SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2404.18508 · 2024

Scalable Event-by-event Processing of Neuromorphic Sensory Signals With Deep State-Space Models

arXiv · PDF · Open in the Atlas

Code that ran

15 of 18 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; 1 honoured the contract we drafted, 3 violated it and 1 has no contract verdict; 10 failures of our instrument; 3 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/2404.18508.json · badge: /api/ran/2404.18508.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
Efficient-Scalable-Machine-Learning/event-ssm canonical (classifier label)
classifier basis: census
15 ran · 3 no recorded run
FunctionStatusWhere it lives
cut_mix_augmentation Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
Efficient-Scalable-Machine-Learning/event-ssm/event_ssm/transform.py
code served (permissive licence) · get_code("801be17d5a9e3d5d")
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 tuple. 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

# Create structured arrays to represent event streams
# Events have fields: 't' (timestamp), and 3 other fields (x, y, p for typical event cameras)
def create_event_stream(num_events):
    """Create a structured array with event data"""
    events = np.zeros(num_events, dtype=[('t', np.float32), ('x', np.int16), ('y', np.int16), ('p', np.int8)])
    events['t'] = np.sort(np.random.rand(num_events) * 1000)  # timestamps in range [0, 1000)
    events['x'] = np.random.randint(0, 640, num_events)
    events['y'] = np.random.randint(0, 480, num_events)
    events['p'] = np.random.randint(0, 2, num_events)
    return events

# Create batch of event streams
batch_size = 2
events = [create_event_stream(50) for _ in range(batch_size)]
events_list = [create_event_stream(50) for _ in range(batch_size)]

# Create targets (class labels or one-hot vectors)
targets = np.array([0.0, 1.0])

ARGS = {
    "events": events_list,
    "targets": targets,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type tuple — a tuple output has no array shape of its own; none is 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
cut_mix_augmentation_time Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
Efficient-Scalable-Machine-Learning/event-ssm/event_ssm/transform.py
code served (permissive licence) · get_code("878ef192b1e8192e")
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 tuple. 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

# Create structured arrays for events with 't' field and other event data
def create_event_stream(num_events, seed):
    np.random.seed(seed)
    # Create structured array with 't' (time) field and other fields
    events = np.zeros(num_events, dtype=[('t', np.float32), ('x', np.int16), ('y', np.int16), ('p', np.int8)])
    events['t'] = np.sort(np.random.uniform(0, 100, num_events)).astype(np.float32)
    events['x'] = np.random.randint(0, 64, num_events)
    events['y'] = np.random.randint(0, 64, num_events)
    events['p'] = np.random.randint(0, 2, num_events)
    return events

# Create batch of event streams
batch_size = 2
num_classes = 5

events = [
    create_event_stream(50, seed=0),
    create_event_stream(60, seed=1)
]

# Create targets as one-hot or probability distributions
targets = np.array([
    [1.0, 0.0, 0.0, 0.0, 0.0],
    [0.0, 0.0, 1.0, 0.0, 0.0]
], dtype=np.float32)

ARGS = {
    'events': events,
    'targets': targets
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type tuple — a tuple output has no array shape of its own; none is 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
discretize_dirac Ran
ran; our contract draft was wrong, not the code
Efficient-Scalable-Machine-Learning/event-ssm/event_ssm/ssm.py
code served (permissive licence) · get_code("9e888aab5e750287")
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 tuple. 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 np

# Construct arguments for discretize_dirac
# Based on the function signature and docstring:
# - Lambda: diagonal state matrix (P,) - complex valued
# - step_delta: discretization step sizes (P,) - float
# - time_delta: discretization step size (float32) - scalar

P = 4  # state dimension

# Lambda: diagonal state matrix, shape (P,), typically complex
Lambda = np.array([-1.0 + 0.5j, -2.0 + 1.0j, -0.5 + 2.0j, -3.0 + 0.1j], dtype=np.complex64)

# step_delta: discretization step sizes, shape (P,), float
step_delta = np.array([0.1, 0.05, 0.2, 0.15], dtype=np.float32)

