SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2002.08155 · EMNLP Findings · 2020

CodeBERT: A Pre-Trained Model for Programming and Natural Languages

Ming Zhou, Ting Liu, Nan Duan, Daya Guo, Daxin Jiang, Ming Gong, Bing Qin, Duyu Tang, Zhangyin Feng, Xiaocheng Feng, Linjun Shou

arXiv · PDF · Open in the Atlas

Code that ran

14 of 20 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 14 have no contract verdict; 0 failures of our instrument; 6 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/2002.08155.json · badge: /api/ran/2002.08155.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
peri-source/peri canonical (classifier label)
classifier basis: census
12 ran · 3 no recorded run
sakirinn/llm4cvd community
Papers with Code archive
1 ran · 0 no recorded run
zhangzwwww/dietcode community
Papers with Code archive
1 ran · 0 no recorded run
zfj1998/CodeBert-Code2Text community
Papers with Code archive
0 ran · 2 no recorded run
microsoft/CodeBERT marked official in Papers with Code
Papers with Code archive
0 ran · 1 no recorded run
FunctionStatusWhere it lives
Model Ran
ran on a synthesized input
sakirinn/llm4cvd/CodeBERT/model.py
pointer only (licence: NONE) · get_code("008726528c96e5c5")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, sakirinn/llm4cvd when it was checked. The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. A run is recorded.

RobertaClassificationHead Ran
ran on a synthesized input
zhangzwwww/dietcode/dietCodeBERT/codesearch/model.py
pointer only (licence: NONE) · get_code("0503ff378bea5dd8")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, zhangzwwww/dietcode when it was checked. The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. A run is recorded.

build_table Ran
ran on a synthesized input
peri-source/peri/peri/special.py
code served (permissive licence) · get_code("ecd29a67eb6f5369")
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 list. 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 numpy as np

def dummy_func(x):
    """A simple function that takes an array and returns a transformed array."""
    return np.sin(x)

ARGS = {
    "func": dummy_func,
    "N": 10
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type list — a list output has no array shape of its own; none is 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
delistify Ran
ran on a synthesized input
peri-source/peri/peri/util.py
code served (permissive licence) · get_code("5a1aedcf3f0c7290")
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 numpy as np

# Test case 1: single element list should be extracted
a1 = ['string']
b1 = None

# Test case 2: multi-element list should remain as list
a2 = ['string', 'other']
b2 = None

# Test case 3: single element numpy array should be extracted
a3 = np.array([1.0])
b3 = None

# Test case 4: multi-element list should remain as list
a4 = [1, 2, 3]
b4 = None

# Test case 5: plain string should remain as string
a5 = 'string'
b5 = None

# Test case 6: when b is a list, convert a to list type
a6 = 'element'
b6 = ['target_type']

# Test case 7: when b is a numpy array, convert a to numpy array type
a7 = 42
b7 = np.array([1, 2, 3])

# Test case 8: when b is a tuple, convert a to tuple type
a8 = 'value'
b8 = ('target', 'type')

ARGS = {
    'a': ['string'],
    'b': None
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy — 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
listify Ran
ran on a synthesized input
peri-source/peri/peri/logger.py
code served (permissive licence) · get_code("f35d31e49f765f57")
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 list, ndim 1, dtype family int; 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
import numpy as np

torch.manual_seed(0)

# The function listify takes a single parameter 'a' and checks if it's a list or tuple.
# If not, it wraps it in a list. If it is, it returns it as-is.
# We need to test with a value that is NOT a list or tuple to make the function meaningful.

ARGS = {
    "a": 42
}
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 list, ndim 1, dtype family int — a list 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
listify Ran
ran on a synthesized input
peri-source/peri/peri/util.py
code served (permissive licence) · get_code("37d657d7803f8bb4")
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 list, ndim 1, dtype family int; 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 numpy as np

# Test case 1: scalar input
a_scalar = 42

# Test case 2: list input
a_list = [1, 2, 3]

# Test case 3: string input (treated as scalar, not iterable)
a_string = 'a'

# Test case 4: numpy array input
a_numpy = np.array([1, 2, 3])

# Test case 5: tuple input
a_tuple = (4, 5, 6)

# Test case 6: None input
a_none = None

# We'll use the scalar case as the primary test
ARGS = {
    'a': 42
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type list, ndim 1, dtype family int — a list 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
load_tiff Ran
ran on a synthesized input
peri-source/peri/peri/initializers.py
code served (permissive licence) · get_code("121597dd382c5a02")
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 ndarray, ndim 3, dtype family int, shape (2, 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 tempfile
import numpy as np
from PIL import Image

