SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2209.12016 · ICML · 2023

Mastering the Unsupervised Reinforcement Learning Benchmark from Pixels

Tim Verbelen, Bart Dhoedt, Aaron Courville, Alexandre Lacoste, Alexandre Piché, Pietro Mazzaglia, Sai Rajeswar

arXiv · PDF · Open in the Atlas

Code that ran

17 of 24 samples (functions and classes) lifted from this paper's linked repositories executed on a synthesized input in Syntology's jail (14 of them constructed an object rather than computing a result; 1 honoured the contract we drafted, 0 violated it and 14 have no contract verdict; 2 failures of our instrument; 7 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/2209.12016.json · badge: /api/ran/2209.12016.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
mazpie/mastering-urlb marked official in Papers with Code
Papers with Code archive
17 ran · 7 no recorded run
FunctionStatusWhere it lives
BernoulliDist Ran
ran on a synthesized input · constructed an object; no result was computed
mazpie/mastering-urlb/agent/dreamer.py
code served (permissive licence) · get_code("3200c778b6d1d0a8")
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, BernoulliDist: 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.distributions as D

torch.manual_seed(0)

# BernoulliDist.__init__ takes logits and probs parameters
# The sample method uses super().sample(), super().probs, and performs tensor operations
# We need to provide either logits or probs (not both, as per Bernoulli convention)

# Create a small logits tensor
logits = torch.randn(2, 4)

ARGS = {
    'logits': logits,
    'probs': None,
}
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 BernoulliDist — 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 BernoulliDist 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
DistLayer Ran
ran on a synthesized input · constructed an object; no result was computed
mazpie/mastering-urlb/agent/dreamer.py
code served (permissive licence) · get_code("acf159e783acb4b1")
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, DistLayer: 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 numpy as np

torch.manual_seed(0)

# Define ARGS dict with all parameters for DistLayer.__init__
ARGS = {
    'in_dim': 8,
    'shape': [4, 3],
    'dist': 'normal',
    'min_std': 0.1,
    'init_std': 0.0,
    'bias': True,
}
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 DistLayer — 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 DistLayer 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
Encoder Ran
ran on a synthesized input · constructed an object; no result was computed
mazpie/mastering-urlb/agent/dreamer.py
code served (permissive licence) · get_code("478e69ff513053f9")
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, Encoder: 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 numpy as np

torch.manual_seed(0)

ARGS = {
    'shapes': {
        'image': (3, 64, 64),  # 3D shape for CNN
        'vector': (16,),        # 1D shape for MLP
    },
    'cnn_keys': '.*',
    'mlp_keys': '.*',
    'act': nn.ELU,
    'norm': 'none',
    'cnn_depth': 48,
    'cnn_kernels': (4, 4, 4, 4),
    'mlp_layers': [400, 400, 400, 400],
}
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 Encoder — 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 Encoder 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
EnsembleRSSM Ran
ran on a synthesized input · constructed an object; no result was computed
mazpie/mastering-urlb/agent/dreamer.py
code served (permissive licence) · get_code("1007de8c2714718e")
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, EnsembleRSSM: the run constructed the object and did not call it, so no result was computed. The image it ran in was not recorded. The code mentions cuda, and the harness aliases cuda to the CPU for every sample; whether this run needed that is 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 = {
    'ensemble': 5,
    'stoch': 30,
    'deter': 200,
    'hidden': 200,
    'discrete': False,
    'act': nn.ELU,
    'norm': 'none',
    'std_act': 'softplus',
    'min_std': 0.1,
    'action_dim': 4,
    'embed_dim': 1536,
    'device': 'cpu',
}
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 EnsembleRSSM — 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 EnsembleRSSM 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
GRUCell Ran
ran on a synthesized input · constructed an object; no result was computed
mazpie/mastering-urlb/agent/dreamer.py
code served (permissive licence) · get_code("3c568e9e0347391a")
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, GRUCell: the run constructed the object and did not call it, so no result was computed. The image it ran in was not recorded. The code mentions cuda, and the harness aliases cuda to the CPU for every sample; whether this run needed that is 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 = {
    'inp_size': 4,
    'size': 3,
    'norm': False,
    'act': nn.Tanh,
    'update_bias': -1,
    'device': 'cpu',
}
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 GRUCell — 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 GRUCell 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
NormLayer Ran
ran on a synthesized input · constructed an object; no result was computed
mazpie/mastering-urlb/agent/dreamer.py
code served (permissive licence) · get_code("e2e72a75f391e079")
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, NormLayer: 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

torch.manual_seed(0)

