SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2405.04342 · ICLR · 2024

The Curse of Diversity in Ensemble-Based Exploration

Aaron Courville, Zhixuan Lin, Evgenii Nikishin, Pierluca Oro

arXiv · PDF · Open in the Atlas

Code that ran

3 of 16 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; 3 honoured the contract we drafted, 0 violated it and 0 have no contract verdict; 0 failures of our instrument; 13 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/2405.04342.json · badge: /api/ran/2405.04342.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
zhixuan-lin/ensemble-rl-discrete canonical (classifier label)
classifier basis: census
2 ran · 7 no recorded run
zhixuan-lin/ensemble-rl-continuous canonical (classifier label)
classifier basis: classifier_2025
1 ran · 6 no recorded run
FunctionStatusWhere it lives
get_gammas Ran
ran, honoured the contract we drafted
zhixuan-lin/ensemble-rl-discrete/ensemble_rl/agents/boot_dqn_agent.py
code served (permissive licence) · get_code("ee6d2e618cac54d1")
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 ArrayImpl, ndim 2, dtype family float, shape (3, 10) 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 jax.numpy as jnp

ARGS = {
    "gamma_max": 0.99,
    "num_gammas": 10,
    "ensemble_size": 3,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
jax — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type ArrayImpl, ndim 2, dtype family float, shape (3, 10) — shape (3, 10) 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_param_count Ran
ran, honoured the contract we drafted
zhixuan-lin/ensemble-rl-continuous/jaxrl/agents/cel/cel_learner.py
code served (permissive licence) · get_code("2ec107672ef4af91")
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 ArrayImpl, 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 jax
import jax.numpy as jnp
from jax import tree_util

# The function get_param_count expects a pytree of parameters
# It uses jax.tree_util.tree_leaves to flatten the tree and get all leaves
# Then it computes the product of shapes for each leaf and sums them

# Create a simple pytree structure using nested dicts and JAX arrays
params = {
    'layer1': {
        'weight': jnp.ones((4, 3)),
        'bias': jnp.ones((3,)),
    },
    'layer2': {
        'weight': jnp.ones((3, 2)),
        'bias': jnp.ones((2,)),
    },
    'scalar': jnp.ones(()),
}

ARGS = {
    'params': params
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
jax — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type ArrayImpl, 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
squarify Ran
ran, honoured the contract we drafted
zhixuan-lin/ensemble-rl-discrete/ensemble_rl/metric_utils.py
code served (permissive licence) · get_code("66a8c225827dbcfc")
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 ArrayImpl, ndim 3, dtype family float, shape (2, 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 jax.numpy as jnp

# Test case 1: 1D input (batch_size only)
x_1d = jnp.array([1.0, 2.0, 3.0])

# Test case 2: 2D input (batch_size, representation_dim)
x_2d = jnp.array([[1.0, 2.0, 3.0, 4.0],
                   [5.0, 6.0, 7.0, 8.0]])

ARGS = {
    "x": x_2d
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
jax — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type ArrayImpl, ndim 3, dtype family float, shape (2, 2, 4) — shape (2, 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
CELLearner Not yet run
unverified (harvested, no recorded run)
zhixuan-lin/ensemble-rl-continuous/jaxrl/agents/cel/cel_learner.py
code served (permissive licence) · get_code("373f13eb372c6f1c")
How it ran

no invocation recorded.

Model Not yet run
unverified (harvested, no recorded run)
zhixuan-lin/ensemble-rl-continuous/jaxrl/agents/cel/cel_learner.py
code served (permissive licence) · get_code("a681d1d465ca2192")
How it ran

no invocation recorded.

_sample_actions Not yet run
unverified (harvested, no recorded run)
zhixuan-lin/ensemble-rl-continuous/jaxrl/agents/cel/cel_learner.py
code served (permissive licence) · get_code("8078c02fd29a891c")
How it ran

no invocation recorded.

_sample_index Not yet run
unverified (harvested, no recorded run)
zhixuan-lin/ensemble-rl-continuous/jaxrl/agents/cel/cel_learner.py
code served (permissive licence) · get_code("e4e6020a9a6a6bbb")
How it ran

no invocation recorded.

