SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2007.08095 · NeurIPS · 2020

Synthesize, Execute and Debug: Learning to Repair for Neural Program Synthesis

U Berkeley, Kavi Gupta, Peter Ebert Christensen, Xinyun Chen

arXiv · PDF · Open in the Atlas

Code that ran

16 of 19 samples (functions and classes) lifted from this paper's linked repositories executed on a synthesized input in Syntology's jail (0 of them constructed an object rather than computing a result; 0 honoured the contract we drafted, 0 violated it and 16 have no contract verdict; 0 failures of our instrument; 3 no recorded run), measured 2026-11-21; repository links from the Papers with Code archive attributed under CC BY-SA, no affiliation; bodies inline only under a permissive licence as GitHub detects it. Not a reproduction of the paper's results.

"Ran" means the function executed on a synthesized input and returned a value. It is not a reproduction of the paper's results. This record, dated, as JSON: /api/ran/2007.08095.json · badge: /api/ran/2007.08095.svg

Each repository's role is a label: a classifier's output recorded on the link, or, where no role is recorded, what the Papers with Code archive marked. The line under it names its basis. Neither says who wrote the code.

RepositoryRoleRan
sunblaze-ucb/SED community (classifier label)
classifier basis not recorded on the link
16 ran · 3 no recorded run
FunctionStatusWhere it lives
by_max_acc Ran
ran on a synthesized input
sunblaze-ucb/SED/program_synthesis/analysis/organize_results.py
code served (permissive licence) · get_code("79c5f129bde8c733")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned DataFrame. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import pandas as pd
import numpy as np

# Create a sample DataFrame with the required columns
# The function groups by columns that are NOT 'Step', 'Accuracy', 'Exact'
# So we need at least one grouping column plus the three required columns

data = {
    'Model': ['ModelA', 'ModelA', 'ModelB', 'ModelB'],
    'Dataset': ['DS1', 'DS1', 'DS1', 'DS1'],
    'Step': [100, 200, 150, 250],
    'Accuracy': [0.85, 0.92, 0.88, 0.95],
    'Exact': [0.80, 0.90, 0.85, 0.93]
}

accuracies = pd.DataFrame(data)

ARGS = {
    'accuracies': accuracies
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, pandas — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type DataFrame — a DataFrame output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a DataFrame output the recorded ndim 2 and dtype family float describe that coercion, not the output, and are not stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
by_max_step Ran
ran on a synthesized input
sunblaze-ucb/SED/program_synthesis/analysis/organize_results.py
code served (permissive licence) · get_code("5f53ac6d6dd92573")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned DataFrame. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import pandas as pd
import numpy as np

# Create a sample DataFrame that mimics the expected structure
# The function groups by all columns except 'Step', 'Accuracy', 'Exact'
# and finds the max Step for each group

data = {
    'Step': [1, 2, 3, 1, 2, 3],
    'Accuracy': [0.85, 0.87, 0.89, 0.80, 0.82, 0.84],
    'Exact': [0.75, 0.77, 0.79, 0.70, 0.72, 0.74],
    'Model': ['A', 'A', 'A', 'B', 'B', 'B'],
    'Dataset': ['X', 'X', 'X', 'X', 'X', 'X']
}

accuracies = pd.DataFrame(data)

ARGS = {
    'accuracies': accuracies
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, pandas — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type DataFrame — a DataFrame output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a DataFrame output the recorded ndim 2 and dtype family float describe that coercion, not the output, and are not stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
evaluate_code Ran
ran on a synthesized input
sunblaze-ucb/SED/program_synthesis/datasets/executor.py
code served (permissive licence) · get_code("fef08668e5cc872a")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned dict. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
import torch.nn as nn

class ExecutionResult:
    def __init__(self, result):
        self.result = result

def mock_do_execute(code, arguments, test_input):
    """Mock executor that simulates code execution"""
    # For testing purposes, return a result based on test_input
    if test_input == "valid":
        return ExecutionResult(42)
    elif test_input == "mismatch":
        return ExecutionResult(99)
    else:
        return ExecutionResult(None)

