SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2606.14580 · 2026

Persuasion Index: A Theory-Guided Framework for Persuasion Analysis

Zhiyang Wang, Julia Mendelsohn, Liancheng Gong, Yiwei Xu

arXiv · PDF

Code that ran

15 of 20 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, 2 violated it and 0 have no contract verdict; 12 failures of our instrument; 5 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/2606.14580.json · badge: /api/ran/2606.14580.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
krystalgong/Persuasion_Index_Code canonical (classifier label)
classifier basis: census
15 ran · 5 no recorded run
FunctionStatusWhere it lives
calculate_dual_weighted_score Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
krystalgong/Persuasion_Index_Code/persuasion_profile.py
code served (permissive licence) · get_code("95645a772a8337e2")
How it ran

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

import numpy as np
import pandas as pd

# Construct sub_ci_df with coefficient data for sub-features
sub_ci_df = pd.DataFrame({
    'coef': [
        0.5,      # intercept
        0.3,      # Sentiment.anger
        0.2,      # Sentiment.joy
        -0.1,     # Emotion.fear
        0.15,     # Emotion.surprise
    ]
}, index=['intercept', 'Sentiment.anger', 'Sentiment.joy', 'Emotion.fear', 'Emotion.surprise'])

# Construct mean_ci_df with coefficient data for category means
mean_ci_df = pd.DataFrame({
    'coef': [
        0.4,      # intercept
        0.25,     # Sentiment.mean
        0.18,     # Emotion.mean
    ]
}, index=['intercept', 'Sentiment.mean', 'Emotion.mean'])

# Construct raw_scores with nested structure
raw_scores = {
    'Sentiment': {
        'anger': 0.8,
        'joy': 0.6,
        'mean': 0.7,
    },
    'Emotion': {
        'fear': 0.4,
        'surprise': 0.9,
        'mean': 0.65,
    }
}

ARGS = {
    'raw_scores': raw_scores,
    'sub_ci_df': sub_ci_df,
    'mean_ci_df': mean_ci_df,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, pandas — 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
cat Ran
ran; our contract draft was wrong, not the code
krystalgong/Persuasion_Index_Code/lexicons/LLM_expansion/build_theory_definitions.py
code served (permissive licence) · get_code("098e97866be4b322")
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

torch.manual_seed(0)

# Construct arguments for the cat function
appeal = "visual_appeal"
dim = 5
sub = "color_distribution"
function = lambda x: x * 2
anchor = torch.tensor([[1.0, 2.0, 3.0, 4.0]])
include = ["feature_a", "feature_b"]
exclude = ["feature_c"]
traps = {
    "trap_1": torch.tensor([[0.1, 0.2, 0.3, 0.4]]),
    "trap_2": torch.tensor([[0.5, 0.6, 0.7, 0.8]])
}
seeds = [42, 123, 456]

ARGS = {
    'appeal': appeal,
    'dim': dim,
    'sub': sub,
    'function': function,
    'anchor': anchor,
    'include': include,
    'exclude': exclude,
    'traps': traps,
    'seeds': seeds
}
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
evidence_named_entities Ran
ran, honoured the contract we drafted
krystalgong/Persuasion_Index_Code/PI_score_generator.py
code served (permissive licence) · get_code("3153e1c9920965e2")
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 math
from typing import Any, Dict, Iterator, Set

# Create a minimal spaCy-like NLP mock object
class MockNLP:
    def __call__(self, text: str):
        return MockDoc(text)

class MockDoc:
    def __init__(self, text: str):
        self.text = text
        self.ents = [
            MockEntity("Alice", "PERSON"),
            MockEntity("OpenAI", "ORG"),
            MockEntity("2024-01-15", "DATE"),
            MockEntity("100", "CARDINAL"),
        ]
    
    def __len__(self):
        return len(self.text.split())

class MockEntity:
    def __init__(self, text: str, label: str):
        self.text = text
        self.label_ = label

# Create a sample text
sample_text = "Alice works at OpenAI in New York on 2024-01-15."

# Create mock NLP
mock_nlp = MockNLP()

# Define ARGS dict with all parameters for evidence_named_entities
ARGS = {
    "text": sample_text,
    "nlp": mock_nlp,
}
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 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
load_env_file Ran
ran; our contract draft was wrong, not the code
krystalgong/Persuasion_Index_Code/pi_config.py
code served (permissive licence) · get_code("8cedabb4e608d1af")
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 PosixPath. 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 os
import tempfile
from pathlib import Path

