SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2605.28375 · 2026

PrionNER: A Named Entity Recognition Dataset for Prion Disease Biomedical Literature

Yuji Matsumoto, Akiko Aizawa, An Dao, Nhan Ly, Thao Tran

arXiv · PDF

Code that ran

22 of 22 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 0 have no contract verdict; 22 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/2605.28375.json · badge: /api/ran/2605.28375.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
daotuanan/PrionNER canonical (classifier label)
classifier basis: census
22 ran · 0 no recorded run
FunctionStatusWhere it lives
bio_to_entities Ran
ran; our contract draft was wrong, not the code
daotuanan/PrionNER/code/evaluate_conll_predictions.py
code served (permissive licence) · get_code("5a82a9027d979c92")
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 2, 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
from typing import Dict, List, Sequence, Tuple

# The function bio_to_entities takes a sequence of BIO tags and converts them to entities
# BIO tags are strings like "B-PER", "I-PER", "O", etc.
# The function iterates through tags and extracts entity spans

# Create a simple valid input: a sequence of BIO tags
tags = [
    "B-PER",      # Begin Person entity at index 0
    "I-PER",      # Inside Person entity at index 1
    "O",          # Outside any entity at index 2
    "B-LOC",      # Begin Location entity at index 3
    "I-LOC",      # Inside Location entity at index 4
    "O",          # Outside any entity at index 5
]

ARGS = {
    "tags": tags
}
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 2, 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
build_tag_vocab Ran
ran; our contract draft was wrong, not the code
daotuanan/PrionNER/code/train_bert_ner.py
code served (permissive licence) · get_code("2b7457bfbe0035ee")
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
from typing import Dict, Iterable, List, Sequence, Tuple

# Setup: create valid arguments for build_tag_vocab
entity_labels = ['PERSON', 'ORG', 'LOCATION', 'DATE']

ARGS = {
    'entity_labels': entity_labels
}
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
compute_metrics_from_tags Ran
ran; our contract draft was wrong, not the code
daotuanan/PrionNER/code/evaluate_conll_predictions.py
code served (permissive licence) · get_code("6720ad9ffde1acbe")
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 torch
import numpy as np
from collections import Counter
from typing import Dict, List, Sequence, Tuple

# Set seed for reproducibility
torch.manual_seed(0)

# Create sample BIO-tagged sequences
# BIO format: B-LABEL (beginning), I-LABEL (inside), O (outside)

gold_sequences = [
    ['B-PER', 'I-PER', 'O', 'B-LOC', 'I-LOC'],
    ['B-ORG', 'O', 'B-PER', 'O', 'O'],
    ['O', 'O', 'B-LOC', 'O', 'O']
]

pred_sequences = [
    ['B-PER', 'I-PER', 'O', 'B-LOC', 'O'],
    ['B-ORG', 'O', 'B-PER', 'I-PER', 'O'],
    ['O', 'B-LOC', 'I-LOC', 'O', 'O']
]

ARGS = {
    'gold_sequences': gold_sequences,
    'pred_sequences': pred_sequences
}
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 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
dedupe_preserve_order Ran
ran; our contract draft was wrong, not the code
daotuanan/PrionNER/code/export_w2ner_dataset.py
code served (permissive licence) · get_code("b5c74fe7c41b8170")
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)

ARGS = {
    "values": [1, 2, 3, 2, 1, 4, 5, 4, 3]
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type 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
extract_spanned_text Ran
ran; our contract draft was wrong, not the code
daotuanan/PrionNER/code/brat_to_json_conll.py
code served (permissive licence) · get_code("5ee3b795e4dd2cf7")
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.

from dataclasses import dataclass
from typing import Sequence

@dataclass(frozen=True)
class Span:
    start: int
    end: int

# Setup
text = "hello world from python"
spans = [
    Span(start=0, end=5),    # "hello"
    Span(start=6, end=11),   # "world"
    Span(start=17, end=23),  # "python"
]

