SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 1803.07055 · 2018

Simple random search provides a competitive approach to reinforcement learning

arXiv · PDF · Open in the Atlas

Code that ran

13 of 15 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 9 have no contract verdict; 4 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/1803.07055.json · badge: /api/ran/1803.07055.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
billchan226/poar-srl-4-robot community (classifier label)
classifier basis not recorded on the link
8 ran · 0 no recorded run
eric-erki/robotics-rl-srl community (classifier label)
classifier basis not recorded on the link
2 ran · 1 no recorded run
kayuksel/pytorch-ars extension (classifier label)
classifier basis: classifier_2025
2 ran · 0 no recorded run
jainpankul/Augmented-Random-Search reimplementation (classifier label)
classifier basis: classifier_2025
1 ran · 0 no recorded run
suvhradipghosh07/Augmented-Random-Search community (classifier label)
classifier basis not recorded on the link
0 ran · 1 no recorded run
FunctionStatusWhere it lives
calculate_loss Ran
ran; our contract draft was wrong, not the code
kayuksel/pytorch-ars/ars_dataparallel.py
pointer only (licence: GPL-3.0) · get_code("f34d83ec198f0b0c")
How it ran

The body of this sample is withheld: GPL-3.0, as GitHub detects it on this paper's repository, kayuksel/pytorch-ars, is not in the permissive set Syntology serves inline. 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.

conv3x3 Ran
ran; our contract draft was wrong, not the code
billchan226/poar-srl-4-robot/srl_zoo/models/models.py
code served (permissive licence) · get_code("00d0cc5eadefcd18")
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 Conv2d. 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)

ARGS = {
    "in_planes": 3,
    "out_planes": 64,
    "stride": 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 Conv2d — a Conv2d output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a Conv2d 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
convertImagePath Ran
ran on a synthesized input
eric-erki/robotics-rl-srl/environments/dataset_generator.py
code served (permissive licence) · get_code("49accadcbda40a13")
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 argparse

# Create a minimal ArgumentParser object with the required 'name' attribute
parser = argparse.ArgumentParser()
parser.add_argument('--name', type=str, default='output')
args = parser.parse_args(['--name', 'dataset'])

# Define the path - this should contain a structure like: .../prefix_123/image.jpg
path = '/data/record_045/image_001.jpg'

# Define record_id_start as an integer
record_id_start = 100

# Map all parameters
ARGS = {
    'args': args,
    'path': path,
    'record_id_start': record_id_start
}
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
encodeOneHot Ran
ran on a synthesized input
billchan226/poar-srl-4-robot/srl_zoo/models/models.py
code served (permissive licence) · get_code("dfab5e99492d29c9")
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 Tensor, ndim 2, dtype family float, shape (2, 4) 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 as th

th.manual_seed(0)

# Create a small tensor with class indices
# The tensor should contain indices (0 to n_dim-1) to be one-hot encoded
# Shape: (batch_size, ) or (batch_size, 1)
tensor = th.tensor([[0], [3]], dtype=th.long)

# Number of dimensions for one-hot encoding
n_dim = 4

ARGS = {
    "tensor": tensor,
    "n_dim": n_dim,
}
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 Tensor, ndim 2, dtype family float, shape (2, 4) — shape (2, 4) 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
findDissimilar Ran
ran on a synthesized input
billchan226/poar-srl-4-robot/srl_zoo/losses/utils.py
code served (permissive licence) · get_code("fe771d1b728a1d21")
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 ndarray, ndim 1, dtype family int, shape (0,) 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 numpy as np

# Set up small numpy arrays for the function parameters
# The function indexes into these arrays, so we need to ensure indices are valid

# Create a minibatch of indices (e.g., batch of 2 samples with indices)
minibatch1 = np.array([0, 1])

# Create another minibatch of indices
minibatch2 = np.array([0, 1])

# Create actions array - needs to be large enough to index with minibatch1 and minibatch2
# Actions are compared, so they should have different values to test the logic
actions = np.array([1, 2, 1, 2])

# Create rewards array - needs length > max(minibatch1) + 1 and > max(minibatch2) + 1
# We access rewards[minibatch2 + 1] and rewards[minibatch1[index] + 1]
# With minibatch1/minibatch2 having max value 1, we need rewards of length >= 3
rewards = np.array([0.5, 1.0, 2.0, 3.0])

# The index parameter selects which element from minibatch1 to use
index = 0

ARGS = {
    "index": index,
    "minibatch1": minibatch1,
    "minibatch2": minibatch2,
    "actions": actions,
    "rewards": rewards
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type ndarray, ndim 1, dtype family int, shape (0,) — shape (0,) 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
findSameActions Ran
ran on a synthesized input
billchan226/poar-srl-4-robot/srl_zoo/losses/utils.py
code served (permissive licence) · get_code("61e7389045c32bd7")
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 ndarray, ndim 1, dtype family int, shape (3,) 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 numpy as np

# Create sample data
np.random.seed(0)

# Create a minibatch: array of indices into the actions array
minibatch = np.array([0, 2, 4, 1, 3])

# Create actions: array where each element represents the action taken at that index
# We'll make it so some actions are repeated to test the function
actions = np.array([1, 2, 1, 3, 1, 2, 3])