# Create a temporary TIFF file with multiple frames
with tempfile.NamedTemporaryFile(suffix='.tiff', delete=False) as tmp:
    tmp_filename = tmp.name

# Create a multi-frame TIFF image
frames = [
    Image.fromarray(np.random.randint(0, 256, (4, 3), dtype=np.uint8)),
    Image.fromarray(np.random.randint(0, 256, (4, 3), dtype=np.uint8)),
]
frames[0].save(tmp_filename, save_all=True, append_images=frames[1:])

ARGS = {
    'filename': tmp_filename
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
PIL, numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type ndarray, ndim 3, dtype family int, shape (2, 4, 3) — shape (2, 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
load_tiff_layer Ran
ran on a synthesized input
peri-source/peri/peri/initializers.py
code served (permissive licence) · get_code("9adb618f6c87b4a4")
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 ndarray, ndim 3, dtype family int, shape (1, 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 tempfile
import numpy as np
from PIL import Image
import os

# Create a temporary TIFF file with multiple layers
temp_dir = tempfile.mkdtemp()
tiff_path = os.path.join(temp_dir, "test.tiff")

# Create a multi-layer TIFF image
# Layer 0: 4x3 array
layer0 = np.random.randint(0, 256, (4, 3), dtype=np.uint8)
img0 = Image.fromarray(layer0)

# Layer 1: 4x3 array
layer1 = np.random.randint(0, 256, (4, 3), dtype=np.uint8)

# Layer 2: 4x3 array
layer2 = np.random.randint(0, 256, (4, 3), dtype=np.uint8)

# Save as multi-layer TIFF
img0.save(tiff_path, save_all=True, append_images=[Image.fromarray(layer1), Image.fromarray(layer2)])

ARGS = {
    "filename": tiff_path,
    "layer": 1
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
PIL, numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type ndarray, ndim 3, dtype family int, shape (1, 4, 3) — shape (1, 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
normalize Ran
ran on a synthesized input
peri-source/peri/peri/initializers.py
code served (permissive licence) · get_code("ba4bfc2758a0045e")
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 ndarray, 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 numpy as np

# Create a small test image array
im = np.array([[10.0, 50.0, 100.0, 200.0],
               [25.0, 75.0, 150.0, 250.0]], dtype=np.float32)

ARGS = {
    'im': im,
    'invert': False,
    'scale': (0.0, 255.0),
    'dtype': np.float64
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type ndarray, 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
oddify Ran
ran on a synthesized input
peri-source/peri/peri/util.py
code served (permissive licence) · get_code("92d0ace70a4248ed")
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)

ARGS = {
    "num": 4
}
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
reduce_add Ran
ran on a synthesized input
peri-source/peri/peri/comp/comp.py
code served (permissive licence) · get_code("dde61ac910691485")
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 1, dtype family int, shape (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
from operator import add
from functools import reduce

torch.manual_seed(0)

# reduce_add takes a single parameter 'x' which is an iterable
# The function applies reduce with add operator across all elements
# So x should be a list/tuple of numbers or tensors that can be added together

# Create a list of small tensors that can be added together
x = [
    torch.tensor([1, 2, 3, 4]),
    torch.tensor([5, 6, 7, 8]),
    torch.tensor([9, 10, 11, 12])
]

ARGS = {
    "x": x
}
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 1, dtype family int, shape (4,) — shape (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
sample Ran
ran on a synthesized input
peri-source/peri/peri/states.py
code served (permissive licence) · get_code("8ce382543ba7073c")
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 ndarray, ndim 1, dtype family float, shape (5,) 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 numpy as np

torch.manual_seed(0)

