SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2408.15545 · ICLR · 2025

SciLitLLM: How to Adapt LLMs for Scientific Literature Understanding

Xiang Wang, Guolin Ke, Linfeng Zhang, Yaorui Shi, Sihang Li, Jin Huang, Jiaxi Zhuang, Xiaochen Cai, Mingjun Xu, Hengxing Cai

arXiv · PDF · Open in the Atlas

Code that ran

13 of 13 samples (functions and classes) lifted from this paper's linked repositories executed on a synthesized input in Syntology's jail (0 of them constructed an object rather than computing a result; 2 honoured the contract we drafted, 0 violated it and 0 have no contract verdict; 11 failures of our instrument; 0 no recorded run); 1 of them was also run on the host, not in the jail, by the peer-agreement lane that set its level, measured 2026-11-22; repository links from the Papers with Code archive attributed under CC BY-SA, no affiliation; bodies inline only under a permissive licence as GitHub detects it. Not a reproduction of the paper's results.

"Ran" means the function executed on a synthesized input and returned a value. It is not a reproduction of the paper's results. This record, dated, as JSON: /api/ran/2408.15545.json · badge: /api/ran/2408.15545.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
dptech-corp/Uni-SMART canonical (classifier label)
classifier basis: census
11 ran · 0 no recorded run
dptech-corp/uni-smart canonical (classifier label)
classifier basis: classifier_2025
2 ran · 0 no recorded run
FunctionStatusWhere it lives
chunk_text Ran
ran; our contract draft was wrong, not the code
dptech-corp/uni-smart/SciLitLLM/cpt/format_grammar_correction.py
code served (permissive licence) · get_code("82a2172211c2ce1f")
How it ran

Ran in Syntology's jail on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1) and returned list, ndim 1, dtype family float; no output shape is stated. The image it ran in was not recorded. The comparison behind this sample's level is a different run: the peer-agreement lane ran it again on the host, not in the jail (report dated 2026-09-05T00:41:09Z). 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

# Define ARGS dict with valid arguments for chunk_text function
ARGS = {
    "text": "The quick brown fox jumps over the lazy dog and runs through the forest",
    "chunk_size": 3
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type 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
clean_text Ran
ran; our contract draft was wrong, not the code
dptech-corp/Uni-SMART/SciLitLLM/sft/helper/parse_pdfs.py
code served (permissive licence) · get_code("2e366df93e0e0280")
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 re

def clean_text(text):
    """
    Clean the extracted text by removing non-alphanumeric characters and converting to lowercase.
    Also, filter out numbers.
    """
    text = re.sub('\\W+', ' ', text)
    text = re.sub('\\d+', '', text)
    text = text.lower()
    return text

# The function takes a single parameter 'text' which should be a string
ARGS = {
    'text': "Hello World! This is a Test123 with Numbers456 and Special@Characters#"
}
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
get_lev_sim Ran
ran, honoured the contract we drafted
dptech-corp/Uni-SMART/SciLitLLM/sft/helper/dedup.py
code served (permissive licence) · get_code("4e655fe9976690e7")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned float, ndim 0, dtype family float, shape () as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import Levenshtein

# Create two sample dictionaries with 'text' and 'answer' keys
d1 = {
    'text': 'hello world',
    'answer': 'yes'
}

d2 = {
    'text': 'hello there',
    'answer': 'no'
}

ARGS = {
    'd1': d1,
    'd2': d2
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
Levenshtein — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type float, ndim 0, dtype family float, shape () — shape () as a later run recorded it
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
get_qa_pair Ran
ran; our contract draft was wrong, not the code
dptech-corp/Uni-SMART/SciLitLLM/sft/helper/infer_sft.py
code served (permissive licence) · get_code("6ced88d74c21da6b")
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

# Construct a data_item with the required structure
data_item = {
    'conversations': [
        {
            'from': 'human',
            'value': 'What is the capital of France?'
        },
        {
            'from': 'gpt',
            'value': 'The capital of France is Paris.'
        }
    ]
}

ARGS = {
    'data_item': data_item
}
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_json_score Ran
ran; our contract draft was wrong, not the code
dptech-corp/Uni-SMART/SciLitLLM/sft/helper/infer_sft.py
code served (permissive licence) · get_code("c8e533eb7a6418fc")
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 re

def parse_json_score(text):
    pattern = '\\{[^{}]*\\}'
    match = re.search(pattern, text)
    try:
        assert match
        content = match.group()
        result = json.loads(content)
        if isinstance(result, dict):
            return result
        else:
            return None
    except (SyntaxError, ValueError):
        return None


ARGS = {
    "text": 'The score is {"accuracy": 0.95, "loss": 0.05} and that is good'
}
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
parse_score Ran
ran, honoured the contract we drafted
dptech-corp/Uni-SMART/SciLitLLM/cpt/quality_control/llama_infer.py
code served (permissive licence) · get_code("978a0952e0e705b2")
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 int, ndim 0, dtype family int, shape () as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import re

def parse_score(text):
    pattern = re.compile('Education(?:al)? score:\\s*(\\d+(?:\\.\\d+)?)(?:\\/\\d+)?')
    match = pattern.search(text)
    if match:
        score_str = match.group(1)
        if '.' in score_str:
            return int(float(score_str))
        else:
            return int(score_str)
    return -1

