SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2605.28093 · 2026

ConRAG: Consensus-Driven Multi-View Retrieval for Multi-Hop Question Answering

Bo Du, Qihuang Zhong, Juhua Liu, Kunfeng Chen, Yikai Zhu

arXiv · PDF · Open in the Atlas

Code that ran

4 of 5 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; 4 failures of our instrument; 1 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/2605.28093.json · badge: /api/ran/2605.28093.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
yikai-zhu/ConRAG canonical (classifier label)
classifier basis: census
4 ran · 1 no recorded run
FunctionStatusWhere it lives
clean_llm_text Ran
ran; our contract draft was wrong, not the code
yikai-zhu/ConRAG/src/conrag/clients.py
code served (permissive licence) · get_code("4ead1fcd6ff93dbf")
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

_ZERO_WIDTH_RE = re.compile('[\\u200B-\\u200D\\uFEFF]')

_FENCE_RE = re.compile('```(?:\\s*\\w+)?\\s*\\n(?P<body>[\\s\\S]*?)\\n\\s*```', re.MULTILINE)

def clean_llm_text(value: object) -> str:
    if not isinstance(value, str):
        return ''
    text = _ZERO_WIDTH_RE.sub('', value.replace('\r\n', '\n').replace('\r', '\n').strip())
    if (match := _FENCE_RE.search(text)):
        text = match.group('body').strip()
    elif text.startswith('```') and text.endswith('```') and (len(text) >= 6):
        text = text[3:-3].strip()
    if text.lower().startswith('json\n'):
        text = text.split('\n', 1)[1].strip()
    return text


# Test case 1: string with markdown fence
test_value_1 = """```python
def hello():
    print("world")
"""

# Test case 2: string with json header
test_value_2 = """json
{"key": "value"}"""

# Test case 3: simple string with zero-width characters
test_value_3 = "hello\u200Bworld"

# Test case 4: string with carriage returns
test_value_4 = "line1\r\nline2\rline3"

# Test case 5: non-string object
test_value_5 = 42

ARGS = {
    "value": test_value_1
}
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
format_memory Ran
ran; our contract draft was wrong, not the code
yikai-zhu/ConRAG/src/conrag/retrieval.py
code served (permissive licence) · get_code("757f85c8282798a3")
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

torch.manual_seed(0)

# Analyze the function:
# - format_memory takes a tuple of strings
# - It iterates over items in the tuple
# - It filters out falsy items (empty strings, None, etc.)
# - It formats each item as "- {item}"
# - It joins them with newlines

ARGS = {
    "memory": ("first memory item", "second memory item", "third memory item")
}
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
render_plan Ran
ran; our contract draft was wrong, not the code
yikai-zhu/ConRAG/src/conrag/planning.py
code served (permissive licence) · get_code("58cba261d2ca371a")
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 typing import TypedDict

NOT_FOUND = 'Information not found'

class PlanStep(TypedDict):
    id: int
    sub_question: str
    dependencies: list[int]

# Construct valid arguments for render_plan
plan = [
    {
        'id': 0,
        'sub_question': 'What is the capital of France?',
        'dependencies': []
    },
    {
        'id': 1,
        'sub_question': 'Is <dep:0> in Europe?',
        'dependencies': [0]
    },
    {
        'id': 2,
        'sub_question': 'What is the population of <dep:0>?',
        'dependencies': [0]
    }
]

answers = {
    0: 'Paris',
    1: 'Yes, it is in Europe',
}

ARGS = {
    'plan': plan,
    'answers': answers
}
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
topological_plan Ran
ran; our contract draft was wrong, not the code
yikai-zhu/ConRAG/src/conrag/planning.py
code served (permissive licence) · get_code("2b5aaafaa2399cad")
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.

from collections import defaultdict, deque
from typing import TypedDict

class PlanStep(TypedDict):
    id: int
    sub_question: str
    dependencies: list[int]

# Construct valid arguments for topological_plan
steps = {
    0: {
        'id': 0,
        'sub_question': 'What is the first step?',
        'dependencies': []
    },
    1: {
        'id': 1,
        'sub_question': 'What is the second step?',
        'dependencies': [0]
    },
    2: {
        'id': 2,
        'sub_question': 'What is the third step?',
        'dependencies': [0]
    },
    3: {
        'id': 3,
        'sub_question': 'What is the fourth step?',
        'dependencies': [1, 2]
    }
}

ARGS = {
    'steps': steps
}
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
clean_text Not yet run
unverified (harvested, no recorded run)
yikai-zhu/ConRAG/src/conrag/common.py
code served (permissive licence) · get_code("e0d187ce27d64511")
How it ran

no invocation recorded.

Also present in other papers' code

Of 5 samples lifted from this paper's linked repositories, 5 carry a function identity and 0 do not. Of the 5 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 5 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

Retrieval-augmented generation (RAG) has emerged as a promising paradigm for enhancing large language models (LLMs) on multihop question answering (QA), which requires reasoning over evidence from multiple documents. Current multi-hop RAG methods generally focus on either query-side task decomposition or corpus-side knowledge graph construction. Despite their progress, these methods still struggle to achieve satisfactory performance on complex multi-hop QA tasks. To this end, we propose ConRAG, a consensusdriven multi-view RAG framework that effectively boosts LLMs on complex multi-hop QA. The core of ConRAG is to systematically optimize both the query and corpus sides and to leverage multi-view evidence (relation, entity, and text signals) for more accurate retrieval. Extensive experiments on three multi-hop QA benchmarks show that ConRAG consistently outperforms all baselines by a clear margin, e.g., up to +26.9% average performance gains over vanilla RAG, and enables Gemma-4-31B to achieve a new state-of-the-art record on the challenging MuSiQue benchmark. * Equal contribution. † Corresponding author. b) Graph-based RAG a) Reasoning-based RAG Retrieval Subgraph Corpus NER

For agents

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

get_harvested_code_for_paper("2605.28093")
get_code_for_paper("2605.28093")
have("2605.28093")

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

curl https://syntology.ai/api/ran/2605.28093.json

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

[![Syntology run record](https://syntology.ai/api/ran/2605.28093.svg)](https://syntology.ai/paper/2605.28093)

Connect an agent — have() is free.