Datasets:
dataset large_stringclasses 1
value | instance_id int64 0 999 | library large_stringclasses 7
values | perturbation_type large_stringclasses 4
values | sample int64 0 99 | model large_stringclasses 3
values | thinking_mode large_stringclasses 1
value | temp large_stringclasses 1
value | role large_stringclasses 1
value | raw_text large_stringlengths 910 142k | extracted_code large_stringlengths 780 142k | parse_ok bool 2
classes | n_tests int64 0 11 | finish_reason large_stringclasses 2
values | n_total_tokens int64 312 8.19k | sampling_params large_stringclasses 1
value |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
ds1000 | 0 | Pandas | Origin | 0 | qwen3-14b-nothink | nothink | 0.6 | tests_py | ```json
{
"tests": [
{
"setup": "import pandas as pd\nimport numpy as np\ndf = pd.DataFrame({'Col1': [1, 4, 7, 10, 13, 16],\n 'Col2': [2, 5, 8, 11, 14, 17],\n 'Col3': [3, 6, 9, 12, 15, 18],\n 'Type': [1, 1, 2, 2, 3, 3]})\nList = [2, 4, 0, 3, 1, 5]",
... | {"tests": [{"setup": "import pandas as pd\nimport numpy as np\ndf = pd.DataFrame({'Col1': [1, 4, 7, 10, 13, 16],\n 'Col2': [2, 5, 8, 11, 14, 17],\n 'Col3': [3, 6, 9, 12, 15, 18],\n 'Type': [1, 1, 2, 2, 3, 3]})\nList = [2, 4, 0, 3, 1, 5]", "check": "expected = pd.Dat... | true | 10 | stop | 2,964 | {"n": 100, "temperature": 0.6, "top_p": 0.8, "top_k": 20, "min_p": 0.0, "max_tokens": 8192, "seed": 42, "enable_thinking": false, "dtype": "bfloat16"} |
ds1000 | 0 | Pandas | Origin | 1 | qwen3-14b-nothink | nothink | 0.6 | tests_py | ```json
{
"tests": [
{
"setup": "import pandas as pd\nimport numpy as np\ndf = pd.DataFrame({'Col1': [1, 4, 7, 10, 13, 16],\n 'Col2': [2, 5, 8, 11, 14, 17],\n 'Col3': [3, 6, 9, 12, 15, 18],\n 'Type': [1, 1, 2, 2, 3, 3]})\nList = [2, 4, 0, 3, 1, 5]",
... | {"tests": [{"setup": "import pandas as pd\nimport numpy as np\ndf = pd.DataFrame({'Col1': [1, 4, 7, 10, 13, 16],\n 'Col2': [2, 5, 8, 11, 14, 17],\n 'Col3': [3, 6, 9, 12, 15, 18],\n 'Type': [1, 1, 2, 2, 3, 3]})\nList = [2, 4, 0, 3, 1, 5]", "check": "expected = pd.Dat... | true | 10 | stop | 2,964 | {"n": 100, "temperature": 0.6, "top_p": 0.8, "top_k": 20, "min_p": 0.0, "max_tokens": 8192, "seed": 42, "enable_thinking": false, "dtype": "bfloat16"} |
ds1000 | 0 | Pandas | Origin | 2 | qwen3-14b-nothink | nothink | 0.6 | tests_py | ```json
{
"tests": [
{
"setup": "import pandas as pd\nimport numpy as np\ndf = pd.DataFrame({'Col1': [1, 4, 7, 10, 13, 16],\n 'Col2': [2, 5, 8, 11, 14, 17],\n 'Col3': [3, 6, 9, 12, 15, 18],\n 'Type': [1, 1, 2, 2, 3, 3]})\nList = [2, 4, 0, 3, 1, 5]",
... | {"tests": [{"setup": "import pandas as pd\nimport numpy as np\ndf = pd.DataFrame({'Col1': [1, 4, 7, 10, 13, 16],\n 'Col2': [2, 5, 8, 11, 14, 17],\n 'Col3': [3, 6, 9, 12, 15, 18],\n 'Type': [1, 1, 2, 2, 3, 3]})\nList = [2, 4, 0, 3, 1, 5]", "check": "expected = pd.Dat... | true | 10 | stop | 2,964 | {"n": 100, "temperature": 0.6, "top_p": 0.8, "top_k": 20, "min_p": 0.0, "max_tokens": 8192, "seed": 42, "enable_thinking": false, "dtype": "bfloat16"} |