# Define test cases
tests = [
    {'input': 'valid', 'output': 42},
    {'input': 'mismatch', 'output': 100},
    {'input': 'none_result', 'output': None},
]

# Define arguments
ARGS = {
    'code': 'def test_func(x):\n    return x * 2',
    'arguments': {'param1': 10, 'param2': [1, 2, 3]},
    'tests': tests,
    'do_execute': mock_do_execute
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type dict — a dict output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a dict output the recorded ndim 0 and dtype family float describe that coercion, not the output, and are not stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
expand_by_beam Ran
ran on a synthesized input
sunblaze-ucb/SED/program_synthesis/models/beam_search.py
code served (permissive licence) · get_code("00ffb35e6b02915b")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned Tensor, ndim 3, dtype family float, shape (6, 4, 3) as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
import torch.nn as nn

torch.manual_seed(0)

# Analyze the function:
# - v: must be a tensor (has .unsqueeze, .repeat, .dim, .view, .shape methods)
# - beam_size: an integer for repetition factor

# Create a small tensor with batch_size=2, some feature dimensions
v = torch.randn(2, 4, 3)  # shape: (batch=2, feat1=4, feat2=3)
beam_size = 3

ARGS = {
    "v": v,
    "beam_size": beam_size,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Tensor, ndim 3, dtype family float, shape (6, 4, 3) — shape (6, 4, 3) as a later run recorded it
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
get_arg_parser Ran
ran on a synthesized input
sunblaze-ucb/SED/program_synthesis/arguments.py
code served (permissive licence) · get_code("881122c4654562fb")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned ArgumentParser. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
import torch.nn as nn
import argparse
import time

torch.manual_seed(0)

# The function signature is: get_arg_parser(title, mode)
# We need to provide values for both parameters

ARGS = {
    'title': 'Test Parser',
    'mode': 'train'
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type ArgumentParser — a ArgumentParser output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a ArgumentParser output the recorded ndim 0 and dtype family float describe that coercion, not the output, and are not stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
get_attn_mask Ran
ran on a synthesized input
sunblaze-ucb/SED/program_synthesis/models/base.py
code served (permissive licence) · get_code("a09108fab4b8d1b4")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned Tensor, ndim 2, dtype family float, shape (3, 5) as a later run recorded it. The image it ran in was not recorded. The code mentions cuda, and the harness aliases cuda to the CPU for every sample; whether this run needed that is not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch

torch.manual_seed(0)

# The function expects:
# - seq_lengths: a list of integers representing sequence lengths for each batch
# - cuda: a boolean flag for whether to move to cuda

ARGS = {
    "seq_lengths": [3, 5, 2],  # List of sequence lengths for batch_size=3
    "cuda": False  # CPU only
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Tensor, ndim 2, dtype family float, shape (3, 5) — shape (3, 5) as a later run recorded it
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
get_exact_match Ran
ran on a synthesized input
sunblaze-ucb/SED/program_synthesis/analysis/load_results.py
code served (permissive licence) · get_code("7eeeb9e1399f3c5c")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned dict. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import os
import json
import tempfile

# Create a temporary directory and file for testing
temp_dir = tempfile.mkdtemp()
test_file = os.path.join(temp_dir, "test_data.jsonl")

# Create a valid JSONL file with the expected structure
# The function expects lines with 'example' and 'code' keys
test_data = [
    {"example": {"code": "def foo(): pass"}, "code": {"code_sequence": "def foo(): pass"}},  # Header line
    {"example": {"code": "x = 1"}, "code": {"code_sequence": "x = 1"}},  # Exact match
    {"example": {"code": "y = 2"}, "code": {"code_sequence": "y = 3"}},  # Not exact match
    {"example": {"code": "z = 4"}, "code": {"code_sequence": "z = 4"}},  # Exact match
]

with open(test_file, 'w') as f:
    for item in test_data:
        f.write(json.dumps(item) + '\n')