# Choose a reference index within the minibatch
index = 1  # This points to minibatch[1] = 2, which has action actions[2] = 1

ARGS = {
    "index": index,
    "minibatch": minibatch,
    "actions": actions
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type ndarray, ndim 1, dtype family int, shape (3,) — shape (3,) 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
forwardModelLoss Ran
ran on a synthesized input
billchan226/poar-srl-4-robot/srl_zoo/losses/losses.py
code served (permissive licence) · get_code("649af1e84314be4e")
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 Tensor, 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 torch.nn as nn

torch.manual_seed(0)

# Create tiny tensors for next_states_pred and next_states
# Batch size: 2, Dimension: 4
next_states_pred = torch.randn(2, 4)
next_states = torch.randn(2, 4)

# Weight coefficient
weight = 0.5

# LossManager stand-in class
class LossManager:
    def __init__(self):
        self.losses = {}
    
    def addToLosses(self, loss_name, weight, loss_value):
        """Add a loss to the manager"""
        self.losses[loss_name] = (weight, loss_value)

loss_manager = LossManager()

ARGS = {
    'next_states_pred': next_states_pred,
    'next_states': next_states,
    'weight': weight,
    'loss_manager': loss_manager
}
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 Tensor, 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
getSRLDim Ran
ran on a synthesized input
eric-erki/robotics-rl-srl/state_representation/models.py
code served (permissive licence) · get_code("305f5edcff073267")
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 int, ndim 0, dtype family int, 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 json
import os
import tempfile
from collections import OrderedDict

# Create a temporary directory and config file for the path-based test
temp_dir = tempfile.mkdtemp()
log_folder = temp_dir + '/'

# Create a valid exp_config.json file
exp_config = OrderedDict([
    ('state-dim', 128),
    ('other_key', 'value')
])

config_path = log_folder + 'exp_config.json'
with open(config_path, 'w') as f:
    json.dump(exp_config, f)

# Create a model path that points to a file in the log folder
model_path = log_folder + 'model.pkl'

# Create a stand-in gym env object with the required method
class StandInEnv:
    def getGroundTruthDim(self):
        return 64

env_object = StandInEnv()

# Define ARGS with both parameter options
ARGS = {
    'path': model_path,
    'env_object': env_object
}
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 int, ndim 0, dtype family int, 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
get_batch Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
kayuksel/pytorch-ars/ars_dataparallel.py
pointer only (licence: GPL-3.0) · get_code("bd8890576c4c1b92")
How it ran

The body of this sample is withheld: GPL-3.0, as GitHub detects it on this paper's repository, kayuksel/pytorch-ars, is not in the permissive set Syntology serves inline. 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.

inverseModelLoss Ran
ran on a synthesized input
billchan226/poar-srl-4-robot/srl_zoo/losses/losses.py
code served (permissive licence) · get_code("e44e088cf66053b3")
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 Tensor, 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 torch.nn as nn

torch.manual_seed(0)

# actions_pred: predicted actions logits, shape (batch_size, num_actions)
# Should be output of a model, typically (batch, num_classes)
actions_pred = torch.randn(2, 4)

# actions_st: true action indices, shape (batch_size, 1)
# CrossEntropyLoss expects (batch_size,) after squeeze, so we need (batch_size, 1)
actions_st = torch.tensor([[1], [3]])

# weight: coefficient to weight the loss
weight = 0.5

# loss_manager: needs addToLosses method that takes (loss_name, weight, loss_value)
class LossManager:
    def __init__(self):
        self.losses = {}
    
    def addToLosses(self, name, weight, loss_value):
        self.losses[name] = (weight, loss_value)

loss_manager = LossManager()

ARGS = {
    'actions_pred': actions_pred,
    'actions_st': actions_st,
    'weight': weight,
    'loss_manager': loss_manager
}
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 Tensor, 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
mkdir Ran
ran; our contract draft was wrong, not the code
jainpankul/Augmented-Random-Search/ARS.py
pointer only (licence: NONE) · get_code("6b3bc12652205efd")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, jainpankul/Augmented-Random-Search when it was checked. 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.

overSampling Ran
ran on a synthesized input
billchan226/poar-srl-4-robot/srl_zoo/losses/utils.py
code served (permissive licence) · get_code("5e577a99c2ee9cb3")
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. 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 numpy as np

def findDissimilar(i, minibatch1, minibatch2, actions, rewards):
    """
    Stand-in function that returns indices where rewards are dissimilar.
    Returns indices j from minibatch2 where reward differs from minibatch1[i].
    """
    dissimilar_indices = []
    for j in range(len(minibatch2)):
        if rewards[i] != rewards[j]:
            dissimilar_indices.append(j)
    return dissimilar_indices

# Setup parameters
batch_size = 2

# Create mini-batch list: list of numpy arrays representing minibatches
# Each minibatch is a 2D array where rows are observations
m_list = [
    np.array([[1.0, 2.0, 3.0, 4.0], 
              [5.0, 6.0, 7.0, 8.0]]),
    np.array([[9.0, 10.0, 11.0, 12.0], 
              [13.0, 14.0, 15.0, 16.0]])
]