# time_delta: scalar discretization step size, float32
time_delta = np.float32(0.01)

ARGS = {
    "Lambda": Lambda,
    "step_delta": step_delta,
    "time_delta": time_delta
}
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 tuple — a tuple output has no array shape of its own; none is 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
discretize_zoh Ran
ran, violated the contract we drafted
Efficient-Scalable-Machine-Learning/event-ssm/event_ssm/ssm.py
code served (permissive licence) · get_code("bb3aeee6201c06dd")
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 tuple, ndim 2, dtype family float; no output shape is stated. The image it ran in was not recorded. If your copy returns the same type and ndim 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 np
import numpy as onp

# Set seed for reproducibility
onp.random.seed(0)

# Lambda: diagonal state matrix (P,)
# Should be complex-valued for continuous-time SSM
P = 4
Lambda = np.array(onp.random.randn(P) + 1j * onp.random.randn(P), dtype=np.complex64)

# step_delta: discretization step sizes (P,)
# Should be positive real values
step_delta = np.array(onp.random.rand(P) + 0.1, dtype=np.float32)

# time_delta: (float32) discretization step size (scalar)
time_delta = np.float32(0.01)

ARGS = {
    "Lambda": Lambda,
    "step_delta": step_delta,
    "time_delta": time_delta,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
jax, numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type tuple, ndim 2, dtype family float — a tuple output has no array shape of its own; none is stated
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
evaluation_step Ran
ran; our contract draft was wrong, not the code
Efficient-Scalable-Machine-Learning/event-ssm/event_ssm/train_utils.py
code served (permissive licence) · get_code("7905962f8670b1da")
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 tuple, ndim 1, dtype family float; no output shape is stated. The image it ran in was not recorded. If your copy returns the same type and ndim 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 jaxtyping import Array
from typing import Any, Dict
from flax.training import train_state
from flax import linen as nn
import optax

# Define a minimal Flax model
class SimpleModel(nn.Module):
    @nn.compact
    def __call__(self, inputs, integration_timesteps, lengths, training):
        x = nn.Dense(8)(inputs)
        x = nn.relu(x)
        x = nn.Dense(4)(x)  # 4 output classes
        return x

# Create an extended TrainState with key and model_state
class TrainState(train_state.TrainState):
    key: Array
    model_state: Dict

# Initialize model and parameters
model = SimpleModel()
key = jax.random.PRNGKey(0)
dummy_inputs = jnp.ones((2, 3))  # batch_size=2, input_dim=3
dummy_timesteps = jnp.ones((2, 1))
dummy_lengths = jnp.ones((2,))
params = model.init(key, dummy_inputs, dummy_timesteps, dummy_lengths, False)['params']

# Create optimizer and optimizer state
optimizer = optax.adam(learning_rate=0.001)
opt_state = optimizer.init(params)

# Create TrainState instance
train_state_obj = TrainState(
    step=0,
    apply_fn=model.apply,
    params=params,
    tx=optimizer,
    opt_state=opt_state,
    key=key,
    model_state={}
)

# Create batch: [inputs, targets, integration_timesteps, lengths]
batch_inputs = jnp.ones((2, 3))  # batch_size=2, input_dim=3
batch_targets = jnp.array([[1.0, 0.0, 0.0, 0.0], [0.0, 1.0, 0.0, 0.0]])  # batch_size=2, num_classes=4 (one-hot)
batch_timesteps = jnp.ones((2, 1))
batch_lengths = jnp.array([2, 2])
batch = (batch_inputs, batch_targets, batch_timesteps, batch_lengths)

ARGS = {
    'train_state': train_state_obj,
    'batch': batch,
    'distributed': False
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
flax, jax, jaxtyping, optax — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type tuple, ndim 1, dtype family float — a tuple output has no array shape of its own; none is stated
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
evaluation_step Ran
ran; our contract draft was wrong, not the code
Efficient-Scalable-Machine-Learning/event-ssm/run_evaluation.py
code served (permissive licence) · get_code("1b34fdae353a499d")
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 tuple. 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 optax

# Set seed for reproducibility
jax.config.update('jax_platform_name', 'cpu')

# Create a simple apply_fn that mimics a neural network forward pass
def dummy_apply_fn(state_dict, inputs, integration_timesteps, lengths, training):
    """
    Dummy apply function that takes inputs and returns logits.
    state_dict: dict with 'params' and other model state
    inputs: shape (batch_size, input_dim)
    integration_timesteps: shape (batch_size,)
    lengths: shape (batch_size,)
    training: bool
    Returns logits of shape (batch_size, num_classes)
    """
    batch_size = inputs.shape[0]
    num_classes = 5
    # Simple dummy computation: just return random logits
    logits = jnp.ones((batch_size, num_classes))
    return logits

