SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2605.17364 · 2026

NewsLens: A Multi-Agent Framework for Adversarial News Bias Navigation

Joy Bose

arXiv · PDF · Open in the Atlas

Code that ran

5 of 7 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; 5 failures of our instrument; 2 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/2605.17364.json · badge: /api/ran/2605.17364.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
joyboseroy/newslens canonical (classifier label)
classifier basis: census
5 ran · 2 no recorded run
FunctionStatusWhere it lives
domain_of Ran
ran; our contract draft was wrong, not the code
joyboseroy/newslens/newsbiasagents/tagging/scraper.py
code served (permissive licence) · get_code("8c747dc51dc8874b")
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 urllib.parse import urlparse

def domain_of(url: str) -> str:
    try:
        return urlparse(url).netloc.replace('www.', '')
    except Exception:
        return ''

ARGS = {
    'url': 'https://www.example.com/path/to/page'
}
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_bias_label Ran
ran; our contract draft was wrong, not the code
joyboseroy/newslens/newsbiasagents/utils/gdelt_fetcher.py
code served (permissive licence) · get_code("cf8c4ee706b29b2c")
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

OUTLET_BIAS = {'cnn.com': 'left', 'huffpost.com': 'left', 'theguardian.com': 'left', 'msnbc.com': 'left', 'nytimes.com': 'left-center', 'washingtonpost.com': 'left-center', 'bbc.com': 'center', 'bbc.co.uk': 'center', 'reuters.com': 'center', 'apnews.com': 'center', 'bloomberg.com': 'center', 'usatoday.com': 'center', 'wsj.com': 'right-center', 'nypost.com': 'right', 'foxnews.com': 'right', 'breitbart.com': 'right', 'dailywire.com': 'right', 'newsmax.com': 'right', 'thehindu.com': 'left-center', 'ndtv.com': 'center', 'hindustantimes.com': 'center', 'timesofindia.com': 'center', 'republicworld.com': 'right', 'opindia.com': 'right'}

def get_bias_label(url: str) -> str:
    for domain, label in OUTLET_BIAS.items():
        if domain in url:
            return label
    return 'unknown'

ARGS = {
    'url': 'https://www.cnn.com/article/some-news'
}
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
safe_parse Ran
ran; our contract draft was wrong, not the code
joyboseroy/newslens/newsbiasagents/agents/pipeline.py
code served (permissive licence) · get_code("4499c6c2f4306882")
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 logging
from typing import Optional

log = logging.getLogger(__name__)

def safe_parse(raw: Optional[str], agent_name: str='agent') -> dict:
    """Parse JSON from agent output; strip markdown fences if present."""
    if not raw:
        return {}
    cleaned = raw.strip()
    if cleaned.startswith('```'):
        lines = cleaned.split('\n')
        cleaned = '\n'.join(lines[1:-1]) if len(lines) > 2 else cleaned
    try:
        return json.loads(cleaned)
    except json.JSONDecodeError:
        start = cleaned.find('{')
        end = cleaned.rfind('}')
        if start != -1 and end != -1 and (end > start):
            try:
                return json.loads(cleaned[start:end + 1])
            except Exception:
                pass
        log.warning(f'{agent_name}: JSON parse failed on: {cleaned[:200]}')
        return {'parse_error': True, 'raw': cleaned[:300]}


# Test case 1: Valid JSON string
test_raw_1 = '{"key": "value", "number": 42}'

# Test case 2: JSON with markdown fences
test_raw_2 = '''```json
{"action": "move", "target": "location_x"}
'''

# Test case 3: Empty/None string
test_raw_3 = None

# Test case 4: JSON with extra text
test_raw_4 = 'Some text before {"result": true} and after'

# Test case 5: Invalid JSON with agent name
test_raw_5 = 'This is not valid JSON at all'

ARGS = {
    'raw': test_raw_1,
    'agent_name': 'test_agent'
}
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
sanitize Ran
ran; our contract draft was wrong, not the code
joyboseroy/newslens/newsbiasagents/graph/falkor_writer.py
code served (permissive licence) · get_code("b6b967766845baad")
How it ran