ARGS = {
    "text": text,
    "spans": spans,
}
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
filter_annotation_lines Ran
ran; our contract draft was wrong, not the code
daotuanan/PrionNER/code/prepare_raw_split.py
code served (permissive licence) · get_code("d2efbeec4e89afa5")
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
from typing import Iterable, List

# The function filter_annotation_lines takes lines (an Iterable[str])
# It iterates through lines, strips whitespace, skips empty lines,
# skips lines containing '\tChoose_', and keeps the rest

# Create a list of test annotation lines
test_lines = [
    "annotation 1",
    "   ",  # empty/whitespace line
    "\tannotation 2",
    "line with \tChoose_ in it",  # excluded
    "valid annotation 3",
    "",  # empty line
    "\tChoose_something",  # excluded prefix
    "  final annotation  ",  # whitespace to be stripped
]

ARGS = {
    "lines": test_lines
}
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
load_json Ran
ran; our contract draft was wrong, not the code
daotuanan/PrionNER/code/convert_w2ner_output_to_brat.py
code served (permissive licence) · get_code("f5868f251e1785a0")
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
from pathlib import Path

# Create a temporary JSON file with some valid JSON content
with tempfile.NamedTemporaryFile(mode='w', suffix='.json', delete=False, encoding='utf-8') as f:
    json.dump({"key": "value", "number": 42, "list": [1, 2, 3]}, f)
    temp_path = Path(f.name)

ARGS = {
    "path": temp_path
}
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_label_map Ran
ran; our contract draft was wrong, not the code
daotuanan/PrionNER/code/build_coarse_dataset.py
code served (permissive licence) · get_code("4c52123532019b99")
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
from pathlib import Path
from tempfile import TemporaryDirectory

# Create a minimal valid schema JSON structure
schema_dict = {
    "entity_types": [
        {
            "coarse_type": "PERSON",
            "fine_types": [
                {
                    "fine_type": "person_name",
                    "brat_label": "PERSON_NAME"
                },
                {
                    "fine_type": "person_title",
                    "brat_label": "PERSON_TITLE"
                }
            ]
        },
        {
            "coarse_type": "LOCATION",
            "fine_types": [
                {
                    "fine_type": "city",
                    "brat_label": "CITY"
                },
                {
                    "fine_type": "country",
                    "brat_label": "COUNTRY"
                }
            ]
        },
        {
            "coarse_type": "ORGANIZATION",
            "fine_types": [
                {
                    "fine_type": "company",
                    "brat_label": "COMPANY"
                }
            ]
        }
    ]
}

# Create a temporary directory and file with the schema
temp_dir = TemporaryDirectory()
schema_path = Path(temp_dir.name) / "schema.json"
schema_path.write_text(json.dumps(schema_dict), encoding='utf-8')

ARGS = {
    "schema_path": schema_path
}
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_pairs Ran
ran; our contract draft was wrong, not the code
daotuanan/PrionNER/code/prepare_raw_split.py
code served (permissive licence) · get_code("e4b6a2d6715aa493")
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 2, 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 tempfile
from pathlib import Path

# Create a temporary directory with test .txt and .ann files
temp_dir = tempfile.TemporaryDirectory()
input_dir = Path(temp_dir.name)

# Create paired .txt and .ann files
txt_file_1 = input_dir / "document1.txt"
ann_file_1 = input_dir / "document1.ann"
txt_file_2 = input_dir / "document2.txt"
ann_file_2 = input_dir / "document2.ann"

txt_file_1.write_text("Sample text content 1")
ann_file_1.write_text("Sample annotation 1")
txt_file_2.write_text("Sample text content 2")
ann_file_2.write_text("Sample annotation 2")

ARGS = {
    'input_dir': input_dir
}
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 2, 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_w2ner_modules Ran
ran; our contract draft was wrong, not the code
daotuanan/PrionNER/code/run_w2ner_no_dev.py
code served (permissive licence) · get_code("a9afd9edaa0f26eb")
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 sys
from pathlib import Path
from unittest.mock import MagicMock
import tempfile
import os