# Create params (dummy parameters)
params = {
    'w': jnp.ones((4, 5)),
    'b': jnp.zeros(5)
}

# Create model_state (additional state like batch norm stats, etc.)
model_state = {
    'batch_stats': {'mean': jnp.zeros(4), 'var': jnp.ones(4)}
}

# Create a batch with the required structure
batch_size = 2
input_dim = 4
num_classes = 5

inputs = jnp.ones((batch_size, input_dim))
targets = jax.nn.one_hot(jnp.array([0, 1]), num_classes)  # One-hot encoded targets
integration_timesteps = jnp.array([10, 20])
lengths = jnp.array([15, 25])

batch = (inputs, targets, integration_timesteps, lengths)

ARGS = {
    'apply_fn': dummy_apply_fn,
    'params': params,
    'model_state': model_state,
    'batch': batch
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
jax, optax — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type tuple — a tuple output has no array shape of its own; none is 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
event_stream_dataloader Ran
ran on a synthesized input
Efficient-Scalable-Machine-Learning/event-ssm/event_ssm/dataloading.py
code served (permissive licence) · get_code("e0293591386f80aa")
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 tuple, ndim 1, dtype family float; no output shape is stated. The image it ran in was not recorded. If your copy returns the same type and ndim 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.utils.data import Dataset

torch.manual_seed(0)

class SimpleDataset(Dataset):
    """Minimal dataset for testing"""
    def __init__(self, size=8, seq_len=5, feature_dim=4):
        self.size = size
        self.seq_len = seq_len
        self.feature_dim = feature_dim
    
    def __len__(self):
        return self.size
    
    def __getitem__(self, idx):
        x = torch.randn(self.seq_len, self.feature_dim)
        y = torch.tensor(idx % 2, dtype=torch.long)
        return (x, y)

def train_collate_fn(batch):
    """Collate function for training"""
    xs = torch.stack([item[0] for item in batch])
    ys = torch.stack([item[1] for item in batch])
    return (xs, ys)

def eval_collate_fn(batch):
    """Collate function for evaluation"""
    xs = torch.stack([item[0] for item in batch])
    ys = torch.stack([item[1] for item in batch])
    return (xs, ys)

train_dataset = SimpleDataset(size=16, seq_len=5, feature_dim=4)
val_dataset = SimpleDataset(size=8, seq_len=5, feature_dim=4)
test_dataset = SimpleDataset(size=8, seq_len=5, feature_dim=4)

rng = torch.Generator()
rng.manual_seed(0)

ARGS = {
    'train_data': train_dataset,
    'val_data': val_dataset,
    'test_data': test_dataset,
    'batch_size': 2,
    'eval_batch_size': 4,
    'train_collate_fn': train_collate_fn,
    'eval_collate_fn': eval_collate_fn,
    'rng': rng,
    'num_workers': 0,
    'shuffle_training': True,
}
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 tuple, ndim 1, dtype family float — a tuple output has no array shape of its own; none is stated
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
make_DPLR_HiPPO Ran
ran; our contract draft was wrong, not the code
Efficient-Scalable-Machine-Learning/event-ssm/event_ssm/ssm_init.py
code served (permissive licence) · get_code("188626e586495ec6")
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 tuple. 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 np
from jax.numpy.linalg import eigh

ARGS = {
    "N": 4,
}
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 tuple — a tuple output has no array shape of its own; none is 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
make_HiPPO Ran
ran, honoured the contract we drafted
Efficient-Scalable-Machine-Learning/event-ssm/event_ssm/ssm_init.py
code served (permissive licence) · get_code("27e9e05de680bfb5")
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 (5, 5) 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 = {
    "N": 5,
}
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 (5, 5) — shape (5, 5) 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
make_NPLR_HiPPO Ran
ran; our contract draft was wrong, not the code
Efficient-Scalable-Machine-Learning/event-ssm/event_ssm/ssm_init.py
code served (permissive licence) · get_code("17177c94dcd9bfbf")
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 tuple. 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 np