ARGS = {
    "text": "Educational score: 85.5/100 - This is a great result!"
}
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 int, ndim 0, dtype family int, shape () — shape () as a later run recorded it
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
process_chunk Ran
ran; our contract draft was wrong, not the code
dptech-corp/Uni-SMART/SciLitLLM/cpt/format_grammar_correction.py
code served (permissive licence) · get_code("3b6071d2b3a5c4ef")
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

# Sample text chunk to process
sample_chunk = """The quick brown fox jumps over the lazy dog.
This is a test   of  the  formatting  issues.
There are unnecessary    line breaks
and    extra     spaces throughout."""

model_name = "gpt-3.5-turbo"

ARGS = {
    'chunk': sample_chunk,
    'model': model_name
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, openai, 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
process_text_entry Ran
ran; our contract draft was wrong, not the code
dptech-corp/Uni-SMART/SciLitLLM/cpt/format_grammar_correction.py
code served (permissive licence) · get_code("037abb9ab6603b81")
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 concurrent.futures
from openai import OpenAI, AzureOpenAI

# Mock the OpenAI client to avoid actual API calls
class MockCompletion:
    def __init__(self, content):
        self.choices = [MockChoice(content)]

class MockChoice:
    def __init__(self, content):
        self.message = MockMessage(content)

class MockMessage:
    def __init__(self, content):
        self.content = content

class MockOpenAIClient:
    def __init__(self, api_key=None, base_url=None):
        self.api_key = api_key
        self.base_url = base_url
    
    class ChatCompletions:
        @staticmethod
        def create(model, messages):
            return MockCompletion(f"Here is the corrected version of the text: Corrected text from {model}")
    
    @property
    def chat(self):
        class Chat:
            completions = self.ChatCompletions()
        return Chat()

# Monkey patch the client in the module context
client = MockOpenAIClient(api_key='test_key', base_url='http://localhost:8000/v1')

template = 'I have extracted the following raw text from a PDF, but the extraction process has introduced many formatting issues such as unnecessary line breaks, extra spaces, and other artifacts that disrupt the text flow. Could you please help me correct these formatting issues and provide a clean, readable version of the text? Respond with the Corrected Version only.\n\nRaw Text:\n\n{RawText}\n\nStart your response with "Here is the corrected version of the text:".\n'

def chunk_text(text, chunk_size):
    """Splits the text into chunks based on the chunk size (number of words)."""
    words = text.split()
    chunks = [' '.join(words[i:i + chunk_size]) for i in range(0, len(words), chunk_size)]
    return chunks

def process_chunk(chunk, model):
    """Processes a single chunk of text using LLM."""
    messages = [{'role': 'user', 'content': template.replace('{RawText}', chunk)}]
    try:
        completion = client.chat.completions.create(model=model, messages=messages)
        corrected_chunk = completion.choices[0].message.content.strip().replace('Here is the corrected version of the text:', '').strip()
        return corrected_chunk
    except Exception as e:
        print(f'Error processing text chunk: {e}')
        return chunk

def process_text_entry(text, model, chunk_size, max_workers):
    """Processes a single text entry by splitting it into character-based chunks and using LLM to correct it."""
    corrected_text = []
    chunks = chunk_text(text, chunk_size)
    with concurrent.futures.ThreadPoolExecutor(max_workers=max_workers) as executor:
        future_to_chunk = {executor.submit(process_chunk, chunk, model): chunk for chunk in chunks}
        for future in concurrent.futures.as_completed(future_to_chunk):
            try:
                corrected_chunk = future.result()
                corrected_text.append(corrected_chunk)
            except Exception as e:
                print(f'Error processing chunk: {e}')
                corrected_text.append(future_to_chunk[future])
    return ' '.join(corrected_text)

# Setup arguments
ARGS = {
    'text': 'This is a sample text with some formatting issues like extra spaces and line breaks that need to be corrected by the LLM model for better readability and presentation.',
    'model': 'gpt-3.5-turbo',
    'chunk_size': 10,
    'max_workers': 2
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
openai — 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
read_json Ran
ran; our contract draft was wrong, not the code
dptech-corp/Uni-SMART/SciLitLLM/sft/helper/dedup.py
code served (permissive licence) · get_code("7daf38d1d71fe676")
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 json
import tempfile
import os

# Create a temporary JSON file with sample data
temp_dir = tempfile.mkdtemp()
temp_file = os.path.join(temp_dir, 'test_data.json')

# Write sample JSON data to the temporary file
sample_data = [
    {'name': 'Alice', 'age': 30},
    {'name': 'Bob', 'age': 25},
    {'name': 'Charlie', 'age': 35}
]

with open(temp_file, 'w') as f:
    json.dump(sample_data, f)

# Define ARGS with the filename parameter
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 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
read_json Ran
ran; our contract draft was wrong, not the code
dptech-corp/Uni-SMART/SciLitLLM/sft/helper/infer_sft.py
code served (permissive licence) · get_code("953b99c2358aec7b")
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 json
import tempfile
import os

