SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2406.19232 · EMNLP · 2024

RuBLiMP: Russian Benchmark of Linguistic Minimal Pairs

Alena Fenogenova, Ekaterina Taktasheva, Maxim Bazhukov, Kirill Koncha, E. Artemova

arXiv · PDF · Open in the Atlas

Code that ran

14 of 16 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; 1 honoured the contract we drafted, 0 violated it and 0 have no contract verdict; 13 failures of our instrument; 2 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/2406.19232.json · badge: /api/ran/2406.19232.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
natasha/slovnet canonical (classifier label)
classifier basis: census
8 ran · 2 no recorded run
RussianNLP/RuBLiMP canonical (classifier label)
classifier basis: census
6 ran · 0 no recorded run
FunctionStatusWhere it lives
bert_chunks Ran
ran; our contract draft was wrong, not the code
natasha/slovnet/slovnet/bert.py
code served (permissive licence) · get_code("8e7a3b391779e819")
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 re

def bert_chunks(text):
    return re.findall('\\w+|[^\\w\\s]', text)

# The function takes a single parameter 'text' which is a string
# It uses re.findall to find word characters and non-word non-space characters
# So text should be a 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
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
capitalize_indexes Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
RussianNLP/RuBLiMP/src/utils/utils.py
code served (permissive licence) · get_code("b11c77e1e013d613")
How it ran

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

import torch
import torch.nn as nn

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

# Analyze the function:
# - s: str - a string to capitalize
# - ind: List[int] - list of indices where characters should be capitalized

# Create test arguments
ARGS = {
    "s": "hello world",
    "ind": [0, 6, 7]
}
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
capitalize_word Ran
ran; our contract draft was wrong, not the code
RussianNLP/RuBLiMP/src/utils/utils.py
code served (permissive licence) · get_code("beffc6b278a14d3e")
How it ran

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

import torch
import numpy as np

# Set seed for reproducibility
torch.manual_seed(0)

# Analyze the function:
# capitalize_word(old_word: str, new_word: str) -> str
# - Takes two string parameters: old_word and new_word
# - old_word is checked with .isupper(), .islower()
# - new_word is processed and returned
# - Uses getcapital() and capitalize_indexes() helper functions

# Test case 1: old_word is all uppercase
old_word_1 = "HELLO"
new_word_1 = "world"

# Test case 2: old_word is all lowercase
old_word_2 = "hello"
new_word_2 = "WORLD"

# Test case 3: old_word has mixed case
old_word_3 = "HeLLo"
new_word_3 = "world"

# Use the first test case for ARGS
ARGS = {
    "old_word": "HELLO",
    "new_word": "world"
}
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 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_nouns_man_woman Ran
ran; our contract draft was wrong, not the code
RussianNLP/RuBLiMP/src/utils/data_loaders.py
code served (permissive licence) · get_code("78ee6585f194e2b4")
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.

from collections import Counter

# Create a minimal lemmas2ana structure that the function expects
# The function looks for entries with 'pos', 'gr', and semantic information
lemmas2ana = Counter({
    'doctor': [
        {
            'pos': 'S',
            'gr': 'm,anim,nom',
            'sem': 'profession human'
        },
        {
            'pos': 'S',
            'gr': 'f,anim,nom',
            'sem': 'profession human'
        }
    ],
    'teacher': [
        {
            'pos': 'S',
            'gr': 'm,anim,gen',
            'sem': 'profession human'
        },
        {
            'pos': 'S',
            'gr': 'f,anim,gen',
            'sem': 'profession human'
        }
    ],
    'cat': [
        {
            'pos': 'S',
            'gr': 'm,anim,nom',
            'sem': 'animal'
        },
        {
            'pos': 'S',
            'gr': 'f,anim,nom',
            'sem': 'animal'
        }
    ],
    'book': [
        {
            'pos': 'S',
            'gr': 'n,inan,nom',
            'sem': 'object'
        }
    ],
    'actor': [
        {
            'pos': 'V',
            'gr': 'm,anim',
            'sem': 'action'
        }
    ]
})

ARGS = {
    'lemmas2ana': lemmas2ana,
    'anim_only': True
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
none outside the standard library — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type dict — a dict output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a dict output the recorded ndim 0 and dtype family float describe that coercion, not the output, and are not stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
find_nouns_prof_commongender Ran
ran; our contract draft was wrong, not the code
RussianNLP/RuBLiMP/src/utils/data_loaders.py
code served (permissive licence) · get_code("ffbbf48436ea48e9")
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 json
import tempfile
import os
from collections import Counter

# Create a temporary JSON file with test vocabulary data
temp_dir = tempfile.mkdtemp()
temp_vocab_file = os.path.join(temp_dir, 'lemmas_enriched.json')