ARGS = {
    'name': 'layer',
    'dim': 4
}
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 NormLayer — 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 NormLayer 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
OneHotDist Ran
ran on a synthesized input · constructed an object; no result was computed
mazpie/mastering-urlb/agent/dreamer.py
code served (permissive licence) · get_code("66fbe95d6fa0d1a6")
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, OneHotDist: 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
import torch.distributions as D

torch.manual_seed(0)

# The OneHotDist class uses logits parameter in its methods
# We need to provide either logits or probs to the parent class OneHotCategorical
# Looking at the usage: super().logits is accessed, so we need valid logits

# Create a small logits tensor: batch_size=2, num_classes=4
logits = torch.randn(2, 4)

ARGS = {
    'logits': logits,
    'probs': None,
}
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 OneHotDist — 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 OneHotDist 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
Optimizer Ran
ran on a synthesized input · constructed an object; no result was computed
mazpie/mastering-urlb/agent/dreamer.py
code served (permissive licence) · get_code("374c4448ac4f78c9")
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, Optimizer: the run constructed the object and did not call it, so no result was computed. The image it ran in was not recorded. The code mentions cuda, and the harness aliases cuda to the CPU for every sample; whether this run needed that is 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)

# Create a small model with parameters
class SimpleModel(nn.Module):
    def __init__(self):
        super().__init__()
        self.linear1 = nn.Linear(4, 3)
        self.linear2 = nn.Linear(3, 2)
    
    def forward(self, x):
        return self.linear2(torch.relu(self.linear1(x)))

model = SimpleModel()
parameters = model.parameters()

ARGS = {
    'name': 'test_optimizer',
    'parameters': parameters,
    'lr': 0.001,
    'eps': 0.0001,
    'clip': 1.0,
    'wd': 0.01,
    'opt': 'adam',
    'wd_pattern': '.*',
    'use_amp': False,
}
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 Optimizer — 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 Optimizer 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
RequiresGrad Ran
ran on a synthesized input · constructed an object; no result was computed
mazpie/mastering-urlb/agent/dreamer.py
code served (permissive licence) · get_code("eb6cf9a53f925599")
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, RequiresGrad: 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

torch.manual_seed(0)

class SimpleModel(nn.Module):
    def __init__(self):
        super().__init__()
        self.linear = nn.Linear(4, 3)
    
    def forward(self, x):
        return self.linear(x)

model = SimpleModel()

ARGS = {
    'model': model
}
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 RequiresGrad — 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 RequiresGrad 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
SampleDist Ran
ran on a synthesized input · constructed an object; no result was computed
mazpie/mastering-urlb/agent/dreamer.py
code served (permissive licence) · get_code("1ec2bdb7622b9c66")
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, SampleDist: 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.distributions as D

torch.manual_seed(0)

# Create a simple distribution to pass to SampleDist
# Using Normal distribution with batch_shape=(2,) and event_shape=(4,)
# This means we have 2 independent batches, each with 4-dimensional events
loc = torch.zeros(2, 4)
scale = torch.ones(2, 4)
dist = D.Normal(loc, scale)

ARGS = {
    'dist': dist,
    'samples': 100
}
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 SampleDist — 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 SampleDist 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
SquashedNormal Ran
ran on a synthesized input · constructed an object; no result was computed
mazpie/mastering-urlb/agent/dreamer.py
code served (permissive licence) · get_code("bd0264844e671c00")
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, SquashedNormal: 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 torch import distributions as pyd
import math

torch.manual_seed(0)

# Create small tensors for loc and scale parameters
loc = torch.randn(2, 4)  # batch_size=2, dim=4
scale = torch.ones(2, 4) * 0.5  # batch_size=2, dim=4

ARGS = {
    "loc": loc,
    "scale": scale,
}
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 SquashedNormal — 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 SquashedNormal 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
StreamNorm Ran
ran on a synthesized input · constructed an object; no result was computed
mazpie/mastering-urlb/agent/dreamer.py
code served (permissive licence) · get_code("0e5160a9bd6bd23c")
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, StreamNorm: the run constructed the object and did not call it, so no result was computed. The image it ran in was not recorded. The code mentions cuda, and the harness aliases cuda to the CPU for every sample; whether this run needed that is 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)

ARGS = {
    'shape': (4, 3),
    'momentum': 0.99,
    'scale': 1.0,
    'eps': 1e-08,
    'device': 'cpu'
}
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 StreamNorm — 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 StreamNorm 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
TanhTransform Ran
ran on a synthesized input · constructed an object; no result was computed
mazpie/mastering-urlb/agent/dreamer.py
code served (permissive licence) · get_code("b88483dbb52ca5a8")
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, TanhTransform: 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 math
import torch.nn.functional as F
from torch import distributions as pyd

torch.manual_seed(0)

class TanhTransform(pyd.transforms.Transform):
    domain = pyd.constraints.real
    codomain = pyd.constraints.interval(-1.0, 1.0)
    bijective = True
    sign = +1

    def __init__(self, cache_size=1):
        super().__init__(cache_size=cache_size)