ARGS = {
    "N": 8,
}
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 tuple — a tuple output has no array shape of its own; none is 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
map_nested_fn Ran
ran; our contract draft was wrong, not the code
Efficient-Scalable-Machine-Learning/event-ssm/event_ssm/train_utils.py
code served (permissive licence) · get_code("50b98382b2de060e")
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 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 torch
import torch.nn as nn

torch.manual_seed(0)

def sample_fn(k, v):
    """
    A simple function that takes a key and value and returns a transformed value.
    This will be passed to map_nested_fn.
    """
    if isinstance(v, (int, float)):
        return v * 2
    elif isinstance(v, str):
        return v.upper()
    else:
        return v

nested_dict = {
    "layer1": {
        "weight": 5,
        "bias": 3.14,
        "name": "first"
    },
    "layer2": {
        "weight": 10,
        "bias": 2.71,
        "config": {
            "enabled": True,
            "rate": 0.5
        }
    },
    "scalar": 42
}

ARGS = {
    "fn": sample_fn
}
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 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
masked_meanpool Ran
ran, violated the contract we drafted
Efficient-Scalable-Machine-Learning/event-ssm/event_ssm/seq_model.py
code served (permissive licence) · get_code("f2a2429ab2b145a5")
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 1, dtype family float, shape (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
import numpy as np

# Construct tiny test data
# x: input sequence of shape (L, d_model) where L is sequence length, d_model is feature dimension
# lengths: original length of sequence before padding (scalar)

L = 5  # sequence length
d_model = 4  # feature dimension

# Create a small input sequence as JAX array
x = jnp.array(np.random.randn(L, d_model).astype(np.float32))

# Create lengths - the actual unpadded length (must be <= L)
lengths = 3  # Only first 3 positions are real data, rest is padding

ARGS = {
    "x": x,
    "lengths": lengths,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
jax, numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type ArrayImpl, ndim 1, dtype family float, shape (4,) — shape (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
reshape_array_per_device Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
Efficient-Scalable-Machine-Learning/event-ssm/event_ssm/trainer.py
code served (permissive licence) · get_code("d0d2c7cc8a0e8597")
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 4, dtype family float, shape (2, 2, 3, 5) 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 numpy as np

torch.manual_seed(0)

# Create a small tensor with batch size divisible by num_devices
# Using batch_size=4, num_devices=2 so 4 % 2 == 0
# Adding feature dimensions to test reshaping with x.shape[1:]
x = torch.randn(4, 3, 5)  # batch=4, feature1=3, feature2=5
num_devices = 2

ARGS = {
    'x': x,
    'num_devices': num_devices
}
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 Tensor, ndim 4, dtype family float, shape (2, 2, 3, 5) — shape (2, 2, 3, 5) 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
reshape_batch_per_device Ran
ran; our contract draft was wrong, not the code
Efficient-Scalable-Machine-Learning/event-ssm/event_ssm/trainer.py
code served (permissive licence) · get_code("0512fd084f8551e4")
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 dict. 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 torch

torch.manual_seed(0)

# Create a simple pytree structure with JAX arrays
# The function uses tree_map to apply reshape_array_per_device to each leaf
x_data = {
    'array1': jnp.ones((8, 4, 3)),  # batch_size=8, will be split into 4 devices
    'array2': jnp.ones((8, 5, 2)),  # batch_size=8, will be split into 4 devices
}

num_devices = 4

ARGS = {
    'x': x_data,
    'num_devices': num_devices,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
jax, torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type dict — a dict output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a dict 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
timepool Ran
ran, violated the contract we drafted
Efficient-Scalable-Machine-Learning/event-ssm/event_ssm/seq_model.py
code served (permissive licence) · get_code("b03910f5f285bbc7")
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 1, dtype family float, shape (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 np

# Create small test tensors with distinct dimensions
# x: sequence of length L=5, embedding dimension d_model=4
L = 5
d_model = 4
x = np.ones((L, d_model)) * 0.5

# integration_timesteps: shape (L,) - one timestep per sequence position
integration_timesteps = np.array([0.1, 0.2, 0.15, 0.25, 0.3])

