SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2605.16616 · 2026

MLReplicate: Benchmarking Autonomous Research Systems for Machine Learning Reproducibility

Sören Auer, Gollam Rabby, Sasi Kiran Gaddipati, Diyana Muhammed, Farhana Keya

arXiv · PDF · Open in the Atlas

Code that ran

3 of 6 samples (functions and classes) lifted from this paper's linked repositories executed on a synthesized input in Syntology's jail (2 of them constructed an object rather than computing a result; 0 honoured the contract we drafted, 0 violated it and 2 have no contract verdict; 1 failure of our instrument; 3 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/2605.16616.json · badge: /api/ran/2605.16616.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
gsasikiran/MLReplicate-benchmarking community
Papers with Code archive
3 ran · 3 no recorded run
FunctionStatusWhere it lives
Command Ran
ran on a synthesized input · constructed an object; no result was computed
gsasikiran/MLReplicate-benchmarking/AgentLaboratory/mlesolver.py
code served (permissive licence) · get_code("23d1b4f2b50abaf6")
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 an instance of its own class, Command: the run constructed the object and did not call it, so no result was computed. 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
from abc import abstractmethod

class Command:
    def __init__(self):
        self.cmd_type = 'OTHER'

    @abstractmethod
    def docstring(self) -> str:
        pass

    @abstractmethod
    def execute_command(self, *args) -> str:
        pass

    @abstractmethod
    def matches_command(self, cmd_str) -> bool:
        pass

    @abstractmethod
    def parse_command(self, cmd_str) -> tuple:
        pass


class ConcreteCommand(Command):
    def docstring(self) -> str:
        return "This is a concrete command implementation."
    
    def execute_command(self, *args) -> str:
        return f"Executed with args: {args}"
    
    def matches_command(self, cmd_str) -> bool:
        return cmd_str.startswith("cmd_")
    
    def parse_command(self, cmd_str) -> tuple:
        parts = cmd_str.split()
        return tuple(parts)


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 Command — the run constructed an instance of its own class and did not call it; no shape is stated — the instrument coerces any output to an array; for a Command 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
Replace Ran
ran on a synthesized input · constructed an object; no result was computed
gsasikiran/MLReplicate-benchmarking/AgentLaboratory/mlesolver.py
code served (permissive licence) · get_code("d2b82b5479ed387f")
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 an instance of its own class, Replace: the run constructed the object and did not call it, so no result was computed. 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

def extract_prompt(text, marker):
    """Stand-in for extract_prompt function"""
    start = text.find(f'```{marker}')
    if start == -1:
        return ""
    start += len(f'```{marker}') + 1
    end = text.find('```', start)
    if end == -1:
        end = len(text)
    return text[start:end].strip()

def execute_code(code_str):
    """Stand-in for execute_code function"""
    try:
        exec(code_str, {})
        return "Code executed successfully"
    except Exception as e:
        return f"[CODE EXECUTION ERROR] {str(e)}"

torch.manual_seed(0)

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 Replace — the run constructed an instance of its own class and did not call it; no shape is stated — the instrument coerces any output to an array; for a Replace 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_score Ran
ran; our contract draft was wrong, not the code
gsasikiran/MLReplicate-benchmarking/AgentLaboratory/mlesolver.py
code served (permissive licence) · get_code("c7d4030263209148")
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
import numpy as np
from unittest.mock import Mock, patch

torch.manual_seed(0)

def query_model(model_str, system_prompt, openai_api_key, prompt, temp):
    """Mock function that simulates querying a reward model LLM"""
    return "```SCORE\n0.85\n```"

def extract_prompt(text, word):
    """Mock function that extracts a score from formatted text"""
    import re
    pattern = rf'{word}\s+([\d.]+)'
    match = re.search(pattern, text)
    if match:
        return match.group(1)
    return "0.5"

outlined_plan = """
Research Plan:
1. Load the dataset
2. Build a neural network model
3. Train the model for 10 epochs
4. Evaluate on test set
5. Return accuracy score
"""

code = """
import torch
import torch.nn as nn

model = nn.Sequential(
    nn.Linear(10, 5),
    nn.ReLU(),
    nn.Linear(5, 1),
    nn.Sigmoid()
)

optimizer = torch.optim.Adam(model.parameters())
for epoch in range(10):
    pass

accuracy = 0.92
"""

code_return = """
Training completed successfully.
Final accuracy: 0.92
Model saved to checkpoint.pt
"""