# Create a small 3D field (e.g., an image or volume)
field = np.arange(60).reshape(3, 4, 5).astype(np.float32)

# Create flat indices to sample from the raveled field
inds = np.array([0, 5, 10, 15, 20])

# Create a slicer object (3D slice)
slicer = np.s_[1:3, 1:3, 2:4]

ARGS = {
    "field": field,
    "inds": inds,
    "slicer": slicer,
    "flat": True
}
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 ndarray, ndim 1, dtype family float, shape (5,) — shape (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
sanitize Ran
ran on a synthesized input
peri-source/peri/peri/logger.py
code served (permissive licence) · get_code("2e5dd6eaf5b83724")
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

torch.manual_seed(0)

ARGS = {
    'v': [1, 2, 3]
}
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
transform Ran
ran on a synthesized input
peri-source/peri/peri/conf.py
code served (permissive licence) · get_code("38157a1ab1b0b136")
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

torch.manual_seed(0)

def transform(v):
    """
    Translate environment variables to ones corresponding to keys in the
    configuration.  In particular, env variables may be made with
    "PERI_"+key_name: fftw-threads = PERI_FFTW_THREADS. Each env var
    is later checked to see if it has to do with PERI
    """
    return v.lower().replace('_', '-').replace('peri-', '')

ARGS = {
    'v': 'PERI_FFTW_THREADS'
}
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
Beam Not yet run
unverified (harvested, no recorded run)
zfj1998/CodeBert-Code2Text/code/model.py
pointer only (licence: NONE) · get_code("fe20afbd1ed12d33")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, zfj1998/CodeBert-Code2Text when it was checked. The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. No invocation recorded.

Seq2Seq Not yet run
unverified (harvested, no recorded run)
zfj1998/CodeBert-Code2Text/code/model.py
pointer only (licence: NONE) · get_code("09b02d6c43ccfaa8")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, zfj1998/CodeBert-Code2Text when it was checked. The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. No invocation recorded.

Seq2Seq Not yet run
unverified (harvested, no recorded run)
microsoft/CodeBERT/CodeBERT/code2nl/model.py
code served (permissive licence) · get_code("8053f41db1d56bde")
How it ran

no invocation recorded.

fast_j0 Not yet run
unverified (harvested, no recorded run)
peri-source/peri/peri/special.py
code served (permissive licence) · get_code("dee538453dd482d2")
How it ran

no invocation recorded.

fast_j1 Not yet run
unverified (harvested, no recorded run)
peri-source/peri/peri/special.py
code served (permissive licence) · get_code("ed15ecd3ff651e9b")
How it ran

no invocation recorded.

reduce_mul Not yet run
unverified (harvested, no recorded run)
peri-source/peri/peri/comp/comp.py
code served (permissive licence) · get_code("16e86d560b446cd7")
How it ran

no invocation recorded.

Also present in other papers' code

Of 20 samples lifted from this paper's linked repositories, 20 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 (0 with at least one paper dated earlier, 0 the same month, 1 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
Beam
brick of this paper's assembly: helper
1 other paper
1 later · 1 only as an assembly brick
2109.00859 (2021-09)
CodeT5: Identifier-aware Unified Pre-trained Encoder-Decoder Models for Code Understanding and Generation

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

We present CodeBERT, a bimodal pre-trained model for programming language (PL) and natural language (NL). CodeBERT learns general-purpose representations that support downstream NL-PL applications such as natural language code search, code documentation generation, etc. We develop Code-BERT with Transformer-based neural architecture, and train it with a hybrid objective function that incorporates the pre-training task of replaced token detection, which is to detect plausible alternatives sampled from generators. This enables us to utilize both "bimodal" data of NL-PL pairs and "unimodal" data, where the former provides input tokens for model training while the latter helps to learn better generators. We evaluate CodeBERT on two NL-PL applications by fine-tuning model parameters. Results show that CodeBERT achieves state-of-the-art performance on both natural language code search and code documentation generation. Furthermore, to investigate what type of knowledge is learned in CodeBERT, we construct a dataset for NL-PL probing, and evaluate in a zero-shot setting where parameters of pre-trained models are fixed. Results show that CodeBERT performs better than previous pre-trained models on NL-PL probing. 1

For agents

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

get_harvested_code_for_paper("2002.08155")
get_code_for_paper("2002.08155")
have("2002.08155")

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

curl https://syntology.ai/api/ran/2002.08155.json

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

[![Syntology run record](https://syntology.ai/api/ran/2002.08155.svg)](https://syntology.ai/paper/2002.08155)

Connect an agent — have() is free.