# Create a temporary .env file with test content
temp_dir = tempfile.mkdtemp()
temp_env_path = Path(temp_dir) / '.env'

# Write test content to the temporary .env file
env_content = """# This is a comment
API_KEY=secret123
DB_HOST=localhost
DB_PORT='5432'
export LOG_LEVEL="DEBUG"
EMPTY_VAR=
QUOTED_VALUE='quoted_string'
"""

temp_env_path.write_text(env_content, encoding='utf-8')

# Clear any existing test environment variables
for key in ['API_KEY', 'DB_HOST', 'DB_PORT', 'LOG_LEVEL', 'EMPTY_VAR', 'QUOTED_VALUE']:
    os.environ.pop(key, None)

ARGS = {
    'path': str(temp_env_path),
    'override': False
}
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 PosixPath — a PosixPath output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a PosixPath 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_records Ran
ran; our contract draft was wrong, not the code
krystalgong/Persuasion_Index_Code/lexicons/LLM_expansion/postprocess_generation.py
code served (permissive licence) · get_code("671ab98d80ff3d16")
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 3, 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
from pathlib import Path
from typing import Dict, List, Tuple
import pandas as pd
import tempfile
import os

# Create a temporary file with test data
temp_dir = tempfile.mkdtemp()
temp_file = Path(temp_dir) / "test_records.jsonl"

# Write test data to the temporary file
test_data = [
    json.dumps({"id": 1, "name": "item1", "value": 100}),
    json.dumps({"id": 2, "name": "item2", "value": 200}),
    json.dumps({"_meta_status": "active", "record_id": 1, "timestamp": "2024-01-01"}),
    json.dumps({"_meta_status": "inactive", "record_id": 2, "timestamp": "2024-01-02"}),
    "",  # empty line to test stripping
]

with temp_file.open("w") as f:
    f.write("\n".join(test_data))

ARGS = {
    "path": temp_file,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
pandas — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type tuple, ndim 3, 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
load_removed_items Ran
ran; our contract draft was wrong, not the code
krystalgong/Persuasion_Index_Code/lexicons/LLM_expansion/apply_audit.py
code served (permissive licence) · get_code("408158c620962b9e")
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 set. 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 pandas as pd
import tempfile
import os

# Create a temporary CSV file with the required structure
temp_dir = tempfile.mkdtemp()
csv_path = Path(temp_dir) / "audit.csv"

# Create a sample audit CSV with the expected columns
audit_data = {
    'category': ['fruits', 'vegetables', 'fruits', 'dairy'],
    'word_surface': ['Apple', 'Carrot', 'BANANA', 'Milk'],
    'audit_decision': [1, 2, 2, 1]
}

df = pd.DataFrame(audit_data)
df.to_csv(csv_path, index=False)

ARGS = {
    'audit_csv': csv_path
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
pandas — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type set — a set output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a set 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
merge_items Ran
ran; our contract draft was wrong, not the code
krystalgong/Persuasion_Index_Code/lexicons/LLM_expansion/postprocess_generation.py
code served (permissive licence) · get_code("220adb8aa66f3c29")
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 DataFrame. 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
import pandas as pd
import numpy as np

# Create a sample DataFrame with the required columns
items_df = pd.DataFrame({
    'word': ['Hello', 'hello', 'World', 'world  ', 'Test?', 'test?'],
    'category': ['noun', 'noun', 'noun', 'noun', 'verb', 'verb'],
    'slice': [1, 1, 2, 2, 3, 3],
    'register': ['formal', 'formal', 'informal', 'informal', 'formal', 'formal'],
    'type': ['common', 'common', 'common', 'common', 'rare', 'rare'],
    'confidence': ['high', 'medium', 'high', 'high', 'low', 'medium'],
    'rationale': ['reason1', 'reason2', 'reason3', 'reason4', 'reason5', 'reason6'],
    'is_hard_negative': [False, False, True, True, False, False]
})