ARGS = {
    "path": test_file
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
none outside the standard library — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type dict — a dict output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a dict output the recorded ndim 0 and dtype family float describe that coercion, not the output, and are not stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
get_rev_vocab Ran
ran on a synthesized input
sunblaze-ucb/SED/program_synthesis/datasets/data.py
code served (permissive licence) · get_code("74ee36a2c213a620")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned dict. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
import numpy as np

torch.manual_seed(0)

# The function get_rev_vocab takes a vocab dict and reverses it
# It expects vocab to be a dict with keys (strings) and values (indices/integers)
# It iterates through vocab.items() to build the reverse mapping

vocab = {
    "hello": 0,
    "world": 1,
    "test": 2,
    "python": 3
}

ARGS = {
    "vocab": vocab
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type dict — a dict output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a dict output the recorded ndim 0 and dtype family float describe that coercion, not the output, and are not stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
hyp_score Ran
ran on a synthesized input
sunblaze-ucb/SED/program_synthesis/models/beam_search.py
code served (permissive licence) · get_code("5f7285b3d57507d8")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned tuple, ndim 3, dtype family float; no output shape is stated. The image it ran in was not recorded. If your copy returns the same type and ndim on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch

torch.manual_seed(0)

# Create small tensors for the function parameters
batch_size = 2
beam_size = 3

# log_probs: should be shape [batch_size x beam_size]
log_probs = torch.randn(batch_size, beam_size)

# sequence_lengths: should be shape [batch_size x beam_size]
sequence_lengths = torch.randint(1, 20, (batch_size, beam_size)).float()

# penalty_factor: a scalar
penalty_factor = 0.6

ARGS = {
    "log_probs": log_probs,
    "sequence_lengths": sequence_lengths,
    "penalty_factor": penalty_factor,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type tuple, ndim 3, dtype family float — a tuple output has no array shape of its own; none is stated
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
karel_output_ref_mutator_mode Ran
ran on a synthesized input
sunblaze-ucb/SED/program_synthesis/datasets/dataset.py
code served (permissive licence) · get_code("d5825e840b5f66b7")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned str. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
import torch.nn as nn

class Args:
    def __init__(self, model_type=None, iterative_search_use_overfit_model=None):
        self.model_type = model_type
        self.iterative_search_use_overfit_model = iterative_search_use_overfit_model

ARGS = {
    'args': Args(model_type='karel-lgrl-overfit', iterative_search_use_overfit_model=None)
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type str — a str output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a str output the recorded ndim 0 and dtype family float describe that coercion, not the output, and are not stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
length_penalty Ran
ran on a synthesized input
sunblaze-ucb/SED/program_synthesis/models/beam_search.py
code served (permissive licence) · get_code("918b15bc296f5115")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned Tensor, ndim 2, dtype family float, shape (2, 4) as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch

torch.manual_seed(0)

# sequence_lengths should be a tensor of shape [batch_size x beam_size x vocab_size]
# according to the docstring, but the function treats it as if it's just a tensor
# of sequence lengths. Looking at the function body, it just does arithmetic on it,
# so we'll create a tensor of shape [batch_size x beam_size]
sequence_lengths = torch.tensor([
    [5.0, 8.0, 3.0, 10.0],
    [7.0, 4.0, 9.0, 6.0]
], dtype=torch.float32)

# penalty_factor is a scalar that weights the length penalty
penalty_factor = 0.6

ARGS = {
    "sequence_lengths": sequence_lengths,
    "penalty_factor": penalty_factor,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Tensor, ndim 2, dtype family float, shape (2, 4) — shape (2, 4) as a later run recorded it
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
load_vocab Ran
ran on a synthesized input
sunblaze-ucb/SED/program_synthesis/datasets/data.py
code served (permissive licence) · get_code("3d825f33d60f08a3")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned dict. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import tempfile
import os

