SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2208.01893 · 2022

Flow Annealed Importance Sampling Bootstrap

arXiv · PDF · Open in the Atlas

Code that ran

21 of 27 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 21 have no contract verdict; 0 failures of our instrument; 6 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/2208.01893.json · badge: /api/ran/2208.01893.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
lollcat/fab-jax canonical (classifier label)
classifier basis not recorded on the link
4 ran · 1 no recorded run
vislearn/trade community (classifier label)
classifier basis not recorded on the link
14 ran · 5 no recorded run
aimat-lab/coarse-graining-al community (classifier label)
classifier basis not recorded on the link
3 ran · 0 no recorded run
FunctionStatusWhere it lives
ActionScalarTheory Ran
ran on a synthesized input
vislearn/trade/trade/datasets.py
code served (permissive licence) · get_code("da01152e23f9257f")
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 1, dtype family float, shape (2,) 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

torch.manual_seed(0)

# mus: Initial states - should be a batch of 2D lattice field configurations
# Shape: (batch_size, channels, height, width)
# The function uses: .device, .pow(2), .pow(4), torch.roll with dims 2 and 3
mus = torch.randn(2, 1, 4, 4)

# kappas: Hopping parameters - can be float or tensor
# The function converts float to tensor of length len(mus)
# Shape should be compatible with len(mus) = 2
kappas = 0.5

# lambdas: Quadric couplings - can be float or tensor
# The function converts float to tensor of length len(mus)
# Shape should be compatible with len(mus) = 2
lambdas = 0.3

ARGS = {
    "mus": mus,
    "kappas": kappas,
    "lambdas": lambdas,
}
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 Tensor, ndim 1, dtype family float, shape (2,) — shape (2,) 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
broadcasted_where Ran
ran on a synthesized input
lollcat/fab-jax/fabjax/utils/jax_util.py
code served (permissive licence) · get_code("31d10b58f9eba56f")
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, 4, 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 chex

# Create valid test arrays for the broadcasted_where function
# The function requires:
# - valid_samples: a boolean array (can have fewer dimensions than a1, a2)
# - a1, a2: arrays with equal shapes

# Create a1 and a2 with shape (2, 4, 3) - batch=2, dim1=4, dim2=3
a1 = jnp.ones((2, 4, 3), dtype=jnp.float32)
a2 = jnp.zeros((2, 4, 3), dtype=jnp.float32)

# Create valid_samples with shape (2,) - fewer dimensions than a1/a2
# The function will expand_dims to broadcast properly
valid_samples = jnp.array([True, False], dtype=jnp.bool_)

ARGS = {
    "valid_samples": valid_samples,
    "a1": a1,
    "a2": a2,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
chex, 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, 4, 3) — shape (2, 4, 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
call_model_batched Ran
ran on a synthesized input
aimat-lab/coarse-graining-al/main/utils/call_model_batched.py
code served (permissive licence) · get_code("39b9d6445fc9d49f")
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 2, dtype family float, shape (6, 3) as a later run recorded it. The image it ran in was not recorded. The code mentions cuda, and the harness aliases cuda to the CPU for every sample; whether this run needed that is 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 torch.nn as nn

torch.manual_seed(0)

# Create a simple model that accepts input_tensor and optional cond_tensor
class SimpleModel(nn.Module):
    def __init__(self):
        super().__init__()
        self.linear = nn.Linear(4, 3)
    
    def forward(self, x, cond=None):
        # Model can accept conditional tensor as tuple or tensor
        return self.linear(x)

# Create model instance
model = SimpleModel()

# Create input tensor: shape (6, 4) - 6 samples, 4 features
input_tensor = torch.randn(6, 4)

# Create conditional tensor: shape (6, 5) - 6 samples, 5 features
cond_tensor = torch.randn(6, 5)

ARGS = {
    'model': model,
    'input_tensor': input_tensor,
    'cond_tensor': cond_tensor,
    'pass_cond_tensor_as_tuple': True,
    'device': 'cpu',
    'batch_size': 2,
    'move_back_to_cpu': True,
    'do_detach': True,
    'droplast': False,
}
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 Tensor, ndim 2, dtype family float, shape (6, 3) — shape (6, 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
circular_mean Ran
ran on a synthesized input
vislearn/trade/trade/coordinate_transform.py
code served (permissive licence) · get_code("ac688680eef129e1")
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 2, dtype family float, shape (1, 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 torch

torch.manual_seed(0)