ARGS = {
    'items': items_df
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, pandas — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type DataFrame — a DataFrame output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a DataFrame output the recorded ndim 2 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
missing_required_categories Ran
ran; our contract draft was wrong, not the code
krystalgong/Persuasion_Index_Code/persuasion_index/liwc.py
code served (permissive licence) · get_code("bd74616789d876c3")
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

ARGS = {
    'dictionary': {
        'I': ['word1', 'word2'],
        'You': [],
        'Anger': ['angry', 'mad'],
        'Sad': ['sad', 'blue'],
        'Anx': [],
        'Posemo': ['happy', 'good'],
        'Past': ['was', 'went'],
        'See': ['see', 'look'],
        'Ppron': ['he', 'she']
    }
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type list, ndim 1, dtype family float — a list output has no array shape of its own; none is stated
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
model_supports_temperature Ran
ran, violated the contract we drafted
krystalgong/Persuasion_Index_Code/lexicons/LLM_expansion/generator_runner.py
code served (permissive licence) · get_code("21080e01b4b8bb04")
How it ran

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

import re

_NEW_TOKEN_PARAM_PATTERNS = (re.compile('^gpt-5'), re.compile('^o1'), re.compile('^o3'), re.compile('^o4'))

def model_supports_temperature(model: str) -> bool:
    """GPT-5 reasoning models reject temperature != 1.0 / sometimes the field
    altogether. Detect and skip the parameter."""
    return not any((p.match(model) for p in _NEW_TOKEN_PARAM_PATTERNS))


ARGS = {
    "model": "gpt-4-turbo"
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
none outside the standard library — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type bool, ndim 0, dtype family float, shape () — shape () as a later run recorded it
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
model_uses_new_token_param Ran
ran, violated the contract we drafted
krystalgong/Persuasion_Index_Code/lexicons/LLM_expansion/generator_runner.py
code served (permissive licence) · get_code("3ffc5e44fe9290ed")
How it ran

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

import re

_NEW_TOKEN_PARAM_PATTERNS = (re.compile('^gpt-5'), re.compile('^o1'), re.compile('^o3'), re.compile('^o4'))

def model_uses_new_token_param(model: str) -> bool:
    return any((p.match(model) for p in _NEW_TOKEN_PARAM_PATTERNS))

ARGS = {
    "model": "gpt-5-turbo"
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
none outside the standard library — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type bool, ndim 0, dtype family float, shape () — shape () as a later run recorded it
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
normalise_word Ran
ran; our contract draft was wrong, not the code
krystalgong/Persuasion_Index_Code/lexicons/LLM_expansion/postprocess_generation.py
code served (permissive licence) · get_code("6c0bde5062a24dd3")
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 normalise_word(w: str) -> str:
    """Conservative normalisation for dedup: lowercase + collapse whitespace.
    We do NOT strip punctuation because some items are punctuation-bearing
    (e.g. '%', 'right?', "i.e.")."""
    return re.sub('\\s+', ' ', w.strip().lower())

ARGS = {
    'w': '  Hello   World!  '
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
none outside the standard library — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type 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
parse_liwc_file Ran
ran; our contract draft was wrong, not the code
krystalgong/Persuasion_Index_Code/persuasion_index/liwc.py
code served (permissive licence) · get_code("59d6c90243b5d054")
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 tempfile
from pathlib import Path

# Create a temporary LIWC format file for testing
temp_file = tempfile.NamedTemporaryFile(mode='w', suffix='.dic', delete=False, encoding='utf-8')

# Write a simple LIWC dictionary in the compact "Category: terms" format
liwc_content = """# Sample LIWC dictionary
Anger: angry mad furious
Sad: sad depressed gloomy
Posemo: happy joyful glad
You: you your yours
I: i me my
"""

temp_file.write(liwc_content)
temp_file.close()

ARGS = {
    'path': Path(temp_file.name)
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
none outside the standard library — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type 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
slice_key Ran
ran; our contract draft was wrong, not the code
krystalgong/Persuasion_Index_Code/lexicons/LLM_expansion/generator_runner.py
code served (permissive licence) · get_code("76ca2d0aa8afc0e6")
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)

# The function slice_key takes two string parameters and returns their
# concatenation with a tab separator. No complex objects needed.

