SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2006.01938 · 2020

Nurse is Closer to Woman than Surgeon? Mitigating Gender-Biased Proximities in Word Embeddings

arXiv · PDF · Open in the Atlas

Code that ran

12 of 13 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 12 have no contract verdict; 0 failures of our instrument; 1 no recorded run), 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/2006.01938.json · badge: /api/ran/2006.01938.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
TimeTraveller-San/RAN-Debias canonical (classifier label)
classifier basis: census
12 ran · 1 no recorded run
FunctionStatusWhere it lives
dedup Ran
ran on a synthesized input
TimeTraveller-San/RAN-Debias/src/we.py
code served (permissive licence) · get_code("653c902f4192ec34")
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

torch.manual_seed(0)

# The function dedup takes a sequence and removes duplicates while preserving order
# It uses the 'in' operator to check membership and set.add() to track seen items
# The sequence should contain hashable items

ARGS = {
    "seq": [1, 2, 3, 2, 1, 4, 3, 5]
}
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 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
get_nbs Ran
ran on a synthesized input
TimeTraveller-San/RAN-Debias/src/make_neighbours.py
code served (permissive licence) · get_code("3300a5e982458678")
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 int, 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 numpy as np

class MockEmbedding:
    def __init__(self):
        # vecs should be a 2D array where each row is an embedding vector
        # Using shape (10, 4) - 10 words, 4-dimensional embeddings
        self.vecs = np.random.randn(10, 4).astype(np.float32)
        # word_to_idx mapping for the v() method
        self.word_to_idx = {
            'hello': 0,
            'world': 1,
            'test': 2,
            'word': 3,
            'example': 4,
            'python': 5,
            'code': 6,
            'data': 7,
            'vector': 8,
            'embedding': 9
        }
    
    def v(self, word):
        """Return the embedding vector for a given word"""
        idx = self.word_to_idx.get(word, 0)
        return self.vecs[idx]

E = MockEmbedding()
word = 'hello'
k = 5

ARGS = {
    'E': E,
    'word': word,
    'k': k
}
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 1, dtype family int, 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
get_neighbors Ran
ran on a synthesized input
TimeTraveller-San/RAN-Debias/src/gipe.py
code served (permissive licence) · get_code("34c4e09e34e2fc7b")
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

# Create a mock neighborhood structure
# N[word] should be a dict where keys are neighbors
# The function returns list(N[word].keys())[1:k+1]
# This suggests N[word] is an ordered dict-like object with neighbors

# Build a simple neighborhood dictionary
N = {
    'hello': {
        'hello': 1.0,      # index 0 (skipped by [1:k+1])
        'world': 0.95,     # index 1
        'hi': 0.90,        # index 2
        'greetings': 0.85, # index 3
        'hey': 0.80,       # index 4
    },
    'world': {
        'world': 1.0,
        'earth': 0.92,
        'globe': 0.88,
        'planet': 0.84,
    }
}

ARGS = {
    'N': N,
    'word': 'hello',
    'k': 2
}
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
get_ns_idb Ran
ran on a synthesized input
TimeTraveller-San/RAN-Debias/src/RANDebias.py
code served (permissive licence) · get_code("ac609e24a985b508")
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

# The function get_ns_idb takes a word and a neighbor dictionary N
# and returns N[word], so we need to construct a valid neighbor dictionary

# Create a simple neighbor dictionary where keys are words and values are lists of neighbor words
N = {
    "hello": ["hi", "hey", "hallo", "hello_variant"],
    "world": ["earth", "globe", "universe", "planet"],
    "python": ["snake", "programming", "code", "language"],
    "function": ["method", "procedure", "routine", "operation"]
}

