SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 1906.01698 · 2019

Open Sesame: Getting Inside BERT's Linguistic Knowledge

arXiv · PDF · Open in the Atlas

Code that ran

13 of 18 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; 4 honoured the contract we drafted, 0 violated it and 2 have no contract verdict; 7 failures of our instrument; 5 no recorded run); 1 of them was also run on the host, not in the jail, by the peer-agreement lane that set its level, measured 2026-11-22; 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/1906.01698.json · badge: /api/ran/1906.01698.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
yongjie-lin/bert-opensesame canonical (classifier label)
classifier basis: census
13 ran · 5 no recorded run
FunctionStatusWhere it lives
swish Ran
ran; our contract draft was wrong, not the code
yongjie-lin/bert-opensesame/pytorch_pretrained_bert/modeling.py
code served (permissive licence) · get_code("0f786c407fb1ee4c")
How it ran

Ran in Syntology's jail on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1) 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. The comparison behind this sample's level is a different run: the peer-agreement lane ran it again on the host, not in the jail (report dated 2026-09-04T22:33:52Z). 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 swish takes a single parameter 'x' which is used with:
# - torch.sigmoid(x) - requires a tensor
# - x * ... - requires a tensor that supports multiplication
# So x should be a tensor

x = torch.randn(2, 4)

ARGS = {
    "x": x
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1
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
cached_path Ran
ran; our contract draft was wrong, not the code
yongjie-lin/bert-opensesame/pytorch_pretrained_bert/file_utils.py
pointer only (licence: Apache-2.0); the stored copy get_code serves, from sejong-rcv/TAME, is pointer only (licence: NONE) · get_code("349d780dd4a89a37")
How it ran

The body of this sample is withheld: this paper's link allows the body inline (Apache-2.0), but get_code serves the stored copy, first harvested from sejong-rcv/TAME, and GitHub detected no licence in that repository 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.

filename_to_url Ran
ran; our contract draft was wrong, not the code
yongjie-lin/bert-opensesame/pytorch_pretrained_bert/file_utils.py
pointer only (licence: Apache-2.0); the stored copy get_code serves, from sejong-rcv/TAME, is pointer only (licence: NONE) · get_code("db0ac56aaf6e35e6")
How it ran

The body of this sample is withheld: this paper's link allows the body inline (Apache-2.0), but get_code serves the stored copy, first harvested from sejong-rcv/TAME, and GitHub detected no licence in that repository 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.

gelu Ran
ran; our contract draft was wrong, not the code
yongjie-lin/bert-opensesame/bertviz/bertviz/pytorch_pretrained_bert/modeling_gpt2.py
code served (permissive licence) · get_code("8d23fbe2b99b840b")
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) 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
import math

torch.manual_seed(0)

# Create a small tensor for testing
x = torch.randn(2, 4)

ARGS = {
    "x": x
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1
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
gelu Ran
ran, honoured the contract we drafted
yongjie-lin/bert-opensesame/bertviz/bertviz/pytorch_pretrained_bert/modeling.py
pointer only (licence: Apache-2.0); the stored copy get_code serves, from tanmay1618/text_to_sql_bert, is pointer only (licence: NONE) · get_code("fdc64f4c72036ae4")
How it ran

The body of this sample is withheld: this paper's link allows the body inline (Apache-2.0), but get_code serves the stored copy, first harvested from tanmay1618/text_to_sql_bert, and GitHub detected no licence in that repository 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.

gelu Ran
ran, honoured the contract we drafted
yongjie-lin/bert-opensesame/pytorch_pretrained_bert/modeling.py
pointer only (licence: Apache-2.0); the stored copy get_code serves, from shmsw25/qa-hard-em, is pointer only (licence: NONE) · get_code("40e9fee2e0b7e278")
How it ran

The body of this sample is withheld: this paper's link allows the body inline (Apache-2.0), but get_code serves the stored copy, first harvested from shmsw25/qa-hard-em, and GitHub detected no licence in that repository 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.

load_vocab Ran
ran; our contract draft was wrong, not the code
yongjie-lin/bert-opensesame/pytorch_pretrained_bert/tokenization.py
code served (permissive licence) · get_code("fb95c4b13cdf89a2")
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 OrderedDict. 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 collections
import tempfile
import os