# alpha is used with .sin(), .cos(), .mean(dim=0), and .view(1, -1)
# It should be a tensor with at least 2 dimensions (batch, features)
# batch size: 2, feature dimension: 4
alpha = torch.randn(2, 4)

ARGS = {
    "alpha": alpha
}
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 Tensor, ndim 2, dtype family float, shape (1, 4) — shape (1, 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
consistency_check_KL Ran
ran on a synthesized input
vislearn/trade/trade/losses.py
code served (permissive licence) · get_code("b8eaabf1c9db7b92")
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 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 torch
import torch.nn as nn

torch.manual_seed(0)

# Create small log probability tensors (batch_size=2, num_classes=4)
log_probs_beta = torch.randn(2, 4)
log_probs_beta_0 = torch.randn(2, 4)

# Create scalar beta values
beta = torch.tensor(0.5)
beta_0 = torch.tensor(1.0)

# Create a correction function that takes beta and beta_0 and returns a scalar tensor
def correction(b, b0):
    return torch.tensor(0.1)

ARGS = {
    "log_probs_beta": log_probs_beta,
    "log_probs_beta_0": log_probs_beta_0,
    "correction": correction,
    "beta": beta,
    "beta_0": beta_0,
}
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 Tensor, 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
constructor_subnet_2Dconv_plain Ran
ran on a synthesized input
vislearn/trade/trade/models.py
code served (permissive licence) · get_code("226d82cc38425813")
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 Sequential. 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)

ARGS = {
    'c_in': 3,
    'c_out': 5,
    'c_hidden': 16,
    'activation_type': 'ReLU'
}
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 Sequential — a Sequential output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a Sequential output the recorded ndim 1 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
constructor_subnet_fc_configured Ran
ran on a synthesized input
vislearn/trade/trade/models.py
code served (permissive licence) · get_code("a16486276f5c5f09")
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 partial. 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
from functools import partial

torch.manual_seed(0)

# Define ARGS dict with all parameters for constructor_subnet_fc_configured
ARGS = {
    'd_hidden': 16,
    'activation_type': 'ReLU'
}
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 partial — a partial output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a partial 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
constructor_subnet_fc_plain Ran
ran on a synthesized input
vislearn/trade/trade/models.py
code served (permissive licence) · get_code("cadccdbc4e5b12fb")
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 Sequential. 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)

ARGS = {
    'c_in': 10,
    'c_out': 5,
    'c_hidden': 8,
    'activation_type': 'ReLU'
}
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 Sequential — a Sequential output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a Sequential output the recorded ndim 1 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
dActionScalarTheory_dkappa Ran
ran on a synthesized input
vislearn/trade/trade/datasets.py
code served (permissive licence) · get_code("3e92b55934524c65")
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 1, dtype family float, shape (2,) 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

torch.manual_seed(0)

# Analyze the function:
# - mus: used with torch.roll and torch operations on dims 2 and 3, then summed over dims [1, 2, 3]
#   This suggests mus is a 4D tensor (batch, dim1, dim2, dim3)
# - parameter_list: optional, type List[torch.Tensor | float], not used in function body
# - device: optional, type str, not used in function body

# Create a small 4D tensor: (batch=2, dim1=3, dim2=4, dim3=5)
mus = torch.randn(2, 3, 4, 5)

# parameter_list can be None or a list of tensors/floats
parameter_list = None

# device can be None or a string
device = None

ARGS = {
    "mus": mus,
    "parameter_list": parameter_list,
    "device": device,
}
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 Tensor, ndim 1, dtype family float, shape (2,) — shape (2,) 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
generate_grid Ran
ran on a synthesized input
aimat-lab/coarse-graining-al/main/active_learning/active_learning_managers.py
code served (permissive licence) · get_code("2327b3b50e622283")
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 2, dtype family float, shape (8, 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 torch

torch.manual_seed(0)

# uniform_ranges: list of tuples or Tensor of shape (D,2)
# Each tuple is (min, max) for that dimension
# Let's use a list of tuples for 3 dimensions
uniform_ranges = [
    (0.0, 1.0),
    (-1.0, 1.0),
    (10.0, 20.0),
]

