SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2501.16142 · ICLR · 2025

Towards General-Purpose Model-Free Reinforcement Learning

Amy Zhang, Scott Fujimoto, Yuandong Tian, Michael Rabbat, Pierluca Oro

arXiv · PDF · Open in the Atlas

Code that ran

5 of 9 samples (functions and classes) lifted from this paper's linked repositories executed on a synthesized input in Syntology's jail (4 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; 1 failure of our instrument; 4 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/2501.16142.json · badge: /api/ran/2501.16142.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
sfujim/TD7 community
Papers with Code archive
4 ran · 1 no recorded run
copy not recorded community
not in the paper's declared repositories
1 ran · 0 no recorded run
facebookresearch/MRQ community
Papers with Code archive
0 ran · 3 no recorded run
FunctionStatusWhere it lives
Actor Ran
ran on a synthesized input · constructed an object; no result was computed
sfujim/TD7/TD7.py
code served (permissive licence) · get_code("e41e9160d3bfd93f")
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, Actor: 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
import torch.nn.functional as F

torch.manual_seed(0)

# Define parameters for Actor
state_dim = 4
action_dim = 3
zs_dim = 5
hdim = 6
activ = F.relu

# Create small sample tensors for forward pass testing
batch_size = 2
state = torch.randn(batch_size, state_dim)
zs = torch.randn(batch_size, zs_dim)

ARGS = {
    'state_dim': state_dim,
    'action_dim': action_dim,
    'zs_dim': zs_dim,
    'hdim': hdim,
    'activ': activ
}
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 Actor — 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 Actor 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
Critic Ran
ran on a synthesized input · constructed an object; no result was computed
sfujim/TD7/TD7.py
code served (permissive licence) · get_code("32addd2524f26a0b")
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, Critic: 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
import torch.nn.functional as F

torch.manual_seed(0)

def AvgL1Norm(x, eps=1e-08):
    return x / x.abs().mean(-1, keepdim=True).clamp(min=eps)

class Critic(nn.Module):

    def __init__(self, state_dim, action_dim, zs_dim=256, hdim=256, activ=F.elu):
        super(Critic, self).__init__()
        self.activ = activ
        self.q01 = nn.Linear(state_dim + action_dim, hdim)
        self.q1 = nn.Linear(2 * zs_dim + hdim, hdim)
        self.q2 = nn.Linear(hdim, hdim)
        self.q3 = nn.Linear(hdim, 1)
        self.q02 = nn.Linear(state_dim + action_dim, hdim)
        self.q4 = nn.Linear(2 * zs_dim + hdim, hdim)
        self.q5 = nn.Linear(hdim, hdim)
        self.q6 = nn.Linear(hdim, 1)

    def forward(self, state, action, zsa, zs):
        sa = torch.cat([state, action], 1)
        embeddings = torch.cat([zsa, zs], 1)
        q1 = AvgL1Norm(self.q01(sa))
        q1 = torch.cat([q1, embeddings], 1)
        q1 = self.activ(self.q1(q1))
        q1 = self.activ(self.q2(q1))
        q1 = self.q3(q1)
        q2 = AvgL1Norm(self.q02(sa))
        q2 = torch.cat([q2, embeddings], 1)
        q2 = self.activ(self.q4(q2))
        q2 = self.activ(self.q5(q2))
        q2 = self.q6(q2)
        return torch.cat([q1, q2], 1)

state_dim = 4
action_dim = 3
zs_dim = 8
hdim = 16
batch_size = 2

state = torch.randn(batch_size, state_dim)
action = torch.randn(batch_size, action_dim)
zsa = torch.randn(batch_size, zs_dim)
zs = torch.randn(batch_size, zs_dim)

ARGS = {
    "state_dim": state_dim,
    "action_dim": action_dim,
    "zs_dim": zs_dim,
    "hdim": hdim,
    "activ": F.elu,
}
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 Critic — 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 Critic 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
Hyperparameters Ran
ran on a synthesized input · constructed an object; no result was computed
sfujim/TD7/TD7.py
code served (permissive licence) · get_code("bd1d987513dd5174")
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, Hyperparameters: 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.functional as F
from dataclasses import dataclass
from typing import Callable

torch.manual_seed(0)

@dataclass
class Hyperparameters:
    batch_size: int = 256
    buffer_size: int = 1000000.0
    discount: float = 0.99
    target_update_rate: int = 250
    exploration_noise: float = 0.1
    target_policy_noise: float = 0.2
    noise_clip: float = 0.5
    policy_freq: int = 2
    alpha: float = 0.4
    min_priority: float = 1
    lmbda: float = 0.1
    max_eps_when_checkpointing: int = 20
    steps_before_checkpointing: int = 750000.0
    reset_weight: float = 0.9
    zs_dim: int = 256
    enc_hdim: int = 256
    enc_activ: Callable = F.elu
    encoder_lr: float = 0.0003
    critic_hdim: int = 256
    critic_activ: Callable = F.elu
    critic_lr: float = 0.0003
    actor_hdim: int = 256
    actor_activ: Callable = F.relu
    actor_lr: float = 0.0003

