SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2502.15280 · ICML · 2025

Hyperspherical Normalization for Scalable Deep Reinforcement Learning

Peter Stone, Jaegul Choo, Hojoon Lee, Donghu Kim, Takuma Seno, Youngdo Lee

arXiv · PDF · Open in the Atlas

Code that ran

4 of 6 samples (functions and classes) lifted from this paper's linked repositories executed on a synthesized input in Syntology's jail (2 of them constructed an object rather than computing a result; 1 honoured the contract we drafted, 0 violated it and 2 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/2502.15280.json · badge: /api/ran/2502.15280.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
naumix/BiggerRegularizedOptimistic community
Papers with Code archive
3 ran · 1 no recorded run
SonyResearch/simba community
Papers with Code archive
1 ran · 1 no recorded run
FunctionStatusWhere it lives
Model Ran
ran on a synthesized input · constructed an object; no result was computed
naumix/BiggerRegularizedOptimistic/jaxrl/bro/actor.py
code served (permissive licence) · get_code("f73fa4e26e899f5b")
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, Model: 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 jax
import jax.numpy as jnp
import flax
import flax.linen as nn
import optax
from typing import Any, Dict, Optional, Sequence

# Define a minimal nn.Module for testing
class SimpleModel(nn.Module):
    features: int = 4
    
    @nn.compact
    def __call__(self, x):
        x = nn.Dense(self.features)(x)
        return x

# Create a simple model definition
model_def = SimpleModel(features=4)

# Create dummy input for initialization
# The init method expects (rng, *inputs) where inputs are example arrays
rng = jax.random.PRNGKey(0)
dummy_input = jnp.ones((2, 3))  # batch_size=2, input_dim=3

# Create the optimizer
tx = optax.adam(learning_rate=0.001)

# Initialize the model to get params and opt_state
variables = model_def.init(rng, dummy_input)
params = variables['params']
opt_state = tx.init(params)

# Build the ARGS dictionary
ARGS = {
    'step': 1,
    'apply_fn': model_def,
    'params': params,
    'tx': tx,
    'opt_state': opt_state
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
flax, jax, optax — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Model — 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 Model 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
SaveState Ran
ran on a synthesized input · constructed an object; no result was computed
naumix/BiggerRegularizedOptimistic/jaxrl/bro/actor.py
code served (permissive licence) · get_code("95538ae49008d867")
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, SaveState: 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 flax
import optax
import jax
import jax.numpy as jnp

# Create a minimal Params (FrozenDict structure)
# params should be a FrozenDict mapping parameter names to values
params_dict = {
    "layer1": jnp.array([[1.0, 2.0, 3.0, 4.0],
                         [5.0, 6.0, 7.0, 8.0]]),
    "layer2": jnp.array([0.1, 0.2, 0.3, 0.4])
}
params = flax.core.FrozenDict(params_dict)

# Create a minimal OptState
# For this, we need to create an optimizer and its state
# Using a simple Adam optimizer as an example
optimizer = optax.adam(learning_rate=0.001)

# Initialize opt_state by computing it from params
# We need to flatten the params to compute opt_state
flat_params = jax.tree_util.tree_leaves(params)
opt_state = optimizer.init(params)

ARGS = {
    "params": params,
    "opt_state": opt_state
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
flax, jax, optax — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type SaveState — 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 SaveState 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
orthogonal_init Ran
ran; our contract draft was wrong, not the code
SonyResearch/simba/scale_rl/networks/layers.py
code served (permissive licence) · get_code("5ac49b3249fc3ab0")
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) and returned function. 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 jax.numpy as jnp
import flax.linen as nn

ARGS = {
    "scale": jnp.sqrt(2.0)
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1
Imports
flax, jax — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type function — a function output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a function 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
tree_norm Ran
ran, honoured the contract we drafted
naumix/BiggerRegularizedOptimistic/jaxrl/bro/actor.py
code served (permissive licence) · get_code("f325964f1797f651")
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

