SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2501.09409 · EMNLP · 2025

Mind the Inclusivity Gap: Multilingual Gender-Neutral Translation Evaluation with MGENTE

Matteo Negri, Anne Lauscher, Beatrice Savoldi, Luisa Bentivogli, Giuseppe Attanasio, Bruno Kessler, Janiça Hackenbuchner, Eleonora Cupin, Eleni Gkovedarou, Andrea ♢△, ♢ Fondazione

arXiv · PDF · Open in the Atlas

Code that ran

8 of 8 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 8 have no contract verdict; 0 failures of our instrument; 0 no recorded run), measured 2026-11-22; 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/2501.09409.json · badge: /api/ran/2501.09409.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
g8a9/mgente-gap canonical (classifier label)
classifier basis not recorded on the link
8 ran · 0 no recorded run
FunctionStatusWhere it lives
build_prompt_filename Ran
ran on a synthesized input
g8a9/mgente-gap/src/utils.py
code served (permissive licence) · get_code("ebe6cbdc25ae4caf")
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

ARGS = {
    'use_system': True,
    'use_guidelines': False,
    'n_shots': 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 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
find_spans Ran
ran on a synthesized input
g8a9/mgente-gap/src/attribute.py
code served (permissive licence) · get_code("b4650a8d3700da15")
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 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

def find_spans(l: list, sl: list):
    """Find the first span that matches the sublist sl in list l."""
    assert len(l) >= len(sl)
    for i in range(len(l) - len(sl) + 1):
        if l[i:i + len(sl)] == sl:
            return (i, i + len(sl))

# Construct valid arguments for find_spans
# The function finds the first occurrence of sublist sl in list l
# Both parameters should be lists that support slicing and equality comparison

l = [1, 2, 3, 4, 5, 6, 7]
sl = [4, 5, 6]

ARGS = {
    "l": l,
    "sl": sl
}
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 tuple, ndim 1, dtype family int — 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
load_csv_input_data Ran
ran on a synthesized input
g8a9/mgente-gap/src/gnt_eval/gpt4_evaluate.py
code served (permissive licence) · get_code("c8dc606ac8d55986")
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 csv
import tempfile
import os

# Create a temporary CSV file for testing
temp_dir = tempfile.mkdtemp()
csv_filename = os.path.join(temp_dir, "test_data.csv")

# Write sample CSV data
with open(csv_filename, 'w', newline='') as f:
    writer = csv.DictWriter(f, fieldnames=['name', 'age', 'city'])
    writer.writeheader()
    writer.writerow({'name': 'Alice', 'age': '30', 'city': 'New York'})
    writer.writerow({'name': 'Bob', 'age': '25', 'city': 'Los Angeles'})
    writer.writerow({'name': 'Charlie', 'age': '35', 'city': 'Chicago'})

ARGS = {
    'filename': csv_filename
}
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 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
load_schema Ran
ran on a synthesized input
g8a9/mgente-gap/src/gnt_eval/local_evaluate.py
code served (permissive licence) · get_code("6e6d57bc5ad80823")
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 tempfile
import os

# Create a temporary file with schema content
temp_dir = tempfile.mkdtemp()
temp_file = os.path.join(temp_dir, "test_schema.txt")

schema_content = """
{
  "type": "object",
  "properties": {
    "name": {"type": "string"},
    "age": {"type": "integer"}
  }
}
"""

with open(temp_file, 'w') as f:
    f.write(schema_content)

ARGS = {
    "filename": 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 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
load_system_prompt Ran
ran on a synthesized input
g8a9/mgente-gap/src/gnt_eval/local_evaluate.py
code served (permissive licence) · get_code("9bfc3913b678e212")
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 tempfile
import os

# Create a temporary file with some system prompt content
temp_file = tempfile.NamedTemporaryFile(mode='w', delete=False, suffix='.txt')
temp_file.write("You are a helpful assistant.")
temp_file.close()

ARGS = {
    'filename': temp_file.name
}
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
log_arguments Ran
ran on a synthesized input
g8a9/mgente-gap/src/utils.py
code served (permissive licence) · get_code("e704018225145fc1")
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 function. 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 logging
import torch

# Set random seed for reproducibility
torch.manual_seed(0)

# Create a logger instance
test_logger = logging.getLogger("test_logger")
test_logger.setLevel(logging.INFO)

# Add a simple handler to capture logs
handler = logging.StreamHandler()
handler.setLevel(logging.INFO)
test_logger.addHandler(handler)

# Define the ARGS dictionary with the only parameter: logger
ARGS = {
    "logger": test_logger
}
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 function — a function output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a function 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
sample_demonstrations Ran
ran on a synthesized input
g8a9/mgente-gap/src/prompts.py
code served (permissive licence) · get_code("f48178e55ad2587c")
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 random

random.seed(0)

demonstrations = [
    {
        'type': 'gendered',
        'user': 'What is the capital of France?',
        'assistant': 'The capital of France is Paris.'
    },
    {
        'type': 'gendered',
        'user': 'Who was the first president?',
        'assistant': 'George Washington was the first president of the United States.'
    },
    {
        'type': 'gendered',
        'user': 'What is 2 + 2?',
        'assistant': '2 + 2 equals 4.'
    },
    {
        'type': 'neutral',
        'user': 'Explain photosynthesis.',
        'assistant': 'Photosynthesis is the process by which plants convert light energy into chemical energy.'
    },
    {
        'type': 'neutral',
        'user': 'What is the speed of light?',
        'assistant': 'The speed of light is approximately 299,792,458 meters per second.'
    },
    {
        'type': 'neutral',
        'user': 'Define gravity.',
        'assistant': 'Gravity is a natural phenomenon by which all things with mass are brought toward each other.'
    }
]

ARGS = {
    'n': 4,
    'demonstrations': demonstrations
}
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 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
sanitize_model_name Ran
ran on a synthesized input
g8a9/mgente-gap/src/utils.py
code served (permissive licence) · get_code("59fcc0625c0d1b78")
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

def sanitize_model_name(model_name):
    return model_name.replace('/', '--')

ARGS = {
    'model_name': 'bert-base/uncased'
}
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

Also present in other papers' code

Of 8 samples lifted from this paper's linked repositories, 8 carry a function identity and 0 do not. Of the 8 distinct functions with an identity, 0 are also present in other papers whose code we hold (0 with at least one paper dated earlier, 0 the same month, 0 later) and 8 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.

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

Avoiding the propagation of undue (binary) gender inferences and default masculine language remains a key challenge towards inclusive multilingual technologies, particularly when translating into languages with extensive gendered morphology. Gender-neutral translation (GNT) represents a linguistic strategy towards fairer communication across languages. However, research on GNT is limited to a few resources and language pairs. To address this gap, we introduce MGENTE, an expert-curated resource, and use it to conduct the first systematic multilingual evaluation of inclusive translation with state-of-the-art instruction-following language models (LMs). Experiments on enes/de/it/el reveal that while models can recognize when neutrality is appropriate, they cannot consistently produce neutral translations, limiting their usability. To probe this behavior, we enrich our evaluation with interpretability analyses that identify task-relevant features and offer initial insights into the internal dynamics of LM-based GNT.

For agents

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

get_harvested_code_for_paper("2501.09409")
get_code_for_paper("2501.09409")
have("2501.09409")

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

curl https://syntology.ai/api/ran/2501.09409.json

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

[![Syntology run record](https://syntology.ai/api/ran/2501.09409.svg)](https://syntology.ai/paper/2501.09409)

Connect an agent — have() is free.