SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2004.08013 · 2020

How recurrent networks implement contextual processing in sentiment analysis

arXiv · PDF · Open in the Atlas

Code that ran

15 of 19 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; 0 honoured the contract we drafted, 0 violated it and 15 have no contract verdict; 0 failures of our instrument; 4 no recorded run), measured 2026-11-21; 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/2004.08013.json · badge: /api/ran/2004.08013.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
google-research/computation-thru-dynamics community (classifier label)
classifier basis not recorded on the link
15 ran · 4 no recorded run
FunctionStatusWhere it lives
affine Ran
ran on a synthesized input
google-research/computation-thru-dynamics/integrator_rnn_tutorial/rnn.py
code served (permissive licence) · get_code("aced411c40f5de2f")
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 (3,) 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

# Create params dict with weight matrix and bias vector
params = {
    'wO': jnp.array([[1.0, 2.0, 3.0, 4.0],
                     [5.0, 6.0, 7.0, 8.0],
                     [9.0, 10.0, 11.0, 12.0]]),  # shape (3, 4)
    'bO': jnp.array([0.1, 0.2, 0.3])  # shape (3,)
}

# Create input vector x
x = jnp.array([1.0, 2.0, 3.0, 4.0])  # shape (4,)

ARGS = {
    'params': params,
    'x': x
}
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 (3,) — shape (3,) 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
diag_gaussian_log_likelihood Ran
ran on a synthesized input
google-research/computation-thru-dynamics/lfads_tutorial/distributions.py
code served (permissive licence) · get_code("896edacf34bb8bd8")
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 (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 np

# Create small test arrays for the function parameters
# z: The value to compute log-likelihood (shape: batch=2, dims=4)
z = np.array([[1.0, 2.0, 3.0, 4.0],
              [0.5, 1.5, 2.5, 3.5]])

# mean: The mean of the Gaussian (can be scalar or same shape as z)
mean = np.array([[0.0, 0.0, 0.0, 0.0],
                 [0.0, 0.0, 0.0, 0.0]])

# logvar: Log variance of the Gaussian (can be scalar or same shape as z)
logvar = np.array([[0.0, 0.0, 0.0, 0.0],
                   [0.0, 0.0, 0.0, 0.0]])

# varmin: Minimum variance (scalar, default is 1e-16)
varmin = 1e-16

ARGS = {
    'z': z,
    'mean': mean,
    'logvar': logvar,
    'varmin': varmin
}
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 (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
diag_gaussian_sample Ran
ran on a synthesized input
google-research/computation-thru-dynamics/lfads_tutorial/distributions.py
code served (permissive licence) · get_code("0b1d13b86e3d6a9e")
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
from jax import random

# Set seed for reproducibility
seed = 0
key = random.PRNGKey(seed)

# Create small test arrays for mean and logvar
mean = np.array([1.0, 2.0, 3.0, 4.0])
logvar = np.array([-1.0, -0.5, 0.0, 0.5])

# Define varmin (optional parameter, use default)
varmin = 1e-16

# Construct ARGS dictionary with all parameters
ARGS = {
    "key": key,
    "mean": mean,
    "logvar": logvar,
    "varmin": varmin
}
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
get_fp_loss_fun Ran
ran on a synthesized input
google-research/computation-thru-dynamics/fixed_point_finder/fixed_points.py
code served (permissive licence) · get_code("34a73dab2ffc9aa7")
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 PjitFunction. 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
from jax import grad, jacrev, jit, lax, random, vmap

# Define a simple RNN one-step update function
# This function takes a hidden state h of shape (batch_size, hidden_dim)
# and returns the next hidden state of the same shape
def simple_rnn_step(h):
    """Simple RNN one-step update: h_t+1 = tanh(W @ h_t + b)"""
    # h shape: (batch_size, hidden_dim)
    # We'll use a simple linear transformation followed by tanh
    batch_size, hidden_dim = h.shape
    
    # Create simple weight matrix and bias
    W = jnp.ones((hidden_dim, hidden_dim)) * 0.5
    b = jnp.zeros(hidden_dim)
    
    # Apply transformation to each batch element
    # Since this is called with vmap(rnn_fun, in_axes=(0,)), 
    # it receives individual hidden states of shape (hidden_dim,)
    h_next = jnp.tanh(jnp.dot(W, h) + b)
    return h_next

ARGS = {
    "rnn_fun": simple_rnn_step
}
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 PjitFunction — a PjitFunction output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a PjitFunction 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_kl_warmup_fun Ran
ran on a synthesized input
google-research/computation-thru-dynamics/lfads_tutorial/optimize.py
code served (permissive licence) · get_code("2cc94ae064dc4dd7")
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 jax.numpy as np

