SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2406.04127 · 2024

Are We Done with MMLU?

arXiv · PDF · Open in the Atlas

Code that ran

6 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; 0 honoured the contract we drafted, 0 violated it and 4 have no contract verdict; 2 failures of our instrument; 2 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/2406.04127.json · badge: /api/ran/2406.04127.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
wildeval/zeroeval community (classifier label)
classifier basis not recorded on the link
5 ran · 2 no recorded run
copy not recorded community
not in the paper's declared repositories
1 ran · 0 no recorded run
FunctionStatusWhere it lives
clear_output Ran
ran on a synthesized input
wildeval/zeroeval/src/unified_utils.py
code served (permissive licence) · get_code("27c94753060ee864")
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

torch.manual_seed(0)

# Analyze the function:
# - output: can be a list (gets indexed [0]) or a string
# - model_name: a string used for customization (not used in current logic)
# The function expects output to ultimately be a string with some tokens to replace

# Create test arguments
output_str = "Hello <|endoftext|> world <pad> test <end_of_turn> end"
model_name = "test_model"

ARGS = {
    "output": output_str,
    "model_name": model_name
}
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
create_model_path Ran
ran; our contract draft was wrong, not the code
this paper's copy was not recorded; identical code first harvested from aryopg/mmlu-redux
pointer only; the stored copy get_code serves, from aryopg/mmlu-redux, is pointer only (licence: CC-BY-4.0) · get_code("7d2a2984ad40114e")
How it ran

The body of this sample is withheld: this paper's copy of the code was not recorded (its link names no repository or source path), so no licence was read for it. The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. A run is recorded.

generate_choice_string Ran
ran on a synthesized input
wildeval/zeroeval/src/_TEMPLATES.py
code served (permissive licence) · get_code("ff5fe0e6642b312e")
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

def generate_choice_string(choices):
    choice_string = ''
    for i, choice in enumerate(choices):
        choice_string += f'- ({chr(65 + i)}) {choice}\n'
    return choice_string

ARGS = {
    'choices': ['Option 1', 'Option 2', 'Option 3', 'Option 4']
}
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
infer_maybe_lora Ran
ran on a synthesized input
wildeval/zeroeval/src/unified_infer.py
code served (permissive licence) · get_code("041b90250a61bffa")
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 tempfile
import json
import os

# Create a temporary directory with LoRA adapter structure
temp_dir = tempfile.mkdtemp()
adapter_dir = os.path.join(temp_dir, "test_lora_model")
os.makedirs(adapter_dir, exist_ok=True)

# Create a valid adapter_config.json
adapter_config = {
    "base_model_name_or_path": "gpt2",
    "lora_alpha": 16,
    "lora_dropout": 0.05,
    "r": 8,
    "target_modules": ["c_attn"],
    "task_type": "CAUSAL_LM"
}

adapter_config_path = os.path.join(adapter_dir, "adapter_config.json")
with open(adapter_config_path, 'w') as f:
    json.dump(adapter_config, f)

ARGS = {
    "model_name": adapter_dir
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
huggingface_hub — 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
result_format Ran
ran on a synthesized input
wildeval/zeroeval/src/task_configs.py
code served (permissive licence) · get_code("e7a65f4765f03bd1")
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)

class Args:
    def __init__(self, data_name):
        self.data_name = data_name

# Test case 1: alpaca_eval
output_item_1 = {
    'output': ['generated text here'],
    'other_field': 'value'
}
args_1 = Args('alpaca_eval')

# Test case 2: zebra-grid
output_item_2 = {
    'solution': 'some solution',
    'other_field': 'value'
}
args_2 = Args('zebra-grid')

# Test case 3: wildbench_v2-hard
output_item_3 = {
    'conversation_input': 'input',
    'references': ['ref1', 'ref2'],
    'length': 100,
    'checklist': ['item1'],
    'avg_score': 0.85,
    'var_score': 0.05,
    'other_field': 'value'
}
args_3 = Args('wildbench_v2-hard')

# Test case 4: other
output_item_4 = {
    'output': 'some output',
    'other_field': 'value'
}
args_4 = Args('other_dataset')

ARGS = {
    'output_item': output_item_1,
    'args': args_1
}
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
sanitize_args Ran
ran; our contract draft was wrong, not the code
wildeval/zeroeval/src/unified_infer.py
code served (permissive licence) · get_code("8526e7d6eb6a6bd3")
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 Args. 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 argparse

# Create a simple namespace object to act as args
class Args:
    def __init__(self):
        self.download_dir = 'default'

args = Args()

ARGS = {
    'args': args
}
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 Args — a Args output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a Args 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_conv_template Not yet run
unverified (harvested, no recorded run)
wildeval/zeroeval/src/fastchat_conversation.py
code served (permissive licence) · get_code("2e17447aecaa3b2e")
How it ran

no invocation recorded.

mapping_task_names Not yet run
unverified (harvested, no recorded run)
wildeval/zeroeval/src/task_configs.py
code served (permissive licence) · get_code("cfd68db505da9692")
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, 3 are also present in other papers whose code we hold (2 with at least one paper dated earlier, 2 the same month, 2 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.

FunctionAlso present inEarliest-datedOthers named
get_conv_template 69 other papers
28 dated earlier, 8 the same month, 28 later, 4 same year, order not determinable, 1 undated
1911.03705 (2019-11)
CommonGen: A Constrained Text Generation Challenge for Generative Commonsense Reasoning
2112.05682 (2021-12)
Self-attention Does Not Need $O(n^2)$ Memory
2204.07705 (2022-04)
SUPER-NATURALINSTRUCTIONS: Generalization via Declarative Instructions on 1600+ NLP Tasks
and 66 more not named here
sanitize_args 3 other papers
1 the same month, 2 later
2406.04770 (2024-06)
WildBench: Benchmarking LLMs with Challenging Tasks from Real Users in the Wild
2408.14774 (2024-08)
Instruct-SkillMix: A Powerful Pipeline for LLM Instruction Tuning
2608.06310 (2026-08)
RRC: Unlocking Generative Reward Models in LLM Reinforcement Learning via Ranking-Based Reward Construction
create_model_path 1 other paper
1 dated earlier
2009.03300 (2020-09)
Measuring Massive Multitask Language Understanding

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

Maybe not. We identify and analyse errors in the popular Massive Multitask Language Understanding (MMLU) benchmark. Even though MMLU is widely adopted, our analysis demonstrates numerous ground truth errors that obscure the true capabilities of LLMs. For example, we find that 57% of the analysed questions in the Virology subset contain errors. To address this issue, we introduce a comprehensive framework for identifying dataset errors using a novel error annotation protocol. Then, we create MMLU-Redux, which is a subset of 5,700 manually re-annotated questions across all 57 MMLU subjects. We estimate that 6.49% of MMLU questions contain errors. Using MMLU-Redux, we demonstrate significant discrepancies with the model performance metrics that were originally reported. Our results strongly advocate for revising MMLU's error-ridden questions to enhance its future utility and reliability as a benchmark. https://huggingface.co/datasets/edinburgh-dawg/mmlu-redux-2.0.

For agents

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

get_harvested_code_for_paper("2406.04127")
get_code_for_paper("2406.04127")
have("2406.04127")

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

curl https://syntology.ai/api/ran/2406.04127.json

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

[![Syntology run record](https://syntology.ai/api/ran/2406.04127.svg)](https://syntology.ai/paper/2406.04127)

Connect an agent — have() is free.