SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2306.09205 · ICLR · 2024

Reward-Free Curricula for Training Robust World Models

Ingmar Posner, Marc Rigter, Minqi Jiang

arXiv · PDF · Open in the Atlas

Code that ran

5 of 7 samples (functions and classes) lifted from this paper's linked repositories executed on a synthesized input in Syntology's jail (1 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; 2 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/2306.09205.json · badge: /api/ran/2306.09205.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
danijar/dreamerv2 canonical (classifier label)
classifier basis: census
4 ran · 1 no recorded run
marc-rigter/waker canonical (classifier label)
classifier basis: classifier_2025
1 ran · 1 no recorded run
FunctionStatusWhere it lives
WAKER Ran
ran on a synthesized input · constructed an object; no result was computed
marc-rigter/waker/waker/waker.py
code served (permissive licence) · get_code("a99d14ad43461b31")
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, WAKER: 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 numpy as np
import tensorflow as tf
import collections

# Create a mock common module with get_dict_key function
class common:
    @staticmethod
    def get_dict_key(env_params):
        """Convert env_params array to a hashable key"""
        return tuple(env_params.astype(float))

# Define ARGS for EnvSampler.__init__
task = "terrain"
interval = 10000
gamma_uncert_reduct = 0.95

ARGS = {
    "task": task,
    "interval": interval,
    "gamma_uncert_reduct": gamma_uncert_reduct
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, tensorflow — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type WAKER — 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 WAKER 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_act Ran
ran; our contract draft was wrong, not the code
danijar/dreamerv2/dreamerv2/common/nets.py
pointer only (licence: MIT); the stored copy get_code serves, from RU-Automated-Reasoning-Group/MUN, is pointer only (licence: NONE) · get_code("8f39adcc1afcac90")
How it ran

The body of this sample is withheld: this paper's link allows the body inline (MIT), but get_code serves the stored copy, first harvested from RU-Automated-Reasoning-Group/MUN, and GitHub detected no licence in that repository 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.

lambda_return Ran
ran on a synthesized input
danijar/dreamerv2/dreamerv2/common/other.py
code served (permissive licence) · get_code("8cd9b6c638b90ac8")
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 EagerTensor, 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 tensorflow as tf

# Set seed for reproducibility
tf.random.set_seed(0)

# Create small tensors with distinct dimensions to probe shapes
batch_size = 2
time_steps = 3
feature_dim = 4

# reward: shape (time_steps, batch_size, feature_dim)
reward = tf.random.normal([time_steps, batch_size, feature_dim])

# value: same shape as reward
value = tf.random.normal([time_steps, batch_size, feature_dim])

# pcont: can be scalar or same shape as reward
pcont = 0.99

# bootstrap: shape matching value[-1], which is (batch_size, feature_dim)
bootstrap = tf.random.normal([batch_size, feature_dim])

# lambda_: scalar between 0 and 1
lambda_ = 0.95

# axis: which axis to transpose (0 in this case, but can be tested with other values)
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
tensorflow — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type EagerTensor, 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
schedule Ran
ran on a synthesized input
danijar/dreamerv2/dreamerv2/common/other.py
code served (permissive licence) · get_code("c2659ae89b97624c")
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 float, 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 tensorflow as tf

# Test case 1: Simple float string
string_1 = "0.001"
step_1 = tf.constant(100, dtype=tf.int32)

# Test case 2: Linear schedule
string_2 = "linear(0.1,0.01,1000)"
step_2 = tf.constant(500, dtype=tf.int32)

# Test case 3: Warmup schedule
string_3 = "warmup(100,0.1)"
step_3 = tf.constant(50, dtype=tf.int32)

# Test case 4: Exponential schedule
string_4 = "exp(0.1,0.001,500)"
step_4 = tf.constant(250, dtype=tf.int32)

# Test case 5: Horizon schedule
string_5 = "horizon(10,100,1000)"
step_5 = tf.constant(500, dtype=tf.int32)

ARGS = {
    "string": string_1,
    "step": step_1,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
tensorflow — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type float, 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
static_scan Ran
ran on a synthesized input
danijar/dreamerv2/dreamerv2/common/other.py
code served (permissive licence) · get_code("77ff0f8ba58bad0e")
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 EagerTensor, 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 tensorflow as tf