# Create test vocabulary data
vocab_data = {
    "teach": [
        {"sem": "t:prof d:nag v:action"},
        {"sem": "v:action"}
    ],
    "doctor": [
        {"sem": "t:prof n:person"},
        {"sem": "n:occupation"}
    ],
    "run": [
        {"sem": "v:action"},
        {"sem": "d:nag v:motion"}
    ],
    "lawyer": [
        {"sem": "t:prof d:nag n:person"}
    ],
    "cat": [
        {"sem": "n:animal a:furry"}
    ]
}

# Write to temporary file
with open(temp_vocab_file, 'w', encoding='utf-8') as f:
    json.dump(vocab_data, f)

# Load and prepare the lemmas2ana data structure
def load_vocab_local(filename: str):
    with open(filename, 'r', encoding='utf-8') as f:
        form2analyses = json.load(f)
    for form, analyses in form2analyses.items():
        for analysis in analyses:
            sem_str = analysis['sem']
            analysis['sem'] = set(sem_str.split())
    return Counter(form2analyses)

lemmas2ana_test = load_vocab_local(temp_vocab_file)

# Define ARGS
ARGS = {
    'lemmas2ana': lemmas2ana_test,
    'anim_only': True
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
none outside the standard library — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type dict — a dict output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a dict output the recorded ndim 0 and dtype family float describe that coercion, not the output, and are not stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
format_bio Ran
ran; our contract draft was wrong, not the code
natasha/slovnet/slovnet/bio.py
code served (permissive licence) · get_code("d9def9c954ea4370")
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 = {
    "part": "B",
    "type": "PER"
}
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
format_conll_tag Ran
ran; our contract draft was wrong, not the code
natasha/slovnet/slovnet/conll.py
code served (permissive licence) · get_code("60009883d53fc392")
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 format_conll_tag(pos, feats):
    if not feats:
        return pos
    feats = '|'.join(('%s=%s' % (_, feats[_]) for _ in sorted(feats)))
    return '%s|%s' % (pos, feats)

ARGS = {
    'pos': 'NOUN',
    'feats': {'Case': 'Nom', 'Number': 'Sing', 'Gender': 'Masc'}
}
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
getcapital Ran
ran, honoured the contract we drafted
RussianNLP/RuBLiMP/src/utils/utils.py
code served (permissive licence) · get_code("1b17728d03301242")
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

ARGS = {
    "s": "HelloWorld"
}
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
load_json Ran
ran; our contract draft was wrong, not the code
natasha/slovnet/slovnet/io.py
code served (permissive licence) · get_code("ea0488d1412c27b2")
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 json
import tempfile
import os

# Create a temporary JSON file for the function to read
temp_dir = tempfile.mkdtemp()
json_file_path = os.path.join(temp_dir, "test.json")

# Write a valid JSON file
test_data = {"key": "value", "number": 42, "list": [1, 2, 3]}
with open(json_file_path, 'w', encoding='utf8') as f:
    json.dump(test_data, f)

ARGS = {
    'path': json_file_path,
    'encoding': 'utf8'
}
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
load_vocab Ran
ran; our contract draft was wrong, not the code
RussianNLP/RuBLiMP/src/utils/data_loaders.py
code served (permissive licence) · get_code("dbe2a14f4e641a48")
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 Counter. 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 json
import tempfile
import os
from collections import Counter

# Create a temporary JSON file with the expected structure
temp_dir = tempfile.mkdtemp()
temp_file = os.path.join(temp_dir, 'lemmas_enriched.json')

# Create sample vocabulary data
form2analyses = {
    "run": [
        {"sem": "move fast locomotion"},
        {"sem": "manage operate"}
    ],
    "walk": [
        {"sem": "move slow locomotion"},
        {"sem": "stroll"}
    ],
    "talk": [
        {"sem": "speak communicate"}
    ]
}

# Write to temporary file
with open(temp_file, 'w', encoding='utf-8') as f:
    json.dump(form2analyses, f, ensure_ascii=False)

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 Counter — a Counter output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a Counter 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
parse_bio Ran
ran; our contract draft was wrong, not the code
natasha/slovnet/slovnet/bio.py
code served (permissive licence) · get_code("e686b8ce6606b5a9")
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 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

def parse_bio(tag):
    if '-' in tag:
        part, type = tag.split('-', 1)
    else:
        part = tag
        type = None
    return (part, type)

ARGS = {
    'tag': 'B-PER'
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type tuple, ndim 1, dtype family float — a tuple output has no array shape of its own; none is stated
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
parse_conll_tag Ran
ran; our contract draft was wrong, not the code
natasha/slovnet/slovnet/conll.py
code served (permissive licence) · get_code("b0d5d5c95bce546f")
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 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 torch.nn as nn