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

import torch

def sanitize(val: str) -> str:
    """Escape single quotes for Cypher string literals."""
    return str(val).replace("'", "\\'").replace('"', '\\"')

# The function takes a single string parameter and performs escape replacements
# Test with a string containing both single and double quotes
ARGS = {
    "val": "It's a \"test\" string with quotes"
}
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
write_kg_updates Ran
ran; our contract draft was wrong, not the code
joyboseroy/newslens/newsbiasagents/graph/falkor_writer.py
code served (permissive licence) · get_code("a3458aad7c5a8767")
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 logging
from datetime import datetime

# Configure logging
logging.basicConfig(level=logging.DEBUG)
log = logging.getLogger(__name__)

# Create a minimal mock graph object that has a query method
class MockGraph:
    def query(self, q: str):
        """Mock graph query execution."""
        return None

# Define the arguments for write_kg_updates
ARGS = {
    'kg_updates': {
        'nodes': [
            {
                'id': 'node_1',
                'type': 'Person',
                'label': 'Alice'
            },
            {
                'id': 'node_2',
                'type': 'Organization',
                'label': 'Tech Corp'
            }
        ],
        'edges': [
            {
                'source': 'node_1',
                'target': 'node_2',
                'relationship': 'works for',
                'properties': {
                    'political_stance': 'neutral',
                    'credibility_score': '0.85',
                    'primary_propaganda_tactic': 'none'
                }
            }
        ]
    },
    'article_url': 'https://example.com/article',
    'graph': MockGraph(),
    'dry_run': 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 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
create_sample_data Not yet run
unverified (harvested, no recorded run)
joyboseroy/newslens/newsbiasagents/utils/sample_data.py
code served (permissive licence) · get_code("398a5f9c67022f8f")
How it ran

no invocation recorded.

get_graph Not yet run
unverified (harvested, no recorded run)
joyboseroy/newslens/newsbiasagents/graph/falkor_writer.py
code served (permissive licence) · get_code("73d692dfec720072")
How it ran

no invocation recorded.

Also present in other papers' code

Of 7 samples lifted from this paper's linked repositories, 7 carry a function identity and 0 do not. Of the 7 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 7 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

Media bias detection has predominantly been framed as a classification task: assign a political label to an article or outlet. We argue this framing is too shallow: it identifies that bias exists but not where, how, or crucially, what is structurally omitted. We present NewsLens, a five-agent adversarial pipeline for structured news bias navigation. A Fact Verifier, Progressive Framing Analyst, Conservative Framing Analyst, Propaganda Detector, and Neutral Summarizer collaborate to deconstruct articles into interpretable epistemic maps, exposing ideological omissions, rhetorical manipulation, and framing boundaries. The system is evaluated on 15 articles across four geopolitical event clusters (India-Pakistan Kashmir, Gaza, Climate Policy, Ukraine) using Qwen2.5-3B-Instruct, with crossmodel validation using Mistral 7B on the Kashmir cluster, we report Perspective Divergence Scores (PDS) and Manipulation Indices (MI). Center outlets show the highest mean PDS (Qwen: 0.907 ± 0.020, Mistral: 0.729 on Kashmir subset); conservative-framing outlets show the highest mean MI (0.600 ± 0.245 across both models). Cross-model comparison (Mistral 7B vs Qwen2.5-3B) shows high consistency for high-propaganda content (Republic World ΔPDS=0.125, MI=0.8 both models) and greater variance for nuanced reporting. Mann-Whitney U tests find no statistically significant betweengroup differences at n=15, reported honestly as a sample-size limitation confirmed by post-hoc power analysis. The architecture extends prior lexical-geometric bias work [Patankar & Bose, ICMLA 2017] to agentic LLM reasoning, and is fully reproducible using open-weight models without API keys.

For agents

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

get_harvested_code_for_paper("2605.17364")
get_code_for_paper("2605.17364")
have("2605.17364")

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

curl https://syntology.ai/api/ran/2605.17364.json

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

[![Syntology run record](https://syntology.ai/api/ran/2605.17364.svg)](https://syntology.ai/paper/2605.17364)

Connect an agent — have() is free.