# num_points_per_dim: int - number of equidistant points per dimension
num_points_per_dim = 2

ARGS = {
    "uniform_ranges": uniform_ranges,
    "num_points_per_dim": num_points_per_dim,
}
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 Tensor, ndim 2, dtype family float, shape (8, 3) — shape (8, 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
get_leading_axis_tree Ran
ran on a synthesized input
lollcat/fab-jax/fabjax/utils/jax_util.py
code served (permissive licence) · get_code("06e166b4bd44c5ce")
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 int; 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
import chex

# Create a simple tree structure with JAX arrays
# The function expects a pytree (nested structure of arrays)
# and will check that all arrays have the same leading shape

# Create arrays with distinct dimensions to test the function
array1 = jnp.ones((2, 4, 3))  # shape (2, 4, 3)
array2 = jnp.ones((2, 5, 6))  # shape (2, 5, 6)

# Create a pytree (dict of arrays) with consistent leading axis
tree = {
    'a': array1,
    'b': array2,
}

ARGS = {
    'tree': tree,
    'n_dims': 1,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
chex, jax — 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 int — 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
is_json_serializable Ran
ran on a synthesized input
vislearn/trade/trade/trainables.py
code served (permissive licence) · get_code("9bc73838eb0f1444")
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 bool, 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 torch
import json

def is_json_serializable(value):
    """Check if a value is JSON serializable."""
    try:
        json.dumps(value)
        return True
    except (TypeError, OverflowError):
        return False

# Test with a JSON-serializable value
ARGS = {
    "value": {"key": "value", "number": 42, "list": [1, 2, 3], "nested": {"inner": 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 bool, 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
log_p_scalar_theory Ran
ran on a synthesized input
vislearn/trade/trade/datasets.py
code served (permissive licence) · get_code("44b1de121fc4f5c1")
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 1, dtype family float, shape (2,) 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

torch.manual_seed(0)

# Create a small batch of scalar field configurations
# Shape: [batch_size, spatial_dim1, spatial_dim2, spatial_dim3]
# Using batch_size=2, spatial dimensions=4x4x4 for a small 2D lattice
x = torch.randn(2, 1, 4, 4)

# Beta tensor (inverse temperature) - one value per batch element
beta_tensor = torch.tensor([0.5, 1.0])

# Lambda tensor (quadric coupling) - one value per batch element
lambdas = torch.tensor([0.1, 0.2])

# Device
device = torch.device('cpu')

ARGS = {
    'x': x,
    'beta_tensor': beta_tensor,
    'lambdas': lambdas,
    'device': device
}
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 Tensor, ndim 1, dtype family float, shape (2,) — shape (2,) 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
plot_double_well_2d Ran
ran on a synthesized input
vislearn/trade/trade/plots.py
code served (permissive licence) · get_code("f222aa8da4211121")
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 torch
import torch.nn as nn
import numpy as np

torch.manual_seed(0)

# Create a small energy function that takes samples and returns energies
class EnergyFunction(nn.Module):
    def __init__(self):
        super().__init__()
        self.linear = nn.Linear(2, 1)
    
    def forward(self, x):
        # Return energy values as a 1D tensor
        return self.linear(x).squeeze(-1)

# Create sample data: batch_size=2, dimensions=2 for 2D double well
samples = torch.randn(2, 2)

# Create target energy function
target_energy = EnergyFunction()

# Create reference/test data (optional, can be None)
reference_data = torch.randn(2, 2)

# Create a simple system object (may not be used but included for completeness)
system = None

ARGS = {
    "samples": samples,
    "system": system,
    "target_energy": target_energy,
    "reference_data": reference_data
}
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 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_nd_histogram Ran
ran on a synthesized input
vislearn/trade/trade/plots.py
code served (permissive licence) · get_code("21e05c489f7ca319")
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 torch
import numpy as np
import matplotlib.pyplot as plt
import torch.nn as nn

torch.manual_seed(0)

# Create a simple energy function that takes samples and returns energies
class SimpleEnergyFunction(nn.Module):
    def __init__(self):
        super().__init__()
        self.linear = nn.Linear(4, 1)
    
    def forward(self, x):
        # Returns energy values for each sample
        return self.linear(x).squeeze(-1)