# Create a temporary directory to act as w2ner_root
temp_dir = tempfile.mkdtemp()
w2ner_root = Path(temp_dir)

# Create minimal module files that will be imported
config_content = """
class Config:
    def __init__(self):
        self.hidden_size = 768
        self.num_layers = 2
"""

data_loader_content = """
def create_data_loader(dataset, batch_size=32):
    return []
"""

model_content = """
class W2NERModel:
    def __init__(self, config):
        self.config = config
"""

utils_content = """
def setup_logging():
    pass
"""

# Write the module files
(w2ner_root / 'config.py').write_text(config_content)
(w2ner_root / 'data_loader.py').write_text(data_loader_content)
(w2ner_root / 'model.py').write_text(model_content)
(w2ner_root / 'utils.py').write_text(utils_content)

ARGS = {
    'w2ner_root': w2ner_root
}
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 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
merge_token_indexes Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
daotuanan/PrionNER/code/convert_w2ner_output_to_brat.py
code served (permissive licence) · get_code("278255e94d2f1fe6")
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 2, 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
from typing import Dict, Sequence, List, Tuple

# The function merge_token_indexes takes:
# 1. indexes: Sequence[int] - a sequence of token indices
# 2. token_offsets: Sequence[Dict[str, int]] - a sequence of dicts with 'start' and 'end' keys

# Looking at the function body:
# - It accesses token_offsets[index]['start'] and token_offsets[index]['end']
# - indexes contains valid indices into token_offsets
# - The function sorts indexes and merges consecutive ones based on token offsets

# Build valid arguments
indexes = [0, 1, 2, 4, 5]  # Some indices, with a gap at 3

token_offsets = [
    {'start': 0, 'end': 5},      # index 0
    {'start': 5, 'end': 10},     # index 1
    {'start': 10, 'end': 15},    # index 2
    {'start': 20, 'end': 25},    # index 3
    {'start': 15, 'end': 20},    # index 4
    {'start': 20, 'end': 25},    # index 5
]

ARGS = {
    "indexes": indexes,
    "token_offsets": token_offsets,
}
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 2, 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
normalize_formats Ran
ran; our contract draft was wrong, not the code
daotuanan/PrionNER/code/brat_to_json_conll.py
code served (permissive licence) · get_code("1f662b56e813d43f")
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 numpy as np
from typing import Dict, Sequence, Tuple

# Construct valid arguments for normalize_formats
raw_formats_basic = ('json', 'conll', 'xml')
raw_formats_with_both = ('json', 'both', 'conll')
raw_formats_single = ('json',)
raw_formats_both_only = ('both',)

ARGS = {
    'raw_formats': ('json', 'conll', 'xml')
}
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 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
normalize_label_input_mode Ran
ran; our contract draft was wrong, not the code
daotuanan/PrionNER/code/gliner2_conll_predict.py
code served (permissive licence) · get_code("9d34db2fe7dfc5d8")
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 = {
    "label_input_mode": "short"
}
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
normalize_label_list Ran
ran; our contract draft was wrong, not the code
daotuanan/PrionNER/code/gliner2_conll_predict.py
code served (permissive licence) · get_code("6e7770475cd00bc3")
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
from typing import List

# Define the arguments for normalize_label_list
labels = [
    "B-PER",
    "I-PER",
    "O",
    "B-LOC",
    "S-ORG",
    "E-MISC",
    "  ",
    "PER",
    "B-PER",  # duplicate
    "",
    "LOC",
    "I-LOC"
]

ARGS = {
    "labels": labels
}
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
parse_span_list Ran
ran; our contract draft was wrong, not the code
daotuanan/PrionNER/code/brat_to_json_conll.py
code served (permissive licence) · get_code("8accd4a1e98e7093")
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.