# Create a temporary vocabulary file
temp_dir = tempfile.mkdtemp()
vocab_file = os.path.join(temp_dir, "vocab.txt")

# Write some sample tokens to the vocabulary file
with open(vocab_file, 'w', encoding='utf-8') as f:
    f.write("[PAD]\n")
    f.write("[UNK]\n")
    f.write("[CLS]\n")
    f.write("[SEP]\n")
    f.write("the\n")
    f.write("quick\n")
    f.write("brown\n")
    f.write("fox\n")

ARGS = {
    "vocab_file": vocab_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 OrderedDict — a OrderedDict output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a OrderedDict 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
normalize_string Ran
ran on a synthesized input
yongjie-lin/bert-opensesame/pytorch_pretrained_bert/dd_utils.py
code served (permissive licence) · get_code("5c1d1e5a7edda4cd")
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 unicodedata

def normalize_string(string):
    """
    Normalizes the string from unicode to ascii,
    lowercasing and stripping whitespace
    """

    def unicode_to_ascii(s):
        return ''.join((c for c in unicodedata.normalize('NFD', s) if unicodedata.category(c) != 'Mn'))
    return unicode_to_ascii(string.lower().strip())


# Setup code to construct valid arguments
ARGS = {
    'string': '  Héllo Wörld  '
}
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
prep_batch Ran
ran on a synthesized input
yongjie-lin/bert-opensesame/pytorch_pretrained_bert/dd_utils.py
code served (permissive licence) · get_code("83145131533aa4db")
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. 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
from torch import nn

torch.manual_seed(0)

# x_batch: list of sequences (each sequence is a list of integers)
# Each sequence has different length
x_batch = [
    [1, 2, 3],
    [4, 5]
]

# y_batch: list of target indices (one per sample in batch)
y_batch = [0, 1]

# device: CPU device
device = torch.device('cpu')

ARGS = {
    'x_batch': x_batch,
    'y_batch': y_batch,
    'device': device
}
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 — a tuple 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
url_to_filename Ran
ran; our contract draft was wrong, not the code
yongjie-lin/bert-opensesame/pytorch_pretrained_bert/file_utils.py
pointer only (licence: Apache-2.0); the stored copy get_code serves, from lonePatient/ERNIE-text-classification-pytorch, is pointer only (licence: NONE) · get_code("af64ec220e8bcdbc")
How it ran

The body of this sample is withheld: this paper's link allows the body inline (Apache-2.0), but get_code serves the stored copy, first harvested from lonePatient/ERNIE-text-classification-pytorch, and GitHub detected no licence in that repository 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.

warmup_constant Ran
ran, honoured the contract we drafted
yongjie-lin/bert-opensesame/pytorch_pretrained_bert/optimization.py
pointer only (licence: Apache-2.0); the stored copy get_code serves, from Yueeeeeeee/Self-Supervised-QA, is pointer only (licence: NONE) · get_code("e7d542062316094a")
How it ran

The body of this sample is withheld: this paper's link allows the body inline (Apache-2.0), but get_code serves the stored copy, first harvested from Yueeeeeeee/Self-Supervised-QA, and GitHub detected no licence in that repository 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.

warmup_linear Ran
ran, honoured the contract we drafted
yongjie-lin/bert-opensesame/pytorch_pretrained_bert/optimization.py
code served (permissive licence) · get_code("c58d57224530d17e")
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) and returned float, ndim 0, dtype family float, 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)

# The function warmup_linear takes two parameters:
# - x: a numeric value (float) representing a position/step in training
# - warmup: a float representing the warmup threshold (default 0.002)
# 
# The function uses x and warmup in simple arithmetic comparisons and operations,
# so we just need numeric values.

ARGS = {
    "x": 0.001,
    "warmup": 0.002
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type float, ndim 0, dtype family float, 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
whitespace_tokenize Ran
ran; our contract draft was wrong, not the code
yongjie-lin/bert-opensesame/pytorch_pretrained_bert/tokenization.py
code served (permissive licence) · get_code("da7295883cd7da14")
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 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
import numpy as np

def whitespace_tokenize(text):
    """Runs basic whitespace cleaning and splitting on a peice of text."""
    text = text.strip()
    if not text:
        return []
    tokens = text.split()
    return tokens

# Setup code
torch.manual_seed(0)

# The function takes a single parameter: text (a string)
# It strips whitespace and splits on whitespace
# We need to provide a valid text string