# Create sample data: batch_size=2, dimensions=4
samples = torch.randn(2, 4)

# Create reference/test data: batch_size=2, dimensions=4
reference_data = torch.randn(2, 4)

# Create a simple system object (not directly used in plot_nd_histogram, but kept for completeness)
class SimpleSystem:
    pass

system = SimpleSystem()

# Create energy function
target_energy = SimpleEnergyFunction()

ARGS = {
    "samples": samples,
    "system": system,
    "target_energy": target_energy,
    "reference_data": reference_data
}
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 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
read_config Ran
ran on a synthesized input
vislearn/trade/trade/config.py
code served (permissive licence) · get_code("4b3e1d756ced6adf")
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 torch
import yaml
import tempfile
import os

# Create a temporary YAML file with valid content
temp_dir = tempfile.mkdtemp()
yaml_file_path = os.path.join(temp_dir, 'config.yaml')

config_data = {
    'model': 'bert',
    'learning_rate': 0.001,
    'batch_size': 32,
    'epochs': 10,
    'layers': [64, 128, 256]
}

with open(yaml_file_path, 'w') as f:
    yaml.dump(config_data, f)

ARGS = {
    'file_path': yaml_file_path
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch, yaml — 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
remove_non_serializable Ran
ran on a synthesized input
vislearn/trade/trade/trainables.py
code served (permissive licence) · get_code("05e36a7f85d97e27")
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 torch
import numpy as np
import json

# Test data with both serializable and non-serializable values
test_dict = {
    "name": "test",
    "count": 42,
    "ratio": 3.14,
    "active": True,
    "items": [1, 2, 3],
    "nested": {
        "key1": "value1",
        "key2": 100,
        "tensor": torch.tensor([1, 2, 3]),  # non-serializable
        "numpy_array": np.array([1, 2, 3]),  # non-serializable
    },
    "tensor_list": [torch.tensor([1.0, 2.0]), "valid_string", 5],
    "callable": lambda x: x,  # non-serializable
    "none_value": None,
}

test_list = [
    "string_value",
    42,
    3.14,
    {"nested_key": "nested_value"},
    torch.tensor([1, 2, 3]),  # non-serializable
    np.array([4, 5, 6]),  # non-serializable
    [1, 2, {"inner": "dict"}],
]

ARGS = {
    "obj": test_dict
}
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 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
reverse_kl_loss Ran
ran on a synthesized input
lollcat/fab-jax/fabjax/train/fab_without_buffer.py
code served (permissive licence) · get_code("e2da966a3225548a")
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
import jax.random as jrandom
import chex

# Define sample_and_log_prob_apply function
# Takes params and shape, returns (samples, log_probs)
def q_sample_and_log_prob_apply(params, shape):
    batch_size = shape[0]
    # Simple Gaussian sampler
    key = jrandom.PRNGKey(0)
    samples = jrandom.normal(key, (batch_size, 4))
    log_probs = jnp.sum(-0.5 * samples**2, axis=1)
    return samples, log_probs

# Define log_q_fn_apply function
# Takes params and samples, returns log_probs
def log_q_fn_apply(params, samples):
    # Simple log probability function
    return jnp.sum(-0.5 * samples**2, axis=1)

# Define log_p_fn function
# Takes samples, returns log_probs
def log_p_fn(samples):
    # Simple target log probability (e.g., standard normal)
    return jnp.sum(-0.5 * samples**2, axis=1)

# Create simple params tree (dict of arrays)
params = {
    'w': jnp.array([[0.1, 0.2, 0.3, 0.4],
                     [0.5, 0.6, 0.7, 0.8]]),
    'b': jnp.array([0.01, 0.02])
}

ARGS = {
    'params': params,
    'q_sample_and_log_prob_apply': q_sample_and_log_prob_apply,
    'log_q_fn_apply': log_q_fn_apply,
    'log_p_fn': log_p_fn,
    'batch_size': 2,
    'path_gradient': True
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
chex, 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
sample_without_replacement Ran
ran on a synthesized input
lollcat/fab-jax/fabjax/buffer/prioritised_buffer.py
code served (permissive licence) · get_code("147034b3e56c8828")
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 int, shape (2, 2) 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.random
import jax.lax
import chex
import numpy as np