# Create a temporary vocab file with tab-separated key-value pairs
temp_dir = tempfile.mkdtemp()
vocab_file = os.path.join(temp_dir, "vocab.txt")

with open(vocab_file, 'w') as f:
    f.write("apple\t0\n")
    f.write("banana\t1\n")
    f.write("cherry\t2\n")

ARGS = {
    "filename": vocab_file,
    "mapping": True
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
none outside the standard library — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type dict — a dict output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a dict output the recorded ndim 0 and dtype family float describe that coercion, not the output, and are not stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
relpath Ran
ran on a synthesized input
sunblaze-ucb/SED/program_synthesis/datasets/dataset.py
code served (permissive licence) · get_code("5957df89a5e9f359")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned str. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import os

def relpath(path):
    return os.path.join(os.path.dirname(__file__), path)

# The function takes a single parameter 'path' which is a string
# It uses os.path.dirname(__file__) to get the directory of the current file
# and os.path.join to combine it with the provided path

ARGS = {
    "path": "subdir/file.txt"
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
none outside the standard library — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type str — a str output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a str output the recorded ndim 0 and dtype family float describe that coercion, not the output, and are not stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
replace_pad_with_end Ran
ran on a synthesized input
sunblaze-ucb/SED/program_synthesis/datasets/data.py
code served (permissive licence) · get_code("22048f44250fc46a")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned Tensor, ndim 2, dtype family int, shape (2, 4) as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch

torch.manual_seed(0)

# The function replace_pad_with_end takes a single parameter 'var'
# It expects var to be a tensor that supports:
# - .clone() method
# - indexing with == comparison (var == PAD_TOKEN)
# - assignment via indexing (var[...] = value)

# Create a small tensor with some PAD_TOKEN values
var = torch.tensor([[-1, 0, 2, 3],
                    [1, -1, 5, -1]], dtype=torch.long)

ARGS = {
    'var': var
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Tensor, ndim 2, dtype family int, shape (2, 4) — shape (2, 4) as a later run recorded it
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
separate_strategies Ran
ran on a synthesized input
sunblaze-ucb/SED/program_synthesis/analysis/organize_results.py
code served (permissive licence) · get_code("5c5394d7e87f8bab")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned DataFrame. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import pandas as pd
import numpy as np

# Create a sample DataFrame that matches the expected structure
# The function filters by DataLabel and Model, so we need those columns
# It also uses Accuracy and Exact columns
data = {
    'DataLabel': [
        'nearai,,greedy,,1',
        'nearai,,greedy,,5',
        'nearai,,best_first,,10',
        'nearai32,,greedy,,1',
        'nearai32,,best_first,,5',
        'egnps64,,greedy,,1',
        'egnps64,,best_first,,25',
        'egnpsgood,,greedy,,1',
        'egnpsgood,,best_first,,50',
    ],
    'Model': ['model_a', 'model_a', 'model_a', 'model_b', 'model_b', 'model_a', 'model_a', 'model_b', 'model_b'],
    'Accuracy': [0.85, 0.87, 0.89, 0.82, 0.84, 0.90, 0.91, 0.88, 0.92],
    'Exact': [0.80, 0.82, 0.85, 0.78, 0.80, 0.87, 0.88, 0.85, 0.90],
    'Step': [100, 150, 200, 120, 180, 110, 160, 130, 170],
}

df_orig = pd.DataFrame(data)

# Add some additional columns that might be used in groupby operations
# The function does: groupby(list(set(accuracies) - {'Step', 'Accuracy', 'Exact'}))
# So we need at least one more column that could be grouped
df_orig['ExtraColumn'] = ['val1'] * 9

ARGS = {
    'df_orig': df_orig
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, pandas — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type DataFrame — a DataFrame output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a DataFrame output the recorded ndim 2 and dtype family float describe that coercion, not the output, and are not stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
test_cardinality Ran
ran on a synthesized input
sunblaze-ucb/SED/program_synthesis/datasets/stats.py
code served (permissive licence) · get_code("3c24d293db334783")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned int, ndim 0, dtype family int, shape () as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch

torch.manual_seed(0)