# The word parameter should be a key that exists in N
word = "hello"

ARGS = {
    "word": word,
    "N": N
}
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 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
get_ns_idb Ran
ran on a synthesized input
TimeTraveller-San/RAN-Debias/src/make_neighbours.py
code served (permissive licence) · get_code("8622b41ff65f28b2")
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 numpy as np
from numba import jit

class MockEmbeddings:
    def __init__(self):
        np.random.seed(0)
        self.vecs = np.random.randn(10, 4).astype(np.float64)
        self.words = [f"word_{i}" for i in range(10)]
    
    def v(self, word):
        idx = self.words.index(word)
        return self.vecs[idx]

E = MockEmbeddings()
word = "word_0"
g = np.array([1.0, 0.0, 0.0, 0.0], dtype=np.float64)

ARGS = {
    "E": E,
    "word": word,
    "g": g
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numba, numpy — 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_pair_idb Ran
ran on a synthesized input
TimeTraveller-San/RAN-Debias/src/gipe.py
code served (permissive licence) · get_code("666e7b4bb960c7b3")
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 float64, ndim 0, 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 numpy as np

class MockEmbedding:
    """Mock embedding object that stores vectors and can retrieve them by string key."""
    def __init__(self):
        self.vectors = {
            'w_key': np.array([1.0, 2.0, 3.0, 4.0]),
            'v_key': np.array([2.0, 1.0, 4.0, 3.0])
        }
    
    def v(self, key):
        """Retrieve vector by string key."""
        return self.vectors[key]

# Create embedding object with sample vectors
E = MockEmbedding()

# Create sample vectors
w = 'w_key'  # string key to be resolved via E.v()
v = 'v_key'  # string key to be resolved via E.v()

# Create direction vector g (must be normalized for meaningful computation)
g = np.array([1.0, 0.0, 0.0, 0.0])
g = g / np.linalg.norm(g)  # normalize

ARGS = {
    'w': w,
    'v': v,
    'g': g,
    'E': E
}
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 float64, ndim 0, dtype family float — no output shape is stated: observe_signature.py recorded float ndim 0, which disagrees with the fixture run
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_pair_idb Ran
ran on a synthesized input
TimeTraveller-San/RAN-Debias/src/make_neighbours.py
code served (permissive licence) · get_code("e65013b545c91757")
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 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 numpy as np
from numba import jit

@jit(nopython=False)
def get_pair_idb(w, v, g):
    w_orth = w - np.dot(w, g) * g
    v_orth = v - np.dot(v, g) * g
    dots = np.dot(w, v)
    orth_dots = np.dot(w_orth, v_orth)
    I = (dots - orth_dots / (np.linalg.norm(w_orth) * np.linalg.norm(v_orth))) / dots
    return I

# Create small numpy arrays as arguments
# w, v, g should be 1D vectors (based on usage with np.dot and np.linalg.norm)
np.random.seed(0)

w = np.array([1.0, 2.0, 3.0, 4.0])
v = np.array([2.0, 3.0, 4.0, 5.0])
g = np.array([0.1, 0.2, 0.3, 0.4])

# Normalize g to be a unit vector for stability
g = g / np.linalg.norm(g)

ARGS = {
    'w': w,
    'v': v,
    'g': g
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numba, numpy — 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
init_vector Ran
ran on a synthesized input
TimeTraveller-San/RAN-Debias/src/RANDebias.py
code served (permissive licence) · get_code("3f567dcf4a70ee50")
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 float, 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 copy import deepcopy

torch.manual_seed(0)

class StandInEmbedding:
    """Stand-in class that provides a v() method for word embeddings"""
    def __init__(self):
        self.embeddings = {
            'hello': [0.1, 0.2, 0.3, 0.4],
            'world': [0.5, 0.6, 0.7, 0.8],
        }
    
    def v(self, word):
        """Return embedding vector for a word"""
        if word in self.embeddings:
            return self.embeddings[word]
        return [0.0, 0.0, 0.0, 0.0]

word = 'hello'
E = StandInEmbedding()

ARGS = {
    'word': word,
    'E': E,
}
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 float, 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
parse_float Ran
ran on a synthesized input
TimeTraveller-San/RAN-Debias/src/RANDebias.py
code served (permissive licence) · get_code("7c933676f9aafa31")
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 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
import numpy as np

def parse_float(s):
    """to read the config file properly"""
    try:
        return float(s)
    except:
        n, d = s.split('/')
        return float(n) / float(d)

