SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2208.05083 · ICML · 2022

Reducing Exploitability with Population Based Training

Adam Gleave, Pavel Czempin

arXiv · PDF · Open in the Atlas

Code that ran

5 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 5 have no contract verdict; 0 failures of our instrument; 0 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/2208.05083.json · badge: /api/ran/2208.05083.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
humancompatibleai/reducing-exploitability canonical (classifier label)
classifier basis not recorded on the link
5 ran · 0 no recorded run
FunctionStatusWhere it lives
create_parser Ran
ran on a synthesized input
humancompatibleai/reducing-exploitability/src/dep/evaluate.py
code served (permissive licence) · get_code("0e5de1a0b68887a8")
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 ArgumentParser. 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
import json

def parser_creator(**kwargs):
    """Minimal parser creator that returns an ArgumentParser."""
    return argparse.ArgumentParser(**kwargs)

ARGS = {
    'parser_creator': parser_creator
}
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 ArgumentParser — a ArgumentParser output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a ArgumentParser 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
default_policy_agent_mapping Ran
ran on a synthesized input
humancompatibleai/reducing-exploitability/src/dep/evaluate.py
code served (permissive licence) · get_code("f73e792e0b5295b0")
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 ray.rllib.policy.sample_batch import DEFAULT_POLICY_ID

def default_policy_agent_mapping(unused_agent_id):
    return DEFAULT_POLICY_ID

ARGS = {
    "unused_agent_id": "agent_0"
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
ray — 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
get_checkpoint_file Ran
ran on a synthesized input
humancompatibleai/reducing-exploitability/src/aprl_defense/common/io.py
code served (permissive licence) · get_code("b7426b39719abff3")
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 PosixPath. 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 tempfile
import os
from pathlib import Path

# Create a temporary directory structure that mimics a checkpoint directory
temp_dir = tempfile.mkdtemp()
checkpoints_dir = Path(temp_dir) / 'checkpoints'
checkpoints_dir.mkdir(exist_ok=True)

# Create a specific checkpoint folder with a timestamp-like name
checkpoint_folder = checkpoints_dir / 'checkpoint_000001'
checkpoint_folder.mkdir(exist_ok=True)

# Create a checkpoint file in the folder
checkpoint_file = checkpoint_folder / 'checkpoint.pt'
checkpoint_file.write_text('dummy checkpoint content')

ARGS = {
    'path': temp_dir,
    'specific_folder': False
}
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 PosixPath — a PosixPath output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a PosixPath 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_saved_config Ran
ran on a synthesized input
humancompatibleai/reducing-exploitability/src/aprl_defense/common/io.py
code served (permissive licence) · get_code("0744420afc4631e3")
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
import tempfile
from pathlib import Path
from ray.cloudpickle import cloudpickle
import os

# Create a temporary directory structure for testing
temp_dir = tempfile.mkdtemp()
test_dir = Path(temp_dir) / "subdir" / "checkpoint"
test_dir.mkdir(parents=True, exist_ok=True)

# Create a sample config to pickle
sample_config = {
    "learning_rate": 0.001,
    "batch_size": 32,
    "num_epochs": 10,
    "model_dim": 128
}

# Save the config to params.pkl in the parent directory
config_save_path = test_dir.parent / "params.pkl"
with open(config_save_path, 'wb') as f:
    cloudpickle.dump(sample_config, f)

# Create a test checkpoint path (in the subdir where get_saved_config expects to find it)
checkpoint_path = test_dir / "model.ckpt"
checkpoint_path.touch()

# Define ARGS with the path parameter
ARGS = {
    "path": str(checkpoint_path)
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
ray, 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
save_params Ran
ran on a synthesized input
humancompatibleai/reducing-exploitability/src/aprl_defense/common/rllib_io.py
code served (permissive licence) · get_code("21303ac8a9098ecf")
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 PosixPath. 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 pathlib import Path
import tempfile

# Create a temporary directory for log_dir
temp_dir = tempfile.mkdtemp()
log_dir = Path(temp_dir)

# Create a simple config dict
config = {
    'model': 'test_model',
    'learning_rate': 0.001,
    'batch_size': 32,
    'num_epochs': 10,
    'layers': [64, 128, 256],
    'activation': 'relu'
}

ARGS = {
    'config': config,
    'log_dir': log_dir
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
ray — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type PosixPath — a PosixPath output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a PosixPath 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

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, 1 are also present in other papers whose code we hold (1 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
default_policy_agent_mapping 1 other paper
1 dated earlier
1807.05118 (2018-07)
Tune: A Research Platform for Distributed Model Selection and Training

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

Self-play reinforcement learning has achieved state-of-the-art, and often superhuman, performance in a variety of zero-sum games. Yet prior work has found that policies that are highly capable against regular opponents can fail catastrophically against adversarial policies: an opponent trained explicitly against the victim. Prior defenses using adversarial training were able to make the victim robust to a specific adversary, but the victim remained vulnerable to new ones. We conjecture this limitation was due to insufficient diversity of adversaries seen during training. We analyze a defense using population based training to pit the victim against a diverse set of opponents. We evaluate this defense's robustness against new adversaries in two low-dimensional environments. This defense increases robustness against adversaries, as measured by the number of attacker training timesteps to exploit the victim. Furthermore, we show that robustness is correlated with the size of the opponent population.

For agents

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

get_harvested_code_for_paper("2208.05083")
get_code_for_paper("2208.05083")
have("2208.05083")

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

curl https://syntology.ai/api/ran/2208.05083.json

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

[![Syntology run record](https://syntology.ai/api/ran/2208.05083.svg)](https://syntology.ai/paper/2208.05083)

Connect an agent — have() is free.