# Construct test cases with various outputs
tests = [
    {'output': 'result_1'},
    {'output': 'result_2'},
    {'output': 'result_1'},  # duplicate output
    {'output': 'result_3'},
]

ARGS = {
    'tests': tests
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type int, ndim 0, dtype family int, shape () — shape () as a later run recorded it
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
code_stat Not yet run
unverified (harvested, no recorded run)
sunblaze-ucb/SED/program_synthesis/datasets/stats.py
code served (permissive licence) · get_code("039979ec200d5319")
How it ran

no invocation recorded.

lisp_code_stat Not yet run
unverified (harvested, no recorded run)
sunblaze-ucb/SED/program_synthesis/datasets/stats.py
code served (permissive licence) · get_code("b85637c0fc6f9373")
How it ran

no invocation recorded.

table_of_accuracies Not yet run
unverified (harvested, no recorded run)
sunblaze-ucb/SED/program_synthesis/analysis/load_results.py
code served (permissive licence) · get_code("2e635b1a8221f10c")
How it ran

no invocation recorded.

Also present in other papers' code

Of 19 samples lifted from this paper's linked repositories, 19 carry a function identity and 0 do not. Of the 19 distinct functions with an identity, 1 are also present in other papers whose code we hold (1 with at least one paper dated earlier, 0 the same month, 0 later) and 18 are only here among the 52,977 papers whose code Syntology holds (as of 2026-09-26).

Only here means: among the 52,977 papers whose code Syntology holds (51,887 with lifted samples, 1,090 through their assemblies only), as of 2026-09-26, no other paper's lifted samples or current assembly bricks contain an identical function body after normalisation.

FunctionAlso present inEarliest-datedOthers named
replace_pad_with_end 1 other paper
1 dated earlier
1807.03168 (2018-07)
NAPS: Natural Program Synthesis Dataset

Presence, not direction: an identical function body in two papers' repositories does not say who wrote it, which came from which, or that either paper's method rests on it. Matching is exact after normalisation, so a renamed or edited function is not matched. Dates are when papers appeared, not when code was written. Only .py files the harvester lifted are covered; notebooks are not read. A new-style arXiv id YYMM.NNNN dates a paper to the month 20YY-MM; otherwise Paper.year dates it to the year; otherwise it is undated. Dates are when the paper appeared, not when the code was written. Read 2026-11-21.

Repositories linked to this paper

Some links come from the archived Papers with Code dataset (CC BY-SA 4.0): attribution and licence.

Abstract

The use of deep learning techniques has achieved significant progress for program synthesis from input-output examples. However, when the program semantics become more complex, it still remains a challenge to synthesize programs that are consistent with the specification. In this work, we propose SED, a neural program generation framework that incorporates synthesis, execution, and debugging stages. Instead of purely relying on the neural program synthesizer to generate the final program, SED first produces initial programs using the neural program synthesizer component, then utilizes a neural program debugger to iteratively repair the generated programs. The integration of the debugger component enables SED to modify the programs based on the execution results and specification, which resembles the coding process of human programmers. On Karel, a challenging input-output program synthesis benchmark, SED reduces the error rate of the neural program synthesizer itself by a considerable margin, and outperforms the standard beam search for decoding.

For agents

The same record, over MCP at https://syntology.ai/mcp:

get_harvested_code_for_paper("2007.08095")
get_code_for_paper("2007.08095")
have("2007.08095")

The run record, dated, one paper per request, free:

curl https://syntology.ai/api/ran/2007.08095.json

A badge for a README (the split and the date, never a ratio):

[![Syntology run record](https://syntology.ai/api/ran/2007.08095.svg)](https://syntology.ai/paper/2007.08095)

Connect an agent — have() is free.