ARGS = {
    "text": "  hello world this is a test  "
}
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 float — 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
augment_data_JJ Not yet run
unverified (harvested, no recorded run)
yongjie-lin/bert-opensesame/pytorch_pretrained_bert/dd_utils.py
code served (permissive licence) · get_code("776ea2acae7131ec")
How it ran

no invocation recorded.

load_tf_weights_in_bert Not yet run
unverified (harvested, no recorded run)
yongjie-lin/bert-opensesame/bertviz/bertviz/pytorch_pretrained_bert/modeling.py
code served (permissive licence) · get_code("baa5766f4566aafc")
How it ran

no invocation recorded.

load_tf_weights_in_gpt2 Not yet run
unverified (harvested, no recorded run)
yongjie-lin/bert-opensesame/bertviz/bertviz/pytorch_pretrained_bert/modeling_gpt2.py
code served (permissive licence) · get_code("a62604f618a8cc82")
How it ran

no invocation recorded.

load_tf_weights_in_openai_gpt Not yet run
unverified (harvested, no recorded run)
yongjie-lin/bert-opensesame/bertviz/bertviz/pytorch_pretrained_bert/modeling_openai.py
code served (permissive licence) · get_code("46f305675277e757")
How it ran

no invocation recorded.

warmup_cosine Not yet run
unverified (harvested, no recorded run)
yongjie-lin/bert-opensesame/pytorch_pretrained_bert/optimization.py
pointer only (licence: Apache-2.0); the stored copy get_code serves, from Yueeeeeeee/Self-Supervised-QA, is pointer only (licence: NONE) · get_code("35f7cddf90dd05d4")
How it ran

The body of this sample is withheld: this paper's link allows the body inline (Apache-2.0), but get_code serves the stored copy, first harvested from Yueeeeeeee/Self-Supervised-QA, and GitHub detected no licence in that repository 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.

