SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2401.04518 · 2024

The Critique of Critique

arXiv · PDF · Open in the Atlas

Code that ran

7 of 8 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; 3 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; 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/2401.04518.json · badge: /api/ran/2401.04518.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
gair-nlp/metacritique canonical (classifier label)
classifier basis: census
7 ran · 1 no recorded run
FunctionStatusWhere it lives
acc Ran
ran, honoured the contract we drafted
gair-nlp/metacritique/meta_critique/cal_meta_scores.py
code served (permissive licence) · get_code("e1ea54cd691ed056")
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 torch

torch.manual_seed(0)

# The function `acc` takes `pred` as a parameter
# It iterates through `pred`, counts truthy values, and returns the ratio
# `pred` should be an iterable of boolean-like values

# Create a list of boolean values
pred = [True, False, True, True, False]

ARGS = {
    "pred": pred
}
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 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
cls_post_process Ran
ran; our contract draft was wrong, not the code
gair-nlp/metacritique/meta_critique/utils.py
code served (permissive licence) · get_code("75a2891e73eccd85")
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

torch.manual_seed(0)

ARGS = {
    'out': 'The claim is true based on the evidence provided.'
}
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 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
f1_score Ran
ran, honoured the contract we drafted
gair-nlp/metacritique/meta_critique/cal_meta_scores.py
code served (permissive licence) · get_code("9b5da268fc9d2533")
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 torch
import numpy as np

torch.manual_seed(0)

# The function f1_score takes two parameters: p (precision) and r (recall)
# It performs arithmetic operations on them (division, addition)
# Valid inputs are numeric values (int or float)

# Create precision and recall values
p = 0.8
r = 0.75

ARGS = {
    "p": p,
    "r": r,
}
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 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
merge_outcomes Ran
ran; our contract draft was wrong, not the code
gair-nlp/metacritique/meta_critique/merge_files.py
code served (permissive licence) · get_code("3664cfc54d424136")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned list, ndim 1, dtype family float; no output shape is stated. The image it ran in was not recorded. If your copy returns the same type and ndim on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
import numpy as np

torch.manual_seed(0)

# Construct all_data - a list of dicts that will be modified
all_data = [
    {
        'question': 'What is 2+2?',
        'id': 0
    },
    {
        'question': 'What is 3+3?',
        'id': 1
    }
]

# Construct ref_answer - list of dicts with 'output' key
ref_answer = [
    {'output': 'The answer is 4'},
    {'output': 'The answer is 6'}
]

# Construct ref_critique - list of dicts with 'output' key
ref_critique = [
    {'output': 'This is a correct answer'},
    {'output': 'This is also correct'}
]

# Construct ref_aius - list of dicts with 'output' key containing newline-separated text
ref_aius = [
    {'output': 'point1\npoint2\npoint3'},
    {'output': 'item1\nitem2'}
]

# Construct hyp_critique - list of dicts with 'output' key
hyp_critique = [
    {'output': 'Hypothesis critique 1'},
    {'output': 'Hypothesis critique 2'}
]

# Construct hyp_aius - list of dicts with 'output' key containing newline-separated text
hyp_aius = [
    {'output': 'hyp_item1\nhyp_item2'},
    {'output': 'hyp_a\nhyp_b\nhyp_c'}
]

ARGS = {
    'all_data': all_data,
    'ref_answer': ref_answer,
    'ref_critique': ref_critique,
    'ref_aius': ref_aius,
    'hyp_critique': hyp_critique,
    'hyp_aius': hyp_aius
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type list, ndim 1, dtype family float — a list output has no array shape of its own; none is stated
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
meta_score Ran
ran, honoured the contract we drafted
gair-nlp/metacritique/meta_critique/cal_meta_scores.py
code served (permissive licence) · get_code("d77ead10c19bb201")
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 all_data parameter
# Each item needs 'hypothesis_critique' and 'gpt4_critique' with 'aius' lists
all_data = [
    {
        'hypothesis_critique': {'aius': [1, 2, 3]},
        'gpt4_critique': {'aius': [1, 2]}
    },
    {
        'hypothesis_critique': {'aius': [1, 2]},
        'gpt4_critique': {'aius': [1, 2, 3, 4]}
    }
]