ARGS = {
    "x": x,
    "integration_timesteps": integration_timesteps,
}
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 1, dtype family float, shape (4,) — shape (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
discretize_async Not yet run
unverified (harvested, no recorded run)
Efficient-Scalable-Machine-Learning/event-ssm/event_ssm/ssm.py
code served (permissive licence) · get_code("b84da31f714379cb")
How it ran

no invocation recorded.

masked_timepool Not yet run
unverified (harvested, no recorded run)
Efficient-Scalable-Machine-Learning/event-ssm/event_ssm/seq_model.py
code served (permissive licence) · get_code("de3055769b640635")
How it ran

no invocation recorded.

training_step Not yet run
unverified (harvested, no recorded run)
Efficient-Scalable-Machine-Learning/event-ssm/event_ssm/train_utils.py
code served (permissive licence) · get_code("5f8147f4c2b3b6c4")
How it ran

no invocation recorded.

Also present in other papers' code

Of 18 samples lifted from this paper's linked repositories, 18 carry a function identity and 0 do not. Of the 18 distinct functions with an identity, 5 are also present in other papers whose code we hold (5 with at least one paper dated earlier, 0 the same month, 4 later) and 13 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
make_DPLR_HiPPO 3 other papers
1 dated earlier, 2 later
2402.15584 (2024-02)
State Space Models for Event Cameras
2411.02824 (2024-11)
Layer-Adaptive State Pruning for Deep State Space Models
2602.00534 (2026-02)
AIRE-Prune: Asymptotic Impulse-Response Energy for State Pruning in State Space Models
make_HiPPO 3 other papers
1 dated earlier, 2 later
2402.15584 (2024-02)
State Space Models for Event Cameras
2411.02824 (2024-11)
Layer-Adaptive State Pruning for Deep State Space Models
2602.00534 (2026-02)
AIRE-Prune: Asymptotic Impulse-Response Energy for State Pruning in State Space Models
make_NPLR_HiPPO 3 other papers
1 dated earlier, 2 later
2402.15584 (2024-02)
State Space Models for Event Cameras
2411.02824 (2024-11)
Layer-Adaptive State Pruning for Deep State Space Models
2602.00534 (2026-02)
AIRE-Prune: Asymptotic Impulse-Response Energy for State Pruning in State Space Models
masked_meanpool 3 other papers
1 dated earlier, 2 later
2402.15584 (2024-02)
State Space Models for Event Cameras
2411.02824 (2024-11)
Layer-Adaptive State Pruning for Deep State Space Models
2602.00534 (2026-02)
AIRE-Prune: Asymptotic Impulse-Response Energy for State Pruning in State Space Models
map_nested_fn 2 other papers
2 dated earlier
2212.10544 (2022-12)
Pretraining Without Attention
2312.01538 (2023-12)
Recurrent Distance Filtering for Graph Representation 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

Event-based sensors are well suited for real-time processing due to their fast response times and encoding of the sensory data as successive temporal differences. These and other valuable properties, such as a high dynamic range, are suppressed when the data is converted to a frame-based format. However, most current methods either collapse events into frames or cannot scale up when processing the event data directly event-by-event. In this work, we address the key challenges of scaling up event-by-event modeling of the long event streams emitted by such sensors, which is a particularly relevant problem for neuromorphic computing. While prior methods can process up to a few thousand time steps, our model, based on modern recurrent deep state-space models, scales to event streams of millions of events for both training and inference. We leverage their stable parameterization for learning long-range dependencies, parallelizability along the sequence dimension, and their ability to integrate asynchronous events effectively to scale them up to long event streams. We further augment these with novel event-centric techniques enabling our model to match or beat the state-of-the-art performance on several event stream benchmarks. In the Spiking Speech Commands task, we improve state-of-the-art by a large margin of 7.7% to 88.4%. On the DVS128-Gestures dataset, we achieve competitive results without using frames or convolutional neural networks. Our work demonstrates, for the first time, that it is possible to use fully event-based processing with purely recurrent networks to achieve state-of-the-art task performance in several event-based benchmarks.

For agents

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

get_harvested_code_for_paper("2404.18508")
get_code_for_paper("2404.18508")
have("2404.18508")

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

curl https://syntology.ai/api/ran/2404.18508.json

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

[![Syntology run record](https://syntology.ai/api/ran/2404.18508.svg)](https://syntology.ai/paper/2404.18508)

Connect an agent — have() is free.