ds1000 | 0 | Pandas | Origin | 3 | qwen3-14b-nothink | nothink | 0.6 | tests_py | ```json
{
"tests": [
{
"setup": "import pandas as pd\nimport numpy as np\ndf = pd.DataFrame({'Col1': [1, 4, 7, 10, 13, 16], 'Col2': [2, 5, 8, 11, 14, 17], 'Col3': [3, 6, 9, 12, 15, 18], 'Type': [1, 1, 2, 2, 3, 3]})\nList = [2, 4, 0, 3, 1, 5]",
"check": "expected = pd.DataFrame({'Col1': [7, 13, 1, 10, ... | {"tests": [{"setup": "import pandas as pd\nimport numpy as np\ndf = pd.DataFrame({'Col1': [1, 4, 7, 10, 13, 16], 'Col2': [2, 5, 8, 11, 14, 17], 'Col3': [3, 6, 9, 12, 15, 18], 'Type': [1, 1, 2, 2, 3, 3]})\nList = [2, 4, 0, 3, 1, 5]", "check": "expected = pd.DataFrame({'Col1': [7, 13, 1, 10, 4, 16], 'Col2': [8, 14, 2, 11... | true | 10 | stop | 2,794 | {"n": 100, "temperature": 0.6, "top_p": 0.8, "top_k": 20, "min_p": 0.0, "max_tokens": 8192, "seed": 42, "enable_thinking": false, "dtype": "bfloat16"} |
ds1000 | 0 | Pandas | Origin | 4 | qwen3-14b-nothink | nothink | 0.6 | tests_py | ```json
{
"tests": [
{
"setup": "import pandas as pd\nimport numpy as np\ndf = pd.DataFrame({'Col1': [1, 4, 7, 10, 13, 16],\n 'Col2': [2, 5, 8, 11, 14, 17],\n 'Col3': [3, 6, 9, 12, 15, 18],\n 'Type': [1, 1, 2, 2, 3, 3]})\nList = [2, 4, 0, 3, 1, 5]",
... | {"tests": [{"setup": "import pandas as pd\nimport numpy as np\ndf = pd.DataFrame({'Col1': [1, 4, 7, 10, 13, 16],\n 'Col2': [2, 5, 8, 11, 14, 17],\n 'Col3': [3, 6, 9, 12, 15, 18],\n 'Type': [1, 1, 2, 2, 3, 3]})\nList = [2, 4, 0, 3, 1, 5]", "check": "expected = pd.Dat... | true | 10 | stop | 2,964 | {"n": 100, "temperature": 0.6, "top_p": 0.8, "top_k": 20, "min_p": 0.0, "max_tokens": 8192, "seed": 42, "enable_thinking": false, "dtype": "bfloat16"} |
ds1000 | 0 | Pandas | Origin | 5 | qwen3-14b-nothink | nothink | 0.6 | tests_py | ```json
{
"tests": [
{
"setup": "import pandas as pd\nimport numpy as np\ndf = pd.DataFrame({'Col1': [1, 4, 7, 10, 13, 16],\n 'Col2': [2, 5, 8, 11, 14, 17],\n 'Col3': [3, 6, 9, 12, 15, 18],\n 'Type': [1, 1, 2, 2, 3, 3]})\nList = [2, 4, 0, 3, 1, 5]",
... | {"tests": [{"setup": "import pandas as pd\nimport numpy as np\ndf = pd.DataFrame({'Col1': [1, 4, 7, 10, 13, 16],\n 'Col2': [2, 5, 8, 11, 14, 17],\n 'Col3': [3, 6, 9, 12, 15, 18],\n 'Type': [1, 1, 2, 2, 3, 3]})\nList = [2, 4, 0, 3, 1, 5]", "check": "expected = pd.Dat... | true | 10 | stop | 2,964 | {"n": 100, "temperature": 0.6, "top_p": 0.8, "top_k": 20, "min_p": 0.0, "max_tokens": 8192, "seed": 42, "enable_thinking": false, "dtype": "bfloat16"} |
ds1000 | 0 | Pandas | Origin | 6 | qwen3-14b-nothink | nothink | 0.6 | tests_py | ```json
{
"tests": [
{
"setup": "import pandas as pd\nimport numpy as np\ndf = pd.DataFrame({'Col1': [1, 4, 7, 10, 13, 16],\n 'Col2': [2, 5, 8, 11, 14, 17],\n 'Col3': [3, 6, 9, 12, 15, 18],\n 'Type': [1, 1, 2, 2, 3, 3]})\nList = [2, 4, 0, 3, 1, 5]",