REWARD_MODEL_LLM = "gpt-4"

ARGS = {
    "outlined_plan": outlined_plan,
    "code": code,
    "code_return": code_return,
    "REWARD_MODEL_LLM": REWARD_MODEL_LLM,
    "attempts": 3,
    "openai_api_key": None
}
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 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
Edit Not yet run
unverified (harvested, no recorded run)
gsasikiran/MLReplicate-benchmarking/AgentLaboratory/mlesolver.py
code served (permissive licence) · get_code("293a3841721aff09")
How it ran

no invocation recorded.

MLESolver Not yet run
unverified (harvested, no recorded run)
gsasikiran/MLReplicate-benchmarking/AgentLaboratory/mlesolver.py
code served (permissive licence) · get_code("a32b4e304b5f2f73")
How it ran

no invocation recorded.

code_repair Not yet run
unverified (harvested, no recorded run)
gsasikiran/MLReplicate-benchmarking/AgentLaboratory/mlesolver.py
code served (permissive licence) · get_code("706a3e7eed7b4393")
How it ran

no invocation recorded.

Also present in other papers' code

Of 6 samples lifted from this paper's linked repositories, 6 carry a function identity and 0 do not. Of the 6 distinct functions with an identity, 2 are also present in other papers whose code we hold (2 with at least one paper dated earlier, 0 the same month, 0 later) and 4 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
code_repair
brick of this paper's assembly: helper
1 other paper
1 dated earlier
2503.18102 (2025-03)
AgentRxiv: Towards Collaborative Autonomous Research
get_score
brick of this paper's assembly: metric
1 other paper
1 dated earlier
2503.18102 (2025-03)
AgentRxiv: Towards Collaborative Autonomous Research

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

Autonomous research systems capable of generating complete scientific manuscripts have advanced rapidly, yet robust and realistic evaluation frameworks have failed to keep pace. To bridge this gap, we introduce MLReplicate 5 , an end-to-end benchmark evaluating autonomous research systems on machine learning reproducibility. The benchmark was constructed from ICML 2025 outstanding papers reformulated into standardized input specifications and evaluated across 6 state-of-the-art research systems: AI SCIENTIST-V1, AI SCIENTIST-V2, AGENT LABORATORY, CYCLERESEARCHER, AI RESEARCHER, and TINY SCI-ENTIST, yielding 45 generated manuscripts, with 3 failed experiments. Outputs are assessed using a dual-protocol approach that combines automated conferencestyle review and structured expert human evaluation, while tracking computational cost, runtime, and the amount of required human intervention. The automated conference-style review accepted 10 out of 37 valid submissions. An additional 8 submissions were desk-rejected before review for failing to meet the minimum page threshold. In contrast to automated reviews, human reviewers consistently identified methodological flaws, hallucinated experimental results, and reproducibility failures across all systems, and 59% of accepted automated reviews contained fabricated or unsupported claims. We further find that neither token budget nor computational cost predicts output quality: the cheapest system outperforms the most resource-intensive system in human evaluation, despite a 38-fold difference in input tokens. We thus demonstrate that autonomous research workflow design matters more than the scale of compute. MLReplicate exposes a substantial gap between current autonomous research systems and genuine scientific rigor, and establishes a practical, extensible evaluation framework for systematic progress toward trustworthy AI-driven scientific discovery. * Equal contribution † Equal lead ‡ Equal supervision 5 Code and data are available:https://github.com/gsasikiran/MLReplicate-benchmarking Preprint.

For agents

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

get_harvested_code_for_paper("2605.16616")
get_code_for_paper("2605.16616")
have("2605.16616")

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

curl https://syntology.ai/api/ran/2605.16616.json

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

[![Syntology run record](https://syntology.ai/api/ran/2605.16616.svg)](https://syntology.ai/paper/2605.16616)

Connect an agent — have() is free.