# The function takes a single string parameter 's'
# It attempts to parse it as a float, or if that fails,
# it splits by '/' and divides two floats

ARGS = {
    's': '3.14'
}
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 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
prox_bias Ran
ran on a synthesized input
TimeTraveller-San/RAN-Debias/src/gipe.py
code served (permissive licence) · get_code("85ce3c4d3a2d9bf3")
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 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
import numpy as np

torch.manual_seed(0)

# Construct arguments for prox_bias(vals, l, thresh)
# vals: needs to support comparison (vals > thresh) and len() on result
# l: integer for division
# thresh: scalar threshold value

vals = torch.tensor([0.1, 0.5, 1.2, 2.3, 0.8, 1.5], dtype=torch.float32)
l = 6
thresh = 1.0

ARGS = {
    "vals": vals,
    "l": l,
    "thresh": thresh
}
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 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
read Ran
ran on a synthesized input
TimeTraveller-San/RAN-Debias/src/combine.py
code served (permissive licence) · get_code("16667359f4cb8fda")
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 pickle
import tempfile
import os

# Create a temporary file with pickled data
temp_dir = tempfile.mkdtemp()
temp_file = os.path.join(temp_dir, "test_data.pkl")

# Create some test data and pickle it
test_data = {
    "key1": "value1",
    "key2": [1, 2, 3],
    "key3": {"nested": "dict"}
}

with open(temp_file, 'wb') as f:
    pickle.dump(test_data, f)

# Define ARGS with the fname parameter
ARGS = {
    "fname": temp_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
safe_word Ran
ran on a synthesized input
TimeTraveller-San/RAN-Debias/src/we.py
code served (permissive licence) · get_code("dd4de119837da17b")
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 bool, 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 re

def safe_word(w):
    return re.match('^[a-z_]*$', w) and len(w) < 20 and (not re.match('^_*$', w))

ARGS = {
    'w': 'hello_world'
}
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 bool, 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
to_utf8 Not yet run
unverified (harvested, no recorded run)
TimeTraveller-San/RAN-Debias/src/we.py
code served (permissive licence) · get_code("1ae322ed84456151")
How it ran

no invocation recorded.

Also present in other papers' code

Of 13 samples lifted from this paper's linked repositories, 13 carry a function identity and 0 do not. Of the 13 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 12 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
to_utf8 1 other paper
1 dated earlier
1607.06520 (2016-07)
Man is to Computer Programmer as Woman is to Homemaker? Debiasing Word Embeddings

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

Word embeddings are the standard model for semantic and syntactic representations of words. Unfortunately, these models have been shown to exhibit undesirable word associations resulting from gender, racial, and religious biases. Existing post-processing methods for debiasing word embeddings are unable to mitigate gender bias hidden in the spatial arrangement of word vectors. In this paper, we propose RAN-Debias, a novel gender debiasing methodology which not only eliminates the bias present in a word vector but also alters the spatial distribution of its neighbouring vectors, achieving a bias-free setting while maintaining minimal semantic offset. We also propose a new bias evaluation metric - Gender-based Illicit Proximity Estimate (GIPE), which measures the extent of undue proximity in word vectors resulting from the presence of gender-based predilections. Experiments based on a suite of evaluation metrics show that RAN-Debias significantly outperforms the state-of-the-art in reducing proximity bias (GIPE) by at least 42.02%. It also reduces direct bias, adding minimal semantic disturbance, and achieves the best performance in a downstream application task (coreference resolution).

For agents

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

get_harvested_code_for_paper("2006.01938")
get_code_for_paper("2006.01938")
have("2006.01938")

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

curl https://syntology.ai/api/ran/2006.01938.json

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

[![Syntology run record](https://syntology.ai/api/ran/2006.01938.svg)](https://syntology.ai/paper/2006.01938)

Connect an agent — have() is free.