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

# Create small tensors
logits = jax.numpy.array([
    [1.0, 2.0, 3.0, 0.5],
    [0.1, 0.2, 0.3, 0.4]
], dtype=jax.numpy.float32)

n = 2

ARGS = {
    "key": key,
    "logits": logits,
    "n": n
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
chex, 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 2, dtype family int, shape (2, 2) — shape (2, 2) 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
select_diverse_points Ran
ran on a synthesized input
aimat-lab/coarse-graining-al/main/active_learning/active_learning_managers.py
code served (permissive licence) · get_code("a7e6a591642b921e")
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 2, dtype family float, shape (3, 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 torch

torch.manual_seed(0)

# Create a small tensor with shape (batch=5, dim=4)
tensor = torch.randn(5, 4)

# N should be less than the number of points in tensor
N = 3

ARGS = {
    "tensor": tensor,
    "N": N,
}
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 Tensor, ndim 2, dtype family float, shape (3, 4) — shape (3, 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
shift_angles Ran
ran on a synthesized input
vislearn/trade/trade/coordinate_transform.py
code served (permissive licence) · get_code("c12e51ebb2df519b")
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 ndarray, 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 numpy as np

# Create test arguments for shift_angles function
alpha = np.array([0.5, 1.5, -2.0, 3.14])  # Example angles
shift = np.array([0.1, 0.2, 0.3, 0.4])   # Example shifts

ARGS = {
    "alpha": alpha,
    "shift": shift
}
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 ndarray, 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
build_resnet Not yet run
unverified (harvested, no recorded run)
vislearn/trade/trade/flow.py
code served (permissive licence) · get_code("cfb7a6f3f78317c8")
How it ran

no invocation recorded.

build_subnet Not yet run
unverified (harvested, no recorded run)
vislearn/trade/trade/flow.py
code served (permissive licence) · get_code("19a851e76c47a32b")
How it ran

no invocation recorded.

compute_KL Not yet run
unverified (harvested, no recorded run)
vislearn/trade/trade/losses.py
code served (permissive licence) · get_code("a4982e7e1d9e5669")
How it ran

no invocation recorded.

ct_shift_angle Not yet run
unverified (harvested, no recorded run)
vislearn/trade/trade/coordinate_transform.py
code served (permissive licence) · get_code("8b55b599f66645b6")
How it ran

no invocation recorded.

fab_loss_smc_samples Not yet run
unverified (harvested, no recorded run)
lollcat/fab-jax/fabjax/train/fab_without_buffer.py
code served (permissive licence) · get_code("5e4df7287719cf48")
How it ran

no invocation recorded.

get_correction_function Not yet run
unverified (harvested, no recorded run)
vislearn/trade/trade/losses.py
code served (permissive licence) · get_code("01444979755c6722")
How it ran

no invocation recorded.

Also present in other papers' code

Of 27 samples lifted from this paper's linked repositories, 27 carry a function identity and 0 do not. Of the 27 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 27 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 2026-11-22.

Repositories linked to this paper

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

Abstract

Normalizing flows are tractable density models that can approximate complicated target distributions, e.g. Boltzmann distributions of physical systems. However, current methods for training flows either suffer from mode-seeking behavior, use samples from the target generated beforehand by expensive MCMC methods, or use stochastic losses that have high variance. To avoid these problems, we augment flows with annealed importance sampling (AIS) and minimize the mass-covering $α$-divergence with $α=2$, which minimizes importance weight variance. Our method, Flow AIS Bootstrap (FAB), uses AIS to generate samples in regions where the flow is a poor approximation of the target, facilitating the discovery of new modes. We apply FAB to multimodal targets and show that we can approximate them very accurately where previous methods fail. To the best of our knowledge, we are the first to learn the Boltzmann distribution of the alanine dipeptide molecule using only the unnormalized target density, without access to samples generated via Molecular Dynamics (MD) simulations: FAB produces better results than training via maximum likelihood on MD samples while using 100 times fewer target evaluations. After reweighting the samples, we obtain unbiased histograms of dihedral angles that are almost identical to the ground truth.

For agents

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

get_harvested_code_for_paper("2208.01893")
get_code_for_paper("2208.01893")
have("2208.01893")

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

curl