from dataclasses import dataclass
from typing import Dict, Iterable, List, Sequence, Tuple

@dataclass(frozen=True)
class Span:
    start: int
    end: int

def parse_span_list(raw: str) -> Tuple[Span, ...]:
    spans: List[Span] = []
    for chunk in raw.split(';'):
        start_str, end_str = chunk.split()
        start = int(start_str)
        end = int(end_str)
        if start >= end:
            raise ValueError(f'Invalid span {chunk!r}.')
        spans.append(Span(start=start, end=end))
    return tuple(spans)

ARGS = {
    'raw': '0 5;10 20;3 8'
}
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 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
read_conll Ran
ran; our contract draft was wrong, not the code
daotuanan/PrionNER/code/evaluate_conll_predictions.py
code served (permissive licence) · get_code("5149436dbbc79df7")
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.

from pathlib import Path
import tempfile

# Create a temporary CoNLL-formatted file
with tempfile.NamedTemporaryFile(mode='w', suffix='.conll', delete=False, encoding='utf-8') as f:
    f.write("The\tDET\n")
    f.write("cat\tNOUN\n")
    f.write("sat\tVERB\n")
    f.write("\n")
    f.write("Dogs\tNOUN\n")
    f.write("run\tVERB\n")
    f.write("\n")
    temp_path = f.name

conll_path = Path(temp_path)

ARGS = {
    'path': conll_path
}
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 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
read_schema_labels Ran
ran; our contract draft was wrong, not the code
daotuanan/PrionNER/code/gliner2_conll_predict.py
code served (permissive licence) · get_code("ccc172ad50696aa8")
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 json
from pathlib import Path
from tempfile import TemporaryDirectory
import os

# Create a temporary directory and schema file
temp_dir = TemporaryDirectory()
schema_path = os.path.join(temp_dir.name, "schema.json")

# Create a valid schema JSON file
schema_data = {
    "entity_types": [
        {
            "coarse_type": "PERSON",
            "definition": "A person's name",
            "examples": ["John", "Alice"],
            "fine_types": [
                {
                    "fine_type": "PERSON_FIRST",
                    "definition": "First name of a person",
                    "examples": ["John", "Alice"]
                },
                {
                    "fine_type": "PERSON_LAST",
                    "definition": "Last name of a person",
                    "examples": ["Smith", "Johnson"]
                }
            ]
        },
        {
            "coarse_type": "LOCATION",
            "definition": "A geographic location",
            "examples": ["New York", "Paris"],
            "fine_types": [
                {
                    "fine_type": "LOCATION_CITY",
                    "definition": "A city name",
                    "examples": ["New York", "Paris"]
                }
            ]
        },
        {
            "coarse_type": "ORGANIZATION",
            "definition": "An organization name",
            "examples": ["Google", "Microsoft"],
            "fine_types": []
        }
    ]
}

# Write schema file
with open(schema_path, 'w', encoding='utf-8') as f:
    json.dump(schema_data, f)

ARGS = {
    "schema_file": schema_path,
    "label_input_mode": "def"
}
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 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
resolve_model_name Ran
ran; our contract draft was wrong, not the code
daotuanan/PrionNER/code/train_bert_ner.py
code served (permissive licence) · get_code("57dfe5261a9134b3")
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

# resolve_model_name takes a single parameter 'raw' which is a string
# The function calls raw.lower() and uses it as a key in MODEL_ALIASES.get()
# It returns either the aliased model name or the original raw string

ARGS = {
    'raw': 'biobert'
}
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
resolve_model_name Ran
ran; our contract draft was wrong, not the code
daotuanan/PrionNER/code/train_gliner2_ner.py
code served (permissive licence) · get_code("ea7bb9a0ef66200b")
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

MODEL_ALIASES = {'gliner2_base': 'fastino/gliner2-base-v1', 'gliner2_large': 'fastino/gliner2-large-v1'}

def resolve_model_name(raw: str) -> str:
    return MODEL_ALIASES.get(raw.lower(), raw)