ARGS = {
    'category': 'training',
    'slice_name': 'batch_001'
}
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
strip_punctuation Ran
ran; our contract draft was wrong, not the code
krystalgong/Persuasion_Index_Code/lexicons/LLM_expansion/apply_audit.py
code served (permissive licence) · get_code("f7ea5c47b4936ba7")
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
import string

def strip_punctuation(w: str) -> str:
    """Strip leading/trailing punctuation, then normalise internal whitespace.
    Leaves apostrophes inside words intact (e.g. "it's") and hyphens inside
    compound words (e.g. "well-known"), since those are part of the token."""
    stripped = w.strip(string.punctuation + ' ')
    return re.sub('\\s+', ' ', stripped)


# The function takes a single parameter: w (a string)
# Let's create test cases for various scenarios

ARGS = {
    "w": "...hello world..."
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
none outside the standard library — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type 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
trap Ran
ran; our contract draft was wrong, not the code
krystalgong/Persuasion_Index_Code/lexicons/LLM_expansion/build_theory_definitions.py
code served (permissive licence) · get_code("02b2e48b8f4bbb79")
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

torch.manual_seed(0)

# The function simply takes three parameters and returns a dict
# 'word' appears to be a string or identifier
# 'drift' appears to be a drift type (string)
# 'why' appears to be a reason or explanation (string)

word = "example_word"
drift = "drift_type_a"
why = "this is the reason"

ARGS = {
    "word": word,
    "drift": drift,
    "why": why
}
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
build_messages Not yet run
unverified (harvested, no recorded run)
krystalgong/Persuasion_Index_Code/lexicons/LLM_expansion/generation_prompts.py
code served (permissive licence) · get_code("c18f2c089df0bf5a")
How it ran

no invocation recorded.

build_user_message Not yet run
unverified (harvested, no recorded run)
krystalgong/Persuasion_Index_Code/lexicons/LLM_expansion/generation_prompts.py
code served (permissive licence) · get_code("4b7285a0079d3163")
How it ran

no invocation recorded.

evidence_attribution Not yet run
unverified (harvested, no recorded run)
krystalgong/Persuasion_Index_Code/PI_score_generator.py
code served (permissive licence) · get_code("c5a33d204e15cbc2")
How it ran

no invocation recorded.

evidence_statistical Not yet run
unverified (harvested, no recorded run)
krystalgong/Persuasion_Index_Code/PI_score_generator.py
code served (permissive licence) · get_code("b5d248b85a69b87f")
How it ran

no invocation recorded.

is_hard_negative_slice Not yet run
unverified (harvested, no recorded run)
krystalgong/Persuasion_Index_Code/lexicons/LLM_expansion/generation_prompts.py
code served (permissive licence) · get_code("7427a706014f2dea")
How it ran

no invocation recorded.

Also present in other papers' code

Of 20 samples lifted from this paper's linked repositories, 20 carry a function identity and 0 do not. Of the 20 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 20 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

Identifying persuasive rhetorical cues is critical across domains, from detecting information manipulation and improving AI safety to advancing public health communication. We propose Persuasion Index (PI), a taxonomy of 15 dimensions grounded in persuasion theories from psychology and communication, and one transparent implementation using 55 subfeatures built from lexicons and rule-based detectors. The taxonomy is modular: individual detectors can be replaced while preserving the theoretical structure. By evaluating PI on four public datasets varying in domain, style, and outcome measures, we show that PI provides a shared feature space for interpreting rhetorical patterns associated with persuasionrelated outcomes. Linear models show that PI features carry meaningful predictive signal while remaining computationally lightweight. Dimension-level analyses reveal recurring associations between PI dimensions and persuasion outcomes across datasets, while also highlighting topic-and stance-specific variation. We release PI as an open-source package and web interface for principled and auditable analysis of human and AI-mediated communication. 1

For agents

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

get_harvested_code_for_paper("2606.14580")
get_code_for_paper("2606.14580")
have("2606.14580")

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

curl https://syntology.ai/api/ran/2606.14580.json

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

[![Syntology run record](https://syntology.ai/api/ran/2606.14580.svg)](https://syntology.ai/paper/2606.14580)

Connect an agent — have() is free.