# Analyze the function parse_conll_tag:
# - Takes a single parameter: tag (a string)
# - Checks if '|' is in tag
# - If not present, returns (tag, {})
# - If present, splits on first '|' to get pos and feats
# - Further processes feats by splitting on '|' and creating a dict with '=' splits
# - Returns (pos, feats_dict)

# The function expects a string in CoNLL format
# Example valid tags:
# - "NOUN" (no features)
# - "NOUN|Case=Nom|Number=Sing" (with features)

ARGS = {
    'tag': 'NOUN|Case=Nominative|Number=Singular|Gender=Masculine'
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type tuple, ndim 1, dtype family float — a tuple output has no array shape of its own; none is stated
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
safe_wordpiece Ran
ran; our contract draft was wrong, not the code
natasha/slovnet/slovnet/bert.py
code served (permissive licence) · get_code("ae02bed8c6ac0e37")
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

class MockVocab:
    def __init__(self):
        self.item_ids = {
            'hello': 0,
            'world': 1,
            '##world': 2,
            'he': 3,
            '##llo': 4,
            'wo': 5,
            '##rld': 6,
            'h': 7,
            '##e': 8,
            'w': 9,
            '##o': 10,
            '##r': 11,
            '##l': 12,
            '##d': 13,
        }

vocab = MockVocab()
text = "hello"

ARGS = {
    'text': text,
    'vocab': vocab
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type 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
wordpiece Ran
ran; our contract draft was wrong, not the code
natasha/slovnet/slovnet/bert.py
code served (permissive licence) · get_code("70e55146aa0bea49")
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.

# Analyze the function to understand what it needs:
# - text: a string to tokenize
# - vocab: an object with an `item_ids` attribute (dict-like, supporting `in` operator)
# - prefix: a string prefix (has default '##')

# The function treats vocab.item_ids as a set/dict that supports membership testing

class SimpleVocab:
    def __init__(self, item_ids):
        self.item_ids = item_ids

# Create a simple vocabulary with some tokens
vocab_dict = {
    'hello': 0,
    'world': 1,
    '##world': 2,
    'he': 3,
    '##llo': 4,
}

vocab = SimpleVocab(vocab_dict)

# Simple test text to tokenize
text = "hello"

# Build ARGS dict with all parameters
ARGS = {
    'text': text,
    'vocab': vocab,
    'prefix': '##'
}
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
flatten_cross_entropy Not yet run
unverified (harvested, no recorded run)
natasha/slovnet/slovnet/loss.py
code served (permissive licence) · get_code("2083fcb410f76cfd")
How it ran

no invocation recorded.

masked_flatten_cross_entropy Not yet run
unverified (harvested, no recorded run)
natasha/slovnet/slovnet/loss.py
code served (permissive licence) · get_code("6b1d272e499dcae8")
How it ran

no invocation recorded.

Also present in other papers' code

Of 16 samples lifted from this paper's linked repositories, 16 carry a function identity and 0 do not. Of the 16 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 16 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

Minimal pairs are a well-established approach to evaluating the grammatical knowledge of language models. However, existing resources for minimal pairs address a limited number of languages and lack diversity of language-specific grammatical phenomena. This paper introduces the Russian Benchmark of Linguistic Minimal Pairs (RuBLiMP), which includes 45k pairs of sentences that differ in grammaticality and isolate a morphological, syntactic, or semantic phenomenon. In contrast to existing benchmarks of linguistic minimal pairs, RuBLiMP is created by applying linguistic perturbations to automatically annotated sentences from open text corpora and decontaminating test data. We describe the data collection protocol and present the results of evaluating 25 language models in various scenarios. We find that the widely used LMs for Russian are sensitive to morphological and agreement-oriented contrasts, but fall behind humans on phenomena requiring the understanding of structural relations, negation, transitivity, and tense. RuBLiMP, the codebase, and other materials are publicly available. † Work is partially done while at HSE University. Parse Sentence Extraction (a) root spal 'sleep' Vpervye 'For the first time' kosmonavt 'an astronaut' nevesomosti 'zero gravity' v 'in' obl case nsubj advmod Sentence Annotation Minimal Pair Generation adposition government Vpervye kosmonavt spal v nevesomosti. Vpervye kosmonavt spal v nevesomost'. subject-predicate agreement (number) Vpervye kosmonavt spal v nevesomosti. Vpervye kosmonavt spali v nevesomosti. subject-predicate agreement (gender) Vpervye kosmonavt spal v nevesomosti. Vpervye kosmonavt spalo v nevesomosti.

For agents

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

get_harvested_code_for_paper("2406.19232")
get_code_for_paper("2406.19232")
have("2406.19232")

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

curl https://syntology.ai/api/ran/2406.19232.json

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

[![Syntology run record](https://syntology.ai/api/ran/2406.19232.svg)](https://syntology.ai/paper/2406.19232)

Connect an agent — have() is free.