# Construct precision parameter
# List of dicts with 'verifying_result' key (boolean-like values)
precision = [
    {'verifying_result': True},
    {'verifying_result': True},
    {'verifying_result': False},
    {'verifying_result': True},
    {'verifying_result': False}
]

# Construct recall parameter
# List of dicts with 'verifying_result' key
recall = [
    {'verifying_result': True},
    {'verifying_result': True},
    {'verifying_result': True},
    {'verifying_result': False},
    {'verifying_result': True},
    {'verifying_result': False}
]

ARGS = {
    'all_data': all_data,
    'precision': precision,
    'recall': recall
}
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
read_json Ran
ran; our contract draft was wrong, not the code
gair-nlp/metacritique/meta_critique/utils.py
code served (permissive licence) · get_code("851a9d51cc072ba2")
How it ran

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

import json
import tempfile
import os

# Create a temporary JSON file with sample data
temp_dir = tempfile.mkdtemp()
json_file_path = os.path.join(temp_dir, "test.json")

sample_data = {
    "key1": "value1",
    "key2": [1, 2, 3],
    "key3": {"nested": "data"}
}

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

ARGS = {
    "json_file_path": json_file_path
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
none outside the standard library — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type dict — a dict output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a dict output the recorded ndim 0 and dtype family float describe that coercion, not the output, and are not stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
text_post_process Ran
ran; our contract draft was wrong, not the code
gair-nlp/metacritique/meta_critique/utils.py
code served (permissive licence) · get_code("6d1a493f0cd99513")
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

torch.manual_seed(0)

# The function text_post_process takes a single parameter 'out'
# It expects 'out' to be either a string or something that is not a string
# If it's a string, it processes it; otherwise returns None

# Test case 1: valid string that is not 'failed!'
test_string = "Hello World"

ARGS = {
    'out': test_string
}
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 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
add_args Not yet run
unverified (harvested, no recorded run)
gair-nlp/metacritique/meta_critique/meta_critique.py
code served (permissive licence) · get_code("96f07daca82c6104")
How it ran

no invocation recorded.

Also present in other papers' code

Of 8 samples lifted from this paper's linked repositories, 8 carry a function identity and 0 do not. Of the 8 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 8 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

Critique, as a natural language description for assessing the quality of model-generated content, has played a vital role in the training, evaluation, and refinement of LLMs. However, a systematic method to evaluate the quality of critique is lacking. In this paper, we pioneer the critique of critique, termed MetaCritique, which builds specific quantification criteria. To achieve a reliable evaluation outcome, we propose Atomic Information Units (AIUs), which describe the critique in a more fine-grained manner. MetaCritique aggregates each AIU's judgment for the overall score. Moreover, MetaCritique delivers a natural language rationale for the intricate reasoning within each judgment. Lastly, we construct a meta-evaluation dataset covering 4 tasks across 16 public datasets involving human-written and LLM-generated critiques. Experiments demonstrate that MetaCritique can achieve near-human performance. Our study can facilitate future research in LLM critiques based on our following observations and released resources: (1) superior critiques judged by MetaCritique can lead to better refinements, indicating that it can potentially enhance the alignment of existing LLMs; (2) the leaderboard of critique models reveals that open-source critique models commonly suffer from factuality issues; (3) relevant code and data are publicly available at https://github.com/GAIR-NLP/MetaCritique to support deeper exploration; (4) an API at PyPI with the usage documentation in Appendix C allows users to assess the critique conveniently.

For agents

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

get_harvested_code_for_paper("2401.04518")
get_code_for_paper("2401.04518")
have("2401.04518")

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

curl https://syntology.ai/api/ran/2401.04518.json

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

[![Syntology run record](https://syntology.ai/api/ran/2401.04518.svg)](https://syntology.ai/paper/2401.04518)

Connect an agent — have() is free.