    @staticmethod
    def atanh(x):
        return 0.5 * (x.log1p() - (-x).log1p())

    def __eq__(self, other):
        return isinstance(other, TanhTransform)

    def _call(self, x):
        return x.tanh()

    def _inverse(self, y):
        return self.atanh(y)

    def log_abs_det_jacobian(self, x, y):
        return 2.0 * (math.log(2.0) - x - F.softplus(-2.0 * x))


ARGS = {
    "cache_size": 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 TanhTransform — 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 TanhTransform 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
TruncatedNormal Ran
ran on a synthesized input · constructed an object; no result was computed
mazpie/mastering-urlb/agent/dreamer.py
code served (permissive licence) · get_code("58845dbc07016005")
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, TruncatedNormal: 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
from torch import distributions as pyd

torch.manual_seed(0)

# Create small tensors for loc and scale
# loc and scale are used in the parent Normal distribution
# They need to be broadcastable and have compatible shapes
loc = torch.randn(2, 4)
scale = torch.ones(2, 4) * 0.5

# Define bounds
low = -1.0
high = 1.0

# Epsilon for clamping
eps = 1e-6

ARGS = {
    'loc': loc,
    'scale': scale,
    'low': low,
    'high': high,
    'eps': eps
}
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 TruncatedNormal — 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 TruncatedNormal 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
lambda_return Ran
ran; our contract draft was wrong, not the code
mazpie/mastering-urlb/agent/dreamer.py
code served (permissive licence) · get_code("65b7e7497e68ef56")
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 3, dtype family float, shape (3, 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

torch.manual_seed(0)

# Create minimal tensor arguments
batch_size = 2
time_steps = 3
feature_dim = 4

reward = torch.randn(time_steps, batch_size, feature_dim)
value = torch.randn(time_steps, batch_size, feature_dim)
pcont = 0.99  # Can be a scalar float
bootstrap = torch.randn(batch_size, feature_dim)
lambda_ = 0.95
axis = 0

ARGS = {
    "reward": reward,
    "value": value,
    "pcont": pcont,
    "bootstrap": bootstrap,
    "lambda_": lambda_,
    "axis": axis,
}
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 3, dtype family float, shape (3, 2, 4) — shape (3, 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
static_scan_for_lambda_return Ran
ran; our contract draft was wrong, not the code
mazpie/mastering-urlb/agent/dreamer.py
code served (permissive licence) · get_code("28027f3ba32cb402")
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 3, dtype family float, shape (5, 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
import torch.nn as nn

torch.manual_seed(0)

# Define a simple function that fn will be
def sample_fn(state, *inputs):
    """
    fn takes a state and variable number of inputs.
    It should return a tensor that can be concatenated.
    """
    # Unpack inputs from the lambda generator
    batch_inputs = list(inputs)
    # Stack them and process
    combined = torch.stack(batch_inputs, dim=0)  # Shape: (num_inputs, 1, dim)
    combined = combined.squeeze(1)  # Shape: (num_inputs, dim)
    
    # Simple operation: sum across inputs and add to state
    result = state + combined.sum(dim=0, keepdim=True)  # Shape: (1, dim)
    return result

# Create input tensors
# inputs should be a list of tensors with shape (sequence_length, batch_size, feature_dim)
# Using distinct dimensions: 5 for sequence length, 2 for batch, 4 for features
input_1 = torch.randn(5, 2, 4)
input_2 = torch.randn(5, 2, 4)
inputs = [input_1, input_2]

# Create start state
# Should be compatible with the output shape of fn
start = torch.randn(1, 4)

# Build ARGS dict
ARGS = {
    'fn': sample_fn,
    'inputs': inputs,
    'start': start
}
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 3, dtype family float, shape (5, 2, 4) — shape (5, 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
stop_gradient Ran
ran, honoured the contract we drafted
mazpie/mastering-urlb/agent/dreamer.py
code served (permissive licence) · get_code("17f4154f5359d704")
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

torch.manual_seed(0)

# The function stop_gradient takes a tensor x and calls x.detach() on it
# x must be a torch.Tensor

x = torch.randn(2, 4)

ARGS = {
    "x": x
}
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
ActorCritic Not yet run
unverified (harvested, no recorded run)
mazpie/mastering-urlb/agent/dreamer.py
code served (permissive licence) · get_code("d0b227b9789fce4a")
How it ran

no invocation recorded.

Decoder Not yet run
unverified (harvested, no recorded run)
mazpie/mastering-urlb/agent/dreamer.py
code served (permissive licence) · get_code("f4840712627e4756")
How it ran

no invocation recorded.