_update_jit Not yet run
unverified (harvested, no recorded run)
zhixuan-lin/ensemble-rl-continuous/jaxrl/agents/cel/cel_learner.py
code served (permissive licence) · get_code("61fd4828e83d0558")
How it ran

no invocation recorded.

cosine_distance Not yet run
unverified (harvested, no recorded run)
zhixuan-lin/ensemble-rl-discrete/ensemble_rl/metric_utils.py
code served (permissive licence) · get_code("389e520afe20c705")
How it ran

no invocation recorded.

l2 Not yet run
unverified (harvested, no recorded run)
zhixuan-lin/ensemble-rl-discrete/ensemble_rl/metric_utils.py
code served (permissive licence) · get_code("9ebd5c538050adfc")
How it ran

no invocation recorded.

select_action Not yet run
unverified (harvested, no recorded run)
zhixuan-lin/ensemble-rl-discrete/ensemble_rl/agents/boot_quantile_agent.py
code served (permissive licence) · get_code("5eff8223ef2a7296")
How it ran

no invocation recorded.

target_distribution Not yet run
unverified (harvested, no recorded run)
zhixuan-lin/ensemble-rl-discrete/ensemble_rl/agents/boot_quantile_agent.py
code served (permissive licence) · get_code("4e8df17f4dcd90d6")
How it ran

no invocation recorded.

target_q Not yet run
unverified (harvested, no recorded run)
zhixuan-lin/ensemble-rl-discrete/ensemble_rl/agents/boot_dqn_agent.py
code served (permissive licence) · get_code("9697728fcd0cfeb3")
How it ran

no invocation recorded.

target_q Not yet run
unverified (harvested, no recorded run)
zhixuan-lin/ensemble-rl-discrete/ensemble_rl/agents/dqn_agent.py
code served (permissive licence) · get_code("336b5ba1af3f1cd7")
How it ran

no invocation recorded.

target_update Not yet run
unverified (harvested, no recorded run)
zhixuan-lin/ensemble-rl-continuous/jaxrl/agents/cel/cel_learner.py
code served (permissive licence) · get_code("d7b5436efe6a474c")
How it ran

no invocation recorded.

train Not yet run
unverified (harvested, no recorded run)
zhixuan-lin/ensemble-rl-discrete/ensemble_rl/agents/boot_quantile_agent.py
code served (permissive licence) · get_code("fed787981f691a19")
How it ran

no invocation recorded.

Also present in other papers' code

Of 16 samples lifted from this paper's linked repositories, 16 carry a function identity and 0 do not. Of the 16 distinct functions with an identity, 1 are also present in other papers whose code we hold (1 with at least one paper dated earlier, 1 the same month, 1 later) and 15 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
target_update
brick of this paper's assembly: update_rule
3 other papers
1 dated earlier, 1 the same month, 1 later
2310.17966 (2023-10)
Train Once, Get a Family: State-Adaptive Balances for Offline-to-Online Reinforcement Learning
2405.16158 (2024-05)
Bigger, Regularized, Optimistic: scaling for compute and sample-efficient continuous control
2605.11711 (2026-05)
Debiased Model-based Representations for Sample-efficient Continuous Control

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

We uncover a surprising phenomenon in deep reinforcement learning: training a diverse ensemble of data-sharing agents -a well-established exploration strategycan significantly impair the performance of the individual ensemble members when compared to standard single-agent training. Through careful analysis, we attribute the degradation in performance to the low proportion of self-generated data in the shared training data for each ensemble member, as well as the inefficiency of the individual ensemble members to learn from such highly off-policy data. We thus name this phenomenon the curse of diversity. We find that several intuitive solutions -such as a larger replay buffer or a smaller ensemble size -either fail to consistently mitigate the performance loss or undermine the advantages of ensembling. Finally, we demonstrate the potential of representation learning to counteract the curse of diversity with a novel method named Cross-Ensemble Representation Learning (CERL) in both discrete and continuous control domains. Our work offers valuable insights into an unexpected pitfall in ensemble-based exploration and raises important caveats for future applications of similar approaches.

For agents

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

get_harvested_code_for_paper("2405.04342")
get_code_for_paper("2405.04342")
have("2405.04342")

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

curl https://syntology.ai/api/ran/2405.04342.json

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

[![Syntology run record](https://syntology.ai/api/ran/2405.04342.svg)](https://syntology.ai/paper/2405.04342)

Connect an agent — have() is free.