# Create pairs: list of numpy arrays, one per minibatch
# Initially empty or sparse pairs to trigger oversampling
pairs = [
    np.array([]),  # Empty pairs for first minibatch
    np.array([[0, 1]])  # Some pairs for second minibatch
]

# Function to find dissimilar pairs
function_on_pairs = findDissimilar

# Actions array: one action per observation
actions = np.array([0, 1, 2, 3, 1])

# Rewards array: one reward per observation
rewards = np.array([10.0, 20.0, 10.0, 30.0, 20.0])

ARGS = {
    'batch_size': batch_size,
    'm_list': m_list,
    'pairs': pairs,
    'function_on_pairs': function_on_pairs,
    'actions': actions,
    'rewards': rewards
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type tuple — a tuple output has no array shape of its own; none is 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
roboticPriorsLoss Ran
ran on a synthesized input
billchan226/poar-srl-4-robot/srl_zoo/losses/losses.py
code served (permissive licence) · get_code("4d38f7f1d06a5cfd")
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 Tensor, 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 torch as th
import numpy as np

th.manual_seed(0)

# Stand-in LossManager class
class LossManager:
    def __init__(self):
        self.losses = {}
    
    def addToLosses(self, name, weight, loss_value):
        if name not in self.losses:
            self.losses[name] = []
        self.losses[name].append((weight, loss_value.item() if isinstance(loss_value, th.Tensor) else loss_value))

# Create small tensors
batch_size = 2
state_dim = 4

states = th.randn(batch_size, state_dim)
next_states = th.randn(batch_size, state_dim)

# Create pair indices for dissimilar_pairs and same_actions_pairs
# These should be 2D arrays where each row contains two indices
dissimilar_pairs_batch = [
    np.array([[0, 1]], dtype=np.int64),  # one pair per batch
    np.array([[0, 1]], dtype=np.int64),
]

same_actions_pairs_batch = [
    np.array([[0, 1]], dtype=np.int64),  # one pair per batch
    np.array([[0, 1]], dtype=np.int64),
]

minibatch_idx = 0
weight = 0.5

loss_manager = LossManager()

ARGS = {
    "states": states,
    "next_states": next_states,
    "minibatch_idx": minibatch_idx,
    "dissimilar_pairs": dissimilar_pairs_batch,
    "same_actions_pairs": same_actions_pairs_batch,
    "weight": weight,
    "loss_manager": loss_manager,
}
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 Tensor, 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
explore Not yet run
unverified (harvested, no recorded run)
suvhradipghosh07/Augmented-Random-Search/_ai_.py
code served (permissive licence) · get_code("44c6ee6398a602b8")
How it ran

no invocation recorded.

register Not yet run
unverified (harvested, no recorded run)
eric-erki/robotics-rl-srl/environments/registry.py
code served (permissive licence) · get_code("171decf255010ed6")
How it ran

no invocation recorded.

Also present in other papers' code

Of 15 samples lifted from this paper's linked repositories, 15 carry a function identity and 0 do not. Of the 15 distinct functions with an identity, 2 are also present in other papers whose code we hold (2 with at least one paper dated earlier, 1 the same month, 1 later) and 13 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
conv3x3 686 other papers
37 dated earlier, 8 the same month, 641 later · 31 only as an assembly brick
1406.2661 (2014-06)
Generative Adversarial Networks
1409.0575 (2014-09)
ImageNet Large Scale Visual Recognition Challenge
1412.6572 (2014-12)
Explaining and Harnessing Adversarial Examples
and 683 more not named here
mkdir 1 other paper
1 dated earlier
1703.03864 (2017-03)
Evolution Strategies as a Scalable Alternative to Reinforcement Learning

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

A common belief in model-free reinforcement learning is that methods based on random search in the parameter space of policies exhibit significantly worse sample complexity than those that explore the space of actions. We dispel such beliefs by introducing a random search method for training static, linear policies for continuous control problems, matching state-of-the-art sample efficiency on the benchmark MuJoCo locomotion tasks. Our method also finds a nearly optimal controller for a challenging instance of the Linear Quadratic Regulator, a classical problem in control theory, when the dynamics are not known. Computationally, our random search algorithm is at least 15 times more efficient than the fastest competing model-free methods on these benchmarks. We take advantage of this computational efficiency to evaluate the performance of our method over hundreds of random seeds and many different hyperparameter configurations for each benchmark task. Our simulations highlight a high variability in performance in these benchmark tasks, suggesting that commonly used estimations of sample efficiency do not adequately evaluate the performance of RL algorithms.

For agents

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

get_harvested_code_for_paper("1803.07055")
get_code_for_paper("1803.07055")
have("1803.07055")

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

curl https://syntology.ai/api/ran/1803.07055.json

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

[![Syntology run record](https://syntology.ai/api/ran/1803.07055.svg)](https://syntology.ai/paper/1803.07055)

Connect an agent — have() is free.