# Create a temporary JSON file for the function to read
temp_dir = tempfile.mkdtemp()
temp_file = os.path.join(temp_dir, 'test_data.json')

# Create sample data and write to JSON file
sample_data = [
    {"id": 1, "name": "item1", "value": 10},
    {"id": 2, "name": "item2", "value": 20},
    {"id": 3, "name": "item3", "value": 30}
]

with open(temp_file, 'w') as f:
    json.dump(sample_data, f)

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 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
read_jsonl Ran
ran; our contract draft was wrong, not the code
dptech-corp/Uni-SMART/SciLitLLM/cpt/quality_control/llama_infer.py
code served (permissive licence) · get_code("793cd3b000794433")
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 json
import tempfile
import os

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

# Write sample JSONL data
with open(temp_file, 'w') as f:
    f.write(json.dumps({"id": 1, "text": "hello"}) + "\n")
    f.write(json.dumps({"id": 2, "text": "world"}) + "\n")
    f.write(json.dumps({"id": 3, "text": "test"}) + "\n")

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 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
run_batch Ran
ran; our contract draft was wrong, not the code
dptech-corp/uni-smart/SciLitLLM/sft/gen_synthetic_data.py
code served (permissive licence) · get_code("e73cf9602ff9d98e")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1) and returned coroutine. 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 asyncio
from unittest.mock import AsyncMock, MagicMock

# Create a mock batch object that implements the required interface
class MockBatch:
    def __init__(self):
        self.items = []
    
    async def add(self, endpoint, **kwargs):
        """Mock add method that stores the request"""
        self.items.append({
            'endpoint': endpoint,
            'kwargs': kwargs
        })
    
    async def run(self):
        """Mock run method that returns the batch results"""
        return {
            'status': 'completed',
            'items_processed': len(self.items)
        }

# Create synthetic data
synthetic_data = [
    {
        'system': 'You are a helpful assistant.',
        'user': 'What is 2+2?'
    },
    {
        'system': 'You are a math tutor.',
        'user': 'Explain calculus.'
    }
]

# Construct the ARGS dictionary
ARGS = {
    'batch': MockBatch(),
    'synthetic_data': synthetic_data,
    'model': 'gpt-35-turbo',
    'tpm': 10000
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1
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 coroutine — a coroutine output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a coroutine 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
top_percent_indices Ran
ran; our contract draft was wrong, not the code
dptech-corp/Uni-SMART/SciLitLLM/sft/helper/dedup.py
code served (permissive licence) · get_code("63649262371c079a")
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 numpy as np

# Create a small 2D numpy array to pass as argument A
# Using a 4x3 array to have distinct dimensions
A = np.array([
    [0.1, 0.5, 0.9],
    [0.2, 0.6, 0.8],
    [0.3, 0.7, 0.4],
    [0.15, 0.25, 0.35]
], dtype=np.float32)

# Define the ratio parameter
ratio = 0.05

# Construct ARGS dictionary with all parameters
ARGS = {
    "A": A,
    "ratio": ratio
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type 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

Also present in other papers' code

Of 13 samples lifted from this paper's linked repositories, 13 carry a function identity and 0 do not. Of the 13 distinct functions with an identity, 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 13 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

Scientific literature understanding is crucial for extracting targeted information and garnering insights, thereby significantly advancing scientific discovery. Despite the remarkable success of Large Language Models (LLMs), they face challenges in scientific literature understanding, primarily due to (1) a lack of scientific knowledge and (2) unfamiliarity with specialized scientific tasks. To develop an LLM specialized in scientific literature understanding, we propose a hybrid strategy that integrates continual pre-training (CPT) and supervised fine-tuning (SFT), to simultaneously infuse scientific domain knowledge and enhance instructionfollowing capabilities for domain-specific tasks. In this process, we identify two key challenges: (1) constructing high-quality CPT corpora, and (2) generating diverse SFT instructions. We address these challenges through a meticulous pipeline, including PDF text extraction, parsing content error correction, quality filtering, and synthetic instruction creation. Applying this strategy, we present a suite of LLMs: SciLitLLM, specialized in scientific literature understanding. These models demonstrate promising performance on scientific literature understanding benchmarks. Our contributions are threefold: (1) We present an effective framework that integrates CPT and SFT to adapt LLMs to scientific literature understanding, which can also be easily adapted to other domains. (2) We propose an LLM-based synthesis method to generate diverse and high-quality scientific instructions, resulting in a new instruction set -SciLitIns -for less-represented scientific domains. (3) SciLitLLM achieves promising performance in scientific literature understanding benchmarks. We release the data processing codes 1 and model weights 2 .

For agents

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

get_harvested_code_for_paper("2408.15545")
get_code_for_paper("2408.15545")
have("2408.15545")

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

curl https://syntology.ai/api/ran/2408.15545.json

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

[![Syntology run record](https://syntology.ai/api/ran/2408.15545.svg)](https://syntology.ai/paper/2408.15545)

Connect an agent — have() is free.