ARGS = {
    'lfads_opt_hps': {
        'kl_warmup_start': 0,
        'kl_warmup_end': 1000,
        'kl_min': 0.0,
        'kl_max': 1.0,
    }
}
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 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
get_total_fp_loss_fun Ran
ran on a synthesized input
google-research/computation-thru-dynamics/fixed_point_finder/fixed_points.py
code served (permissive licence) · get_code("c873384736c6ede9")
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 PjitFunction. 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 import vmap, jit

def create_simple_rnn_fun():
    """Create a simple RNN one-step update function."""
    def rnn_fun(h):
        # Simple linear transformation: h_t+1 = 0.9 * h_t + 0.1
        return 0.9 * h + 0.1
    return rnn_fun

rnn_fun = create_simple_rnn_fun()

ARGS = {
    "rnn_fun": rnn_fun,
}
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 PjitFunction — a PjitFunction output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a PjitFunction 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
gru Ran
ran on a synthesized input
google-research/computation-thru-dynamics/fixed_point_finder/rnn.py
code served (permissive licence) · get_code("3976c1f7c4e6dab7")
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

# Determine dimensions based on function usage:
# - h and x are concatenated, so they can be 1D arrays
# - hx = concatenate([h, x]) -> shape will be (h_dim + x_dim,)
# - params['wRUHX'] @ hx -> output shape (2 * hidden_dim,) for r, u split
# - params['wCHX'] @ rhx -> output shape (hidden_dim,)

hidden_dim = 4
input_dim = 3

# params['wRUHX']: shape (2*hidden_dim, hidden_dim + input_dim)
wRUHX = np.ones((2 * hidden_dim, hidden_dim + input_dim))

# params['bRU']: shape (2*hidden_dim,)
bRU = np.zeros(2 * hidden_dim)

# params['wCHX']: shape (hidden_dim, hidden_dim + input_dim)
wCHX = np.ones((hidden_dim, hidden_dim + input_dim))

# params['bC']: shape (hidden_dim,)
bC = np.zeros(hidden_dim)

params = {
    'wRUHX': wRUHX,
    'bRU': bRU,
    'wCHX': wCHX,
    'bC': bC,
}

# h: hidden state, shape (hidden_dim,)
h = np.ones(hidden_dim)

# x: input, shape (input_dim,)
x = np.ones(input_dim)

# bfg: bias on forget gate
bfg = 0.5

ARGS = {
    'params': params,
    'h': h,
    'x': x,
    'bfg': bfg,
}
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
keygen Ran
ran on a synthesized input
google-research/computation-thru-dynamics/fixed_point_finder/utils.py
code served (permissive licence) · get_code("295c6b5c69a5cfa0")
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
from jax import random

# Create a JAX random key
key = random.PRNGKey(0)

# Set number of keys to generate
nkeys = 5

ARGS = {
    "key": key,
    "nkeys": nkeys,
}
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
merge_losses_dicts Ran
ran on a synthesized input
google-research/computation-thru-dynamics/fixed_point_finder/utils.py
code served (permissive licence) · get_code("3fad852c8d126922")
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 numpy as onp

# The function takes a list of dictionaries and merges them
# Each dictionary should have numeric values that can be converted to numpy arrays

ARGS = {
    "list_of_dicts": [
        {
            "loss": 0.5,
            "accuracy": 0.92,
            "val_loss": 0.6
        },
        {
            "loss": 0.45,
            "accuracy": 0.94,
            "val_loss": 0.55
        },
        {
            "loss": 0.42,
            "accuracy": 0.96,
            "val_loss": 0.52
        }
    ]
}
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 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
plot_data_stats Ran
ran on a synthesized input
google-research/computation-thru-dynamics/lfads_tutorial/plotting.py
code served (permissive licence) · get_code("ae64d899719d4eb9")
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 Figure. 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 numpy as onp

torch.manual_seed(0)

# Construct the arguments for plot_data_stats
# The function expects:
# - data_dict: a dict with 'hiddens' key containing neural rates
# - data_bxtxn: spike data with shape (batch, time, neurons)
# - data_dt: time step (scalar)

batch_size = 2
time_steps = 5
num_neurons = 4

# data_dict with 'hiddens' key - neural rate data (batch, time, neurons)
data_dict = {
    'hiddens': onp.random.randn(batch_size, time_steps, num_neurons)
}

# data_bxtxn - spike data (batch, time, neurons) - binary or count data
data_bxtxn = onp.random.poisson(lam=0.5, size=(batch_size, time_steps, num_neurons))

# data_dt - time step in seconds
data_dt = 0.01

ARGS = {
    'data_dict': data_dict,
    'data_bxtxn': data_bxtxn,
    'data_dt': data_dt
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
matplotlib, numpy, torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Figure — a Figure output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a Figure 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
poisson_log_likelihood Ran
ran on a synthesized input
google-research/computation-thru-dynamics/lfads_tutorial/distributions.py
code served (permissive licence) · get_code("a80bb6e1b393fe43")
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 (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 np

# Create small test tensors for spike count data and log rates
x = np.array([[1.0, 2.0, 3.0, 0.0],
              [2.0, 1.0, 0.0, 1.0]])

log_rate = np.array([[-0.5, 0.0, 0.5, 1.0],
                     [0.2, -0.3, 0.1, -0.2]])