DreamerAgent Not yet run
unverified (harvested, no recorded run)
mazpie/mastering-urlb/agent/dreamer.py
code served (permissive licence) · get_code("e6da38f0e6237df9")
How it ran

no invocation recorded.

MLP Not yet run
unverified (harvested, no recorded run)
mazpie/mastering-urlb/agent/dreamer.py
code served (permissive licence) · get_code("9d13094ceaa00c11")
How it ran

no invocation recorded.

WorldModel Not yet run
unverified (harvested, no recorded run)
mazpie/mastering-urlb/agent/dreamer.py
code served (permissive licence) · get_code("5fb819938bf3f347")
How it ran

no invocation recorded.

hard_update_params Not yet run
unverified (harvested, no recorded run)
mazpie/mastering-urlb/agent/dreamer.py
code served (permissive licence) · get_code("c0833f1665786d11")
How it ran

no invocation recorded.

static_scan Not yet run
unverified (harvested, no recorded run)
mazpie/mastering-urlb/agent/dreamer.py
code served (permissive licence) · get_code("c84e0c6f35ae5bd7")
How it ran

no invocation recorded.

Also present in other papers' code

Of 24 samples lifted from this paper's linked repositories, 24 carry a function identity and 0 do not. Of the 24 distinct functions with an identity, 5 are also present in other papers whose code we hold (3 with at least one paper dated earlier, 0 the same month, 5 later) and 19 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
TanhTransform
brick of this paper's assembly: helper
5 other papers
1 dated earlier, 4 later · 4 only as an assembly brick
2004.13649 (2020-04)
Image Augmentation Is All You Need: Regularizing Deep Reinforcement Learning from Pixels
2210.09598 (2022-10)
Planning for Sample Efficient Imitation Learning
2210.10765 (2022-10)
When to Ask for Help: Proactive Interventions in Autonomous Reinforcement Learning
and 2 more not named here
SquashedNormal
brick of this paper's assembly: helper
4 other papers
1 dated earlier, 3 later · 3 only as an assembly brick
2004.13649 (2020-04)
Image Augmentation Is All You Need: Regularizing Deep Reinforcement Learning from Pixels
2210.09598 (2022-10)
Planning for Sample Efficient Imitation Learning
2210.10765 (2022-10)
When to Ask for Help: Proactive Interventions in Autonomous Reinforcement Learning
and 1 more not named here
RequiresGrad
brick of this paper's assembly: helper
2 other papers
2 later · 2 only as an assembly brick
2301.04104 (2023-01)
Mastering Diverse Domains through World Models
2305.15260 (2023-05)
Making Offline RL Online: Collaborative World Models for Offline Visual Reinforcement Learning
hard_update_params
brick of this paper's assembly: helper
2 other papers
1 dated earlier, 1 later · 2 only as an assembly brick
2103.04551 (2021-03)
Behavior From the Void: Unsupervised Active Pre-Training
2406.16255 (2024-06)
Uncertainty-Aware Reward-Free Exploration with General Function Approximation
stop_gradient
brick of this paper's assembly: helper
2 other papers
2 later
2402.07871 (2024-02)
Scaling Laws for Fine-Grained Mixture of Experts
2406.18043 (2024-06)
GenRL: Multimodal-foundation world models for generalization in embodied agents

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

Controlling artificial agents from visual sensory data is an arduous task. Reinforcement learning (RL) algorithms can succeed but require large amounts of interactions between the agent and the environment. To alleviate the issue, unsupervised RL proposes to employ self-supervised interaction and learning, for adapting faster to future tasks. Yet, as shown in the Unsupervised RL Benchmark (URLB; Laskin et al. ( 2021)), whether current unsupervised strategies can improve generalization capabilities is still unclear, especially in visual control settings. In this work, we study the URLB and propose a new method to solve it, using unsupervised model-based RL, for pre-training the agent, and a task-aware finetuning strategy combined with a new proposed hybrid planner, Dyna-MPC, to adapt the agent for downstream tasks. On URLB, our method obtains 93.59% overall normalized performance, surpassing previous baselines by a staggering margin. The approach is empirically evaluated through a large-scale empirical study, which we use to validate our design choices and analyze our models. We also show robust performance on the Real-Word RL benchmark, hinting at resiliency to environment perturbations during adaptation.

For agents

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

get_harvested_code_for_paper("2209.12016")
get_code_for_paper("2209.12016")
have("2209.12016")

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

curl https://syntology.ai/api/ran/2209.12016.json

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

[![Syntology run record](https://syntology.ai/api/ran/2209.12016.svg)](https://syntology.ai/paper/2209.12016)

Connect an agent — have() is free.