# The function tree_norm takes a tree structure (which can be nested dicts, lists, tuples, etc.)
# containing arrays and computes the norm by:
# 1. Extracting all leaves using jax.tree_util.tree_leaves()
# 2. Squaring each leaf, summing their elements
# 3. Summing all those values
# 4. Taking the square root

# We need to construct a tree with some JAX arrays as leaves
tree = {
    'layer1': jnp.array([[1.0, 2.0, 3.0, 4.0],
                         [5.0, 6.0, 7.0, 8.0]]),
    'layer2': {
        'weights': jnp.array([[0.5, 1.5],
                              [2.5, 3.5],
                              [4.5, 5.5],
                              [6.5, 7.5]]),
        'bias': jnp.array([0.1, 0.2, 0.3])
    },
    'layer3': jnp.array([1.0, 2.0])
}

ARGS = {
    'tree': tree
}
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
MLPBlock Not yet run
unverified (harvested, no recorded run)
SonyResearch/simba/scale_rl/networks/layers.py
code served (permissive licence) · get_code("6b074234d267bb66")
How it ran

no invocation recorded.

update_optimistic Not yet run
unverified (harvested, no recorded run)
naumix/BiggerRegularizedOptimistic/jaxrl/bro/actor.py
code served (permissive licence) · get_code("9cc22255440aae4f")
How it ran

no invocation recorded.

Also present in other papers' code

Of 6 samples lifted from this paper's linked repositories, 6 carry a function identity and 0 do not. Of the 6 distinct functions with an identity, 3 are also present in other papers whose code we hold (3 with at least one paper dated earlier, 0 the same month, 3 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
tree_norm
brick of this paper's assembly: helper
5 other papers
2 dated earlier, 3 later · 4 only as an assembly brick
2405.04342 (2024-05)
The Curse of Diversity in Ensemble-Based Exploration
2405.16158 (2024-05)
Bigger, Regularized, Optimistic: scaling for compute and sample-efficient continuous control
2506.05968 (2025-06)
Gradual Transition from Bellman Optimality Operator to Bellman Operator in Online Reinforcement Learning
and 2 more not named here
SaveState
brick of this paper's assembly: helper
4 other papers
1 dated earlier, 3 later · 2 only as an assembly brick
2405.16158 (2024-05)
Bigger, Regularized, Optimistic: scaling for compute and sample-efficient continuous control
2506.05968 (2025-06)
Gradual Transition from Bellman Optimality Operator to Bellman Operator in Online Reinforcement Learning
2603.05066 (2026-03)
Reward-Conditioned Reinforcement Learning
and 1 more not named here
Model
brick of this paper's assembly: helper
2 other papers
1 dated earlier, 1 later
2405.16158 (2024-05)
Bigger, Regularized, Optimistic: scaling for compute and sample-efficient continuous control
2506.05968 (2025-06)
Gradual Transition from Bellman Optimality Operator to Bellman Operator in Online 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

Scaling up the model size and computation has brought consistent performance improvements in supervised learning. However, this lesson often fails to apply to reinforcement learning (RL) because training the model on non-stationary data easily leads to overfitting and unstable optimization. In response, we introduce SimbaV2, a novel RL architecture designed to stabilize optimization by (i) constraining the growth of weight and feature norm by hyperspherical normalization; and (ii) using a distributional value estimation with reward scaling to maintain stable gradients under varying reward magnitudes. Using the soft actor-critic as a base algorithm, SimbaV2 scales up effectively with larger models and greater compute, achieving state-of-the-art performance on 57 continuous control tasks across 4 domains. The code is available at dojeon-ai.github.io/SimbaV2.

For agents

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

get_harvested_code_for_paper("2502.15280")
get_code_for_paper("2502.15280")
have("2502.15280")

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

curl https://syntology.ai/api/ran/2502.15280.json

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

[![Syntology run record](https://syntology.ai/api/ran/2502.15280.svg)](https://syntology.ai/paper/2502.15280)

Connect an agent — have() is free.