... | {"tests": [{"setup": "import pandas as pd\nimport numpy as np\ndf = pd.DataFrame({'Col1': [1, 4, 7, 10, 13, 16],\n 'Col2': [2, 5, 8, 11, 14, 17],\n 'Col3': [3, 6, 9, 12, 15, 18],\n 'Type': [1, 1, 2, 2, 3, 3]})\nList = [2, 4, 0, 3, 1, 5]", "check": "expected = pd.Dat... | true | 10 | stop | 2,964 | {"n": 100, "temperature": 0.6, "top_p": 0.8, "top_k": 20, "min_p": 0.0, "max_tokens": 8192, "seed": 42, "enable_thinking": false, "dtype": "bfloat16"} |
ds1000 | 0 | Pandas | Origin | 7 | qwen3-14b-nothink | nothink | 0.6 | tests_py | ```json
{
"tests": [
{
"setup": "import pandas as pd\nimport numpy as np\ndf = pd.DataFrame({'Col1': [1, 4, 7, 10, 13, 16],\n 'Col2': [2, 5, 8, 11, 14, 17],\n 'Col3': [3, 6, 9, 12, 15, 18],\n 'Type': [1, 1, 2, 2, 3, 3]})\nList = [2, 4, 0, 3, 1, 5]",
... | {"tests": [{"setup": "import pandas as pd\nimport numpy as np\ndf = pd.DataFrame({'Col1': [1, 4, 7, 10, 13, 16],\n 'Col2': [2, 5, 8, 11, 14, 17],\n 'Col3': [3, 6, 9, 12, 15, 18],\n 'Type': [1, 1, 2, 2, 3, 3]})\nList = [2, 4, 0, 3, 1, 5]", "check": "expected = pd.Dat... | true | 10 | stop | 2,964 | {"n": 100, "temperature": 0.6, "top_p": 0.8, "top_k": 20, "min_p": 0.0, "max_tokens": 8192, "seed": 42, "enable_thinking": false, "dtype": "bfloat16"} |
ds1000 | 0 | Pandas | Origin | 8 | qwen3-14b-nothink | nothink | 0.6 | tests_py | ```json
{
"tests": [
{
"setup": "import pandas as pd\nimport numpy as np\ndf = pd.DataFrame({'Col1': [1, 4, 7, 10, 13, 16],\n 'Col2': [2, 5, 8, 11, 14, 17],\n 'Col3': [3, 6, 9, 12, 15, 18],\n 'Type': [1, 1, 2, 2, 3, 3]})\nList = [2, 4, 0, 3, 1, 5]",
... | {"tests": [{"setup": "import pandas as pd\nimport numpy as np\ndf = pd.DataFrame({'Col1': [1, 4, 7, 10, 13, 16],\n 'Col2': [2, 5, 8, 11, 14, 17],\n 'Col3': [3, 6, 9, 12, 15, 18],\n 'Type': [1, 1, 2, 2, 3, 3]})\nList = [2, 4, 0, 3, 1, 5]", "check": "expected = pd.Dat... | true | 10 | stop | 2,964 | {"n": 100, "temperature": 0.6, "top_p": 0.8, "top_k": 20, "min_p": 0.0, "max_tokens": 8192, "seed": 42, "enable_thinking": false, "dtype": "bfloat16"} |
ds1000 | 0 | Pandas | Origin | 9 | qwen3-14b-nothink | nothink | 0.6 | tests_py | ```json
{
"tests": [
{
"setup": "import pandas as pd\nimport numpy as np\ndf = pd.DataFrame({'Col1': [1, 4, 7, 10, 13, 16],\n 'Col2': [2, 5, 8, 11, 14, 17],\n 'Col3': [3, 6, 9, 12, 15, 18],\n 'Type': [1, 1, 2, 2, 3, 3]})\nList = [2, 4, 0, 3, 1, 5]",
... | {"tests": [{"setup": "import pandas as pd\nimport numpy as np\ndf = pd.DataFrame({'Col1': [1, 4, 7, 10, 13, 16],\n 'Col2': [2, 5, 8, 11, 14, 17],\n 'Col3': [3, 6, 9, 12, 15, 18],\n 'Type': [1, 1, 2, 2, 3, 3]})\nList = [2, 4, 0, 3, 1, 5]", "check": "expected = pd.Dat... | true | 10 | stop | 2,964 | {"n": 100, "temperature": 0.6, "top_p": 0.8, "top_k": 20, "min_p": 0.0, "max_tokens": 8192, "seed": 42, "enable_thinking": false, "dtype": "bfloat16"} |
