e6e684
BigCodeBench/1035
LIB
m1
self
q13
sc01
train-BigCodeBench/1035-5
[ { "role": "system", "content": "You are a Python coding agent.\nHow a run works: you read the task and write a complete solution, then the harness runs the task's tests. If a test fails, you see the failure and write a corrected full solution. You get at most 3 attempts. The run ends when the tests pass or ...
{ "task_id": "BigCodeBench/1035", "family": "LIB", "split": "train", "card_id": "train-BigCodeBench/1035-5", "S": 110, "P": 28000, "T_med_s": 14.8862363742, "T_q10_s": 11.9089890993, "T_q90_s": 20.3345848792, "TOK_med": 942.5, "ATT_med": 3, "PFP": 0.0261506276, "overhead_med_s": 6.318054556, ...
measured
v2-latency
train
bigcode/bigcodebench
null
null
End of preview. Expand in Data Studio

anthropochrony-sft: Fine-tuning sets

Part of Anthropochrony: Measuring and Correcting Human-Time Bias in LLM Self-Estimates (Track 1, problem #8). Code, article and reproduction: https://github.com/isHeSatoshi/anthropochrony

Contents. The M1 (measured, curated), M2 (teacher-guessed) and M3 (measured, uncurated) supervised sets: chat messages, label, label source, frame (self or human), speed card id, and the drop log with reasons. M1 has 6,077 training examples (5,214 self-frame, 863 human-frame).

Restrictions. User turns for restricted sources contain the marker {{TASK_TURN}} instead of the statement; human-frame answers for those rows are cut to their first line.

Licenses. Our annotations, labels and run data: CC-BY-NC-4.0 (they derive from Tiny Aya Global outputs and include CC BY-NC 4.0 ClassEval). Upstream task sources keep their own licenses: HumanEval+/MBPP+ Apache-2.0 (HumanEval MIT, MBPP CC-BY-4.0), BigCodeBench Apache-2.0, ClassEval CC BY-NC 4.0, CodeContests CC-BY-4.0, LeetCodeDataset Apache-2.0 metadata with LeetCode-owned statements, TACO Apache-2.0 metadata with third-party statements. Model-generated text is from CohereLabs/tiny-aya-global (CC-BY-NC-4.0 plus the AUP). Teacher guesses are from DeepSeek-V4-Pro.

Intended use. Research on model self-knowledge and time estimation for coding agents. Not for commercial use.

Limitations. One training seed; labels are medians of 8 sampled runs under stated speeds, not measurements on the stated hardware; Python tasks only; the parser audit covered 60 rows and was done by the research agent.

Collection. https://huggingface.co/collections/isHeSatoshi/anthropochrony-data-6af31e776d6f33265b0c33a8

Downloads last month
29

Collection including isHeSatoshi/anthropochrony-sft