ARGS = {
    "x": x,
    "log_rate": log_rate
}
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 (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
read_file Ran
ran on a synthesized input
google-research/computation-thru-dynamics/fixed_point_finder/utils.py
code served (permissive licence) · get_code("670f241a0a278e35")
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 h5py
import numpy as onp
import tempfile
import os

# Create a temporary HDF5 file with test data
temp_dir = tempfile.mkdtemp()
test_file = os.path.join(temp_dir, "test_data.h5")

with h5py.File(test_file, 'w') as hf:
    hf.create_dataset('array1', data=onp.array([[1, 2, 3], [4, 5, 6]]))
    hf.create_dataset('array2', data=onp.array([7, 8, 9, 10]))
    hf.create_dataset('array3', data=onp.array(42.0))

ARGS = {
    'data_fname': test_file
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
h5py, numpy — 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
sigmoid Ran
ran on a synthesized input
google-research/computation-thru-dynamics/fixed_point_finder/rnn.py
code served (permissive licence) · get_code("deee5acedc65539b")
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 (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

# The sigmoid function takes a single parameter 'x' which is used with:
# - np.tanh() operation (expects array-like)
# - arithmetic operations (division by 2.0, addition with 1, multiplication by 0.5)
# So x should be a JAX array

x = jnp.array([[-1.5, 0.0, 1.5, 2.0],
               [-0.5, 0.5, 1.0, -2.0]])

ARGS = {
    "x": x
}
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 (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
sigmoid Ran
ran on a synthesized input
google-research/computation-thru-dynamics/lfads_tutorial/lfads.py
code served (permissive licence) · get_code("48f60e785be2e858")
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

ARGS = {
    "x": np.array([0.5, 1.0, -0.5, 0.0])
}
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
split_data Ran
ran on a synthesized input
google-research/computation-thru-dynamics/lfads_tutorial/utils.py
code served (permissive licence) · get_code("1f5c168fff169bd8")
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 a small numpy array with batch dimension
# Shape: (10, 4) - 10 samples, 4 features
data_b = np.random.randn(10, 4)

# Define train_fraction
train_fraction = 0.7

ARGS = {
    "data_b": data_b,
    "train_fraction": train_fraction,
}
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
plot_data_example Not yet run
unverified (harvested, no recorded run)
google-research/computation-thru-dynamics/lfads_tutorial/plotting.py
code served (permissive licence) · get_code("23408171ea05d4c3")
How it ran

no invocation recorded.

plot_data_pca Not yet run
unverified (harvested, no recorded run)
google-research/computation-thru-dynamics/lfads_tutorial/plotting.py
code served (permissive licence) · get_code("5f40340057d4af85")
How it ran

no invocation recorded.

spikify_data Not yet run
unverified (harvested, no recorded run)
google-research/computation-thru-dynamics/lfads_tutorial/utils.py
code served (permissive licence) · get_code("846c9d4c4a3c7b6c")
How it ran

no invocation recorded.

vrnn Not yet run
unverified (harvested, no recorded run)
google-research/computation-thru-dynamics/integrator_rnn_tutorial/rnn.py
code served (permissive licence) · get_code("61a25573238a87a1")
How it ran

no invocation recorded.

Also present in other papers' code

Of 19 samples lifted from this paper's linked repositories, 19 carry a function identity and 0 do not. Of the 18 distinct functions with an identity, 0 are also present in other papers whose code we hold (0 with at least one paper dated earlier, 0 the same month, 0 later) and 18 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.

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 2026-11-21.

Repositories linked to this paper

Some links come from the archived Papers with Code dataset (CC BY-SA 4.0): attribution and licence.

Abstract

Neural networks have a remarkable capacity for contextual processing--using recent or nearby inputs to modify processing of current input. For example, in natural language, contextual processing is necessary to correctly interpret negation (e.g. phrases such as "not bad"). However, our ability to understand how networks process context is limited. Here, we propose general methods for reverse engineering recurrent neural networks (RNNs) to identify and elucidate contextual processing. We apply these methods to understand RNNs trained on sentiment classification. This analysis reveals inputs that induce contextual effects, quantifies the strength and timescale of these effects, and identifies sets of these inputs with similar properties. Additionally, we analyze contextual effects related to differential processing of the beginning and end of documents. Using the insights learned from the RNNs we improve baseline Bag-of-Words models with simple extensions that incorporate contextual modification, recovering greater than 90% of the RNN's performance increase over the baseline. This work yields a new understanding of how RNNs process contextual information, and provides tools that should provide similar insight more broadly.

For agents

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

get_harvested_code_for_paper("2004.08013")
get_code_for_paper("2004.08013")
have("2004.08013")

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

curl https://syntology.ai/api/ran/2004.08013.json

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

[![Syntology run record](https://syntology.ai/api/ran/2004.08013.svg)](https://syntology.ai/paper/2004.08013)

Connect an agent — have() is free.