Also present in other papers' code

Of 18 samples lifted from this paper's linked repositories, 18 carry a function identity and 0 do not. Of the 17 distinct functions with an identity, 14 are also present in other papers whose code we hold (12 with at least one paper dated earlier, 6 the same month, 14 later) and 3 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
swish 206 other papers
9 dated earlier, 2 the same month, 195 later · 44 only as an assembly brick
1408.5882 (2014-08)
Convolutional Neural Networks for Sentence Classification
1805.12114 (2018-05)
Deep Reinforcement Learning in a Handful of Trials using Probabilistic Dynamics Models
1809.01357 (2018-09)
Zero Shot Learning for Code Education: Rubric Sampling with Deep Learning Inference
and 203 more not named here
gelu 198 other papers
8 dated earlier, 4 the same month, 186 later · 32 only as an assembly brick
1408.5882 (2014-08)
Convolutional Neural Networks for Sentence Classification
1810.04805 (2018-10)
BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding
1901.10125 (2019-01)
Glyce: Glyph-vectors for Chinese Character Representations
and 195 more not named here
gelu 51 other papers
3 dated earlier, 3 the same month, 45 later · 8 only as an assembly brick
1706.03762 (2017-06)
Attention Is All You Need
1810.04805 (2018-10)
BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding
1901.02860 (2019-01)
Transformer-XL: Attentive Language Models Beyond a Fixed-Length Context
and 48 more not named here
whitespace_tokenize 43 other papers
2 dated earlier, 41 later
1408.5882 (2014-08)
Convolutional Neural Networks for Sentence Classification
1810.04805 (2018-10)
BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding
1909.11942 (2019-09)
ALBERT: A Lite BERT for Self-supervised Learning of Language Representations
and 40 more not named here
warmup_constant 28 other papers
2 dated earlier, 1 the same month, 25 later
1408.5882 (2014-08)
Convolutional Neural Networks for Sentence Classification
1903.09588 (2019-03)
Utilizing BERT for Aspect-Based Sentiment Analysis via Constructing Auxiliary Sentence
1906.05807 (2019-06)
Real-Time Open-Domain Question Answering with Dense-Sparse Phrase Index
and 25 more not named here
url_to_filename 20 other papers
2 dated earlier, 18 later · 3 only as an assembly brick
1408.5882 (2014-08)
Convolutional Neural Networks for Sentence Classification
1810.04805 (2018-10)
BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding
1909.11059 (2019-09)
Unified Vision-Language Pre-Training for Image Captioning and VQA
and 17 more not named here
warmup_linear 19 other papers
6 dated earlier, 1 the same month, 12 later
1408.5882 (2014-08)
Convolutional Neural Networks for Sentence Classification
1803.03376 (2018-03)
Learning Approximate Inference Networks for Structured Prediction
1810.04805 (2018-10)
BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding
and 16 more not named here
cached_path 17 other papers
2 dated earlier, 15 later · 3 only as an assembly brick
1408.5882 (2014-08)
Convolutional Neural Networks for Sentence Classification
1810.04805 (2018-10)
BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding
1909.11059 (2019-09)
Unified Vision-Language Pre-Training for Image Captioning and VQA
and 14 more not named here
warmup_cosine 17 other papers
2 dated earlier, 1 the same month, 14 later
1408.5882 (2014-08)
Convolutional Neural Networks for Sentence Classification
1903.09588 (2019-03)
Utilizing BERT for Aspect-Based Sentiment Analysis via Constructing Auxiliary Sentence
1906.05807 (2019-06)
Real-Time Open-Domain Question Answering with Dense-Sparse Phrase Index
and 14 more not named here
filename_to_url 13 other papers
1 dated earlier, 12 later
1408.5882 (2014-08)
Convolutional Neural Networks for Sentence Classification
1909.11059 (2019-09)
Unified Vision-Language Pre-Training for Image Captioning and VQA
2109.00590 (2021-09)
WebQA: Multihop and Multimodal QA
and 10 more not named here
load_tf_weights_in_bert 12 other papers
1 dated earlier, 11 later · 1 only as an assembly brick
1810.04805 (2018-10)
BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding
1908.02265 (2019-08)
ViLBERT: Pretraining Task-Agnostic Visiolinguistic Representations for Vision-and-Language Tasks
1909.00252 (2019-09)
Humor Detection: A Transformer Gets the Last Laugh
and 9 more not named here
load_vocab 11 other papers
1 dated earlier, 10 later
1408.5882 (2014-08)
Convolutional Neural Networks for Sentence Classification
2004.07453 (2020-04)
The Right Tool for the Job: Matching Model and Instance Complexities
2010.06138 (2020-10)
Incorporating BERT into Parallel Sequence Decoding with Adapters
and 8 more not named here
load_tf_weights_in_gpt2 4 other papers
4 later
1908.09355 (2019-08)
Patient Knowledge Distillation for BERT Model Compression
1911.00720 (2019-11)
ZEN: Pre-training Chinese Text Encoder Enhanced by N-gram Representations
2004.05707 (2020-04)
VGCN-BERT: Augmenting BERT with Graph Embedding for Text Classification
and 1 more not named here
load_tf_weights_in_openai_gpt 3 other papers
3 later
1908.09355 (2019-08)
Patient Knowledge Distillation for BERT Model Compression
1911.00720 (2019-11)
ZEN: Pre-training Chinese Text Encoder Enhanced by N-gram Representations
2005.02439 (2020-05)
Contextualizing Hate Speech Classifiers with Post-hoc Explanation

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 2026-11-22.

Repositories linked to this paper

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

Abstract

How and to what extent does BERT encode syntactically-sensitive hierarchical information or positionally-sensitive linear information? Recent work has shown that contextual representations like BERT perform well on tasks that require sensitivity to linguistic structure. We present here two studies which aim to provide a better understanding of the nature of BERT's representations. The first of these focuses on the identification of structurally-defined elements using diagnostic classifiers, while the second explores BERT's representation of subject-verb agreement and anaphor-antecedent dependencies through a quantitative assessment of self-attention vectors. In both cases, we find that BERT encodes positional information about word tokens well on its lower layers, but switches to a hierarchically-oriented encoding on higher layers. We conclude then that BERT's representations do indeed model linguistically relevant aspects of hierarchical structure, though they do not appear to show the sharp sensitivity to hierarchical structure that is found in human processing of reflexive anaphora.

For agents

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

get_harvested_code_for_paper("1906.01698")
get_code_for_paper("1906.01698")
have("1906.01698")

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

curl https://syntology.ai/api/ran/1906.01698.json

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

[![Syntology run record](https://syntology.ai/api/ran/1906.01698.svg)](https://syntology.ai/paper/1906.01698)

Connect an agent — have() is free.