ds1000 | 0 | Pandas | Origin | 10 | qwen3-14b-nothink | nothink | 0.6 | tests_py | ```json
{
"tests": [
{
"setup": "import pandas as pd\nimport numpy as np\ndf = pd.DataFrame({'Col1': [1, 4, 7, 10, 13, 16],\n 'Col2': [2, 5, 8, 11, 14, 17],\n 'Col3': [3, 6, 9, 12, 15, 18],\n 'Type': [1, 1, 2, 2, 3, 3]})\nList = [2, 4, 0, 3, 1, 5]",
... | {"tests": [{"setup": "import pandas as pd\nimport numpy as np\ndf = pd.DataFrame({'Col1': [1, 4, 7, 10, 13, 16],\n 'Col2': [2, 5, 8, 11, 14, 17],\n 'Col3': [3, 6, 9, 12, 15, 18],\n 'Type': [1, 1, 2, 2, 3, 3]})\nList = [2, 4, 0, 3, 1, 5]", "check": "expected = pd.Dat... | true | 10 | stop | 2,925 | {"n": 100, "temperature": 0.6, "top_p": 0.8, "top_k": 20, "min_p": 0.0, "max_tokens": 8192, "seed": 42, "enable_thinking": false, "dtype": "bfloat16"} |
ds1000 | 0 | Pandas | Origin | 11 | qwen3-14b-nothink | nothink | 0.6 | tests_py | ```json
{
"tests": [
{
"setup": "import pandas as pd\nimport numpy as np\ndf = pd.DataFrame({'Col1': [1, 4, 7, 10, 13, 16],\n 'Col2': [2, 5, 8, 11, 14, 17],\n 'Col3': [3, 6, 9, 12, 15, 18],\n 'Type': [1, 1, 2, 2, 3, 3]})\nList = [2, 4, 0, 3, 1, 5]",
... | {"tests": [{"setup": "import pandas as pd\nimport numpy as np\ndf = pd.DataFrame({'Col1': [1, 4, 7, 10, 13, 16],\n 'Col2': [2, 5, 8, 11, 14, 17],\n 'Col3': [3, 6, 9, 12, 15, 18],\n 'Type': [1, 1, 2, 2, 3, 3]})\nList = [2, 4, 0, 3, 1, 5]", "check": "expected = pd.Dat... | true | 10 | stop | 2,964 | {"n": 100, "temperature": 0.6, "top_p": 0.8, "top_k": 20, "min_p": 0.0, "max_tokens": 8192, "seed": 42, "enable_thinking": false, "dtype": "bfloat16"} |
ds1000 | 0 | Pandas | Origin | 12 | qwen3-14b-nothink | nothink | 0.6 | tests_py | ```json
{
"tests": [
{
"setup": "import pandas as pd\nimport numpy as np\ndf = pd.DataFrame({'Col1': [1, 4, 7, 10, 13, 16],\n 'Col2': [2, 5, 8, 11, 14, 17],\n 'Col3': [3, 6, 9, 12, 15, 18],\n 'Type': [1, 1, 2, 2, 3, 3]})\nList = [2, 4, 0, 3, 1, 5]",
... | {"tests": [{"setup": "import pandas as pd\nimport numpy as np\ndf = pd.DataFrame({'Col1': [1, 4, 7, 10, 13, 16],\n 'Col2': [2, 5, 8, 11, 14, 17],\n 'Col3': [3, 6, 9, 12, 15, 18],\n 'Type': [1, 1, 2, 2, 3, 3]})\nList = [2, 4, 0, 3, 1, 5]", "check": "expected = pd.Dat... | true | 10 | stop | 2,964 | {"n": 100, "temperature": 0.6, "top_p": 0.8, "top_k": 20, "min_p": 0.0, "max_tokens": 8192, "seed": 42, "enable_thinking": false, "dtype": "bfloat16"} |
ds1000 | 0 | Pandas | Origin | 13 | qwen3-14b-nothink | nothink | 0.6 | tests_py | ```json
{
"tests": [
{
"setup": "import pandas as pd\nimport numpy as np\ndf = pd.DataFrame({'Col1': [1, 4, 7, 10, 13, 16],\n 'Col2': [2, 5, 8, 11, 14, 17],\n 'Col3': [3, 6, 9, 12, 15, 18],\n 'Type': [1, 1, 2, 2, 3, 3]})\nList = [2, 4, 0, 3, 1, 5]",