ARGS = {
    'raw': 'gliner2_base'
}
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
span_text Ran
ran; our contract draft was wrong, not the code
daotuanan/PrionNER/code/build_coarse_dataset.py
code served (permissive licence) · get_code("2ec905b7f5d5735d")
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.

class Span:
    def __init__(self, start, end):
        self.start = start
        self.end = end

span1 = Span(0, 5)
span2 = Span(10, 15)
span3 = Span(20, 25)

ARGS = {
    'spans': (span1, span2, span3)
}
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
span_texts Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
daotuanan/PrionNER/code/convert_w2ner_output_to_brat.py
code served (permissive licence) · get_code("d22dd17ae17f7ed6")
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

# Construct arguments for span_texts function
# The function takes a text string and an iterable of (start, end) tuples
# and returns a string joining the text spans

text = "The quick brown fox jumps over the lazy dog"
spans = [(0, 3), (4, 9), (10, 15), (20, 23)]

ARGS = {
    "text": text,
    "spans": spans
}
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
strip_footer_markers Ran
ran; our contract draft was wrong, not the code
daotuanan/PrionNER/code/prepare_raw_split.py
code served (permissive licence) · get_code("ccb47bea461c7c85")
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

# Analyze the function:
# - strip_footer_markers takes a single parameter: text (str)
# - It processes the text by iterating through lines
# - It breaks when a line starts with one of the MARKER_PREFIXES
# - It returns the kept lines joined with newlines, with trailing newlines stripped

# Create test cases for the text parameter
test_text_with_marker = """This is line 1
This is line 2
<Related> This line should not be included
This line also should not be included"""

test_text_without_marker = """This is line 1
This is line 2
This is line 3"""

test_text_with_done_marker = """Header line 1
Header line 2
[DONE] This should not appear
Footer line"""

test_text_with_check_marker = """Some content
[CHECK] Stop here
More content"""

# Use one of the test cases
ARGS = {
    "text": test_text_with_marker
}
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

Also present in other papers' code

Of 22 samples lifted from this paper's linked repositories, 22 carry a function identity and 0 do not. Of the 21 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 21 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

Prion diseases are rare, rapidly progressive, and fatal neurodegenerative disorders that remain difficult to diagnose, particularly in their early stages because of nonspecific clinical presentations. However, to our knowledge, there is no publicly available prion-disease-focused dataset designed to capture a broad range of clinically relevant entities from the biomedical literature. We introduce PrionNER, a manually annotated named entity recognition dataset for prion disease clinical information in PubMed abstracts. The current release comprises 317 abstracts, 2,943 sentences, and 6,955 text-bound entity annotations spanning 15 coarse-grained and 31 fine-grained clinically oriented entity types covering diseases, symptoms, diagnostics, findings, anatomy, treatments, and temporal and statistical evidence. Inter-annotator agreement reaches 81.78 exactmatch F1, indicating strong annotation consistency. We benchmark supervised BERT baselines, W2NER, and zero-shot extractors on Pri-onNER. W2NER is the strongest supervised model, and Gemma-4-31B is the strongest zeroshot model, but the benchmark remains challenging, especially for structurally complex mentions and fine-grained clinically adjacent label distinctions. PrionNER provides a clinically grounded benchmark for prion-disease information extraction and supports research on rare-disease biomedical NLP under lowresource, fine-grained, and non-flat extraction conditions. The dataset, annotation guidelines, and evaluation scripts are available at https: //github.com/daotuanan/PrionNER/.

For agents

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

get_harvested_code_for_paper("2605.28375")
get_code_for_paper("2605.28375")
have("2605.28375")

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

curl https://syntology.ai/api/ran/2605.28375.json

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

[![Syntology run record](https://syntology.ai/api/ran/2605.28375.svg)](https://syntology.ai/paper/2605.28375)

Connect an agent — have() is free.