ARGS = {
    "batch_size": 256,
    "buffer_size": 1000000.0,
    "discount": 0.99,
    "target_update_rate": 250,
    "exploration_noise": 0.1,
    "target_policy_noise": 0.2,
    "noise_clip": 0.5,
    "policy_freq": 2,
    "alpha": 0.4,
    "min_priority": 1,
    "lmbda": 0.1,
    "max_eps_when_checkpointing": 20,
    "steps_before_checkpointing": 750000.0,
    "reset_weight": 0.9,
    "zs_dim": 256,
    "enc_hdim": 256,
    "enc_activ": F.elu,
    "encoder_lr": 0.0003,
    "critic_hdim": 256,
    "critic_activ": F.elu,
    "critic_lr": 0.0003,
    "actor_hdim": 256,
    "actor_activ": F.relu,
    "actor_lr": 0.0003,
}
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 Hyperparameters — 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 Hyperparameters 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
LAP Ran
ran on a synthesized input · constructed an object; no result was computed
sfujim/TD7/TD7.py
code served (permissive licence) · get_code("7ffb1daeea0ca926")
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, LAP: 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 numpy as np

torch.manual_seed(0)
np.random.seed(0)

# Define parameters
state_dim = 4
action_dim = 3
device = 'cpu'
max_size = 1000.0
batch_size = 2
max_action = 1.0
normalize_actions = True
prioritized = True

# Build ARGS dictionary with all parameters
ARGS = {
    'state_dim': state_dim,
    'action_dim': action_dim,
    'device': device,
    'max_size': max_size,
    'batch_size': batch_size,
    'max_action': max_action,
    'normalize_actions': normalize_actions,
    'prioritized': prioritized
}
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 LAP — 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 LAP 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
LAP_huber Ran
ran; our contract draft was wrong, not the code
this paper's copy was not recorded; identical code first harvested from sfujim/TD7
pointer only · get_code("f829c1e4fecf264f")
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.

Agent Not yet run
unverified (harvested, no recorded run)
sfujim/TD7/TD7.py
code served (permissive licence) · get_code("34167534733a08e9")
How it ran

no invocation recorded.

Agent Not yet run
unverified (harvested, no recorded run)
facebookresearch/MRQ/MRQ/MRQ.py
pointer only (licence: NOASSERTION) · get_code("186ce51f6191d32c")
How it ran

The body of this sample is withheld: GitHub could not classify the licence of this paper's repository, facebookresearch/MRQ (NOASSERTION). 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. No invocation recorded.

Hyperparameters Not yet run
unverified (harvested, no recorded run)
facebookresearch/MRQ/MRQ/MRQ.py
pointer only (licence: NOASSERTION) · get_code("53d5ca315dd6bdc0")
How it ran

The body of this sample is withheld: GitHub could not classify the licence of this paper's repository, facebookresearch/MRQ (NOASSERTION). 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. No invocation recorded.

ReplayBuffer Not yet run
unverified (harvested, no recorded run)
facebookresearch/MRQ/MRQ/MRQ.py
pointer only (licence: NOASSERTION) · get_code("39dd84fd5db384b3")
How it ran

The body of this sample is withheld: GitHub could not classify the licence of this paper's repository, facebookresearch/MRQ (NOASSERTION). 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. No invocation recorded.

Also present in other papers' code

Of 9 samples lifted from this paper's linked repositories, 9 carry a function identity and 0 do not. Of the 9 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 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.

FunctionAlso present inEarliest-datedOthers named
LAP_huber
brick of this paper's assembly: loss
1 other paper
1 dated earlier
2306.02451 (2023-06)
For SALE: State-Action Representation Learning for Deep 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

Reinforcement learning (RL) promises a framework for near-universal problemsolving. In practice however, RL algorithms are often tailored to specific benchmarks, relying on carefully tuned hyperparameters and algorithmic choices. Recently, powerful model-based RL methods have shown impressive general results across benchmarks but come at the cost of increased complexity and slow run times, limiting their broader applicability. In this paper, we attempt to find a unifying model-free deep RL algorithm that can address a diverse class of domains and problem settings. To achieve this, we leverage model-based representations that approximately linearize the value function, taking advantage of the denser task objectives used by model-based RL while avoiding the costs associated with planning or simulated trajectories. We evaluate our algorithm, MR.Q, on a variety of common RL benchmarks with a single set of hyperparameters and show a competitive performance against domain-specific and general baselines, providing a concrete step towards building general-purpose model-free deep RL algorithms.

For agents

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

get_harvested_code_for_paper("2501.16142")
get_code_for_paper("2501.16142")
have("2501.16142")

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

curl https://syntology.ai/api/ran/2501.16142.json

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

[![Syntology run record](https://syntology.ai/api/ran/2501.16142.svg)](https://syntology.ai/paper/2501.16142)

Connect an agent — have() is free.