# Define a simple function that takes (state, input) and returns new state
def simple_fn(state, inp):
    # state and inp are both single tensors
    # This function just adds them together and returns a new state
    return state + inp

# Create small test tensors
# inputs should have shape (sequence_length, feature_dim)
sequence_length = 2
feature_dim = 4
inputs = tf.constant([[1.0, 2.0, 3.0, 4.0], [5.0, 6.0, 7.0, 8.0]], dtype=tf.float32)

# start is the initial state, should have same shape as individual outputs
start = tf.constant([0.0, 0.0, 0.0, 0.0], dtype=tf.float32)

ARGS = {
    "fn": simple_fn,
    "inputs": inputs,
    "start": start,
    "reverse": False,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
tensorflow — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type EagerTensor, 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
EnvSampler Not yet run
unverified (harvested, no recorded run)
marc-rigter/waker/waker/waker.py
code served (permissive licence) · get_code("040c8b472721aadf")
How it ran

no invocation recorded.

encode_gif Not yet run
unverified (harvested, no recorded run)
danijar/dreamerv2/dreamerv2/common/logger.py
code served (permissive licence) · get_code("8c937e50ed534047")
How it ran

no invocation recorded.

Also present in other papers' code

Of 7 samples lifted from this paper's linked repositories, 7 carry a function identity and 0 do not. Of the 7 distinct functions with an identity, 4 are also present in other papers whose code we hold (3 with at least one paper dated earlier, 0 the same month, 1 later) and 3 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
get_act 2 other papers
2 later
2308.12270 (2023-08)
Language Reward Modulation for Pretraining Reinforcement Learning
2411.02446 (2024-11)
Learning World Models for Unconstrained Goal Navigation
schedule 2 other papers
2 dated earlier
2010.02193 (2020-10)
Mastering Atari with Discrete World Models
2110.14565 (2021-10)
DreamerPro: Reconstruction-Free Model-Based Reinforcement Learning with Prototypical Representations
static_scan 2 other papers
2 dated earlier
2010.02193 (2020-10)
Mastering Atari with Discrete World Models
2110.14565 (2021-10)
DreamerPro: Reconstruction-Free Model-Based Reinforcement Learning with Prototypical Representations
lambda_return 1 other paper
1 dated earlier
2110.14565 (2021-10)
DreamerPro: Reconstruction-Free Model-Based Reinforcement Learning with Prototypical Representations

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

There has been a recent surge of interest in developing generally-capable agents that can adapt to new tasks without additional training in the environment. Learning world models from reward-free exploration is a promising approach, and enables policies to be trained using imagined experience for new tasks. However, achieving a general agent requires robustness across different environments. In this work, we address the novel problem of generating curricula in the reward-free setting to train robust world models. We consider robustness in terms of minimax regret over all environment instantiations and show that the minimax regret can be connected to minimising the maximum error in the world model across environment instances. This result informs our algorithm, WAKER: Weighted Acquisition of Knowledge across Environments for Robustness. WAKER selects environments for data collection based on the estimated error of the world model for each environment. Our experiments demonstrate that WAKER outperforms several baselines, resulting in improved robustness, efficiency, and generalisation.

For agents

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

get_harvested_code_for_paper("2306.09205")
get_code_for_paper("2306.09205")
have("2306.09205")

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

curl https://syntology.ai/api/ran/2306.09205.json

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

[![Syntology run record](https://syntology.ai/api/ran/2306.09205.svg)](https://syntology.ai/paper/2306.09205)

Connect an agent — have() is free.