... | {"tests": [{"setup": "import pandas as pd\nimport numpy as np\ndf = pd.DataFrame({'Col1': [1, 4, 7, 10, 13, 16],\n 'Col2': [2, 5, 8, 11, 14, 17],\n 'Col3': [3, 6, 9, 12, 15, 18],\n 'Type': [1, 1, 2, 2, 3, 3]})\nList = [2, 4, 0, 3, 1, 5]", "check": "expected = pd.Dat... | true | 10 | stop | 2,955 | {"n": 100, "temperature": 0.6, "top_p": 0.8, "top_k": 20, "min_p": 0.0, "max_tokens": 8192, "seed": 42, "enable_thinking": false, "dtype": "bfloat16"} |
DS-1000 test-suite rollouts
Model-generated test suites for all 1000 DS-1000
problems, for the joint coding-and-testing construction (independent code and test experts;
see the mlcb-ocaml campaign).
The matching code rollouts are the domain=ds1000 cells of
samuki-hf/thinking-rollouts
(join on instance_id = DS-1000 test-split row index).
100 rollouts per problem per model: Qwen3-4B / 8B / 14B, non-thinking
(enable_thinking=False), temperature 0.6, top_p 0.8, top_k 20, min_p 0.0, seed 42,
max_tokens 8192, vLLM 0.27.1, bf16.
Test format
The test expert sees only the problem statement. A suite is a JSON object
{"tests": [{"setup": ..., "check": ...}, ...]} (4-10 tests requested):
setup re-creates from scratch the input variables the solution reads (same names as
the problem's setup, own imports, possibly new values); check runs immediately after
the solution in the same namespace and must raise iff result is wrong (for Matplotlib
problems it inspects pyplot state instead). A test executes as
setup + <solution> + check; agreement = no exception. Faithfulness runs the official
reference_code as the solution.
Layout
rollouts/domain=ds1000/role=tests_py/model=<tag>/temp=0.6/data.parquet
Columns
dataset(ds1000),instance_id(DS-1000 test-split row index, 0-999),library,perturbation_type,sample(0-99)model(qwen3-{4b,8b,14b}-nothink),thinking_mode(nothink),temp("0.6"),role(tests_py)raw_text(full model output),extracted_code(canonical suite JSON whenparse_ok, else raw text),parse_ok(last ```json fence parses, schema valid, every snippetast.parses),n_testsfinish_reason,n_total_tokens,sampling_params(JSON)
parse_ok by cell: qwen3-4b 0.944, qwen3-8b 0.961, qwen3-14b 0.971. Invalid rows are
retained; filter on parse_ok.
Faithfulness verdicts
faithfulness/domain=ds1000/model=<tag>/temp=0.6/data.parquet — one row per
(instance_id, sample, test_idx): ref_pass = the test executes cleanly against
DS-1000's official reference_code in the official exec_context frame
(DS-1000 pinned environment, py3.10); error = the exception line otherwise.
A suite is faithful (phi) iff all its tests pass.
| model | per-assertion pass | faithful suites (phi) |
|---|---|---|
| qwen3-14b-nothink | 0.407 | 0.176 |
| qwen3-8b-nothink | 0.304 | 0.100 |
| qwen3-4b-nothink | 0.281 | 0.120 |
Per-library phi ranges from ~0.03 (Sklearn, 4B/8B) to ~0.23 (Numpy, 14B). Failures are content errors (wrong expected literals, mis-predicted dtypes/indices/column structure), not formatting: exact-equality checks on rich objects (DataFrames, arrays) are much harder to write correctly than the stdin/stdout pairs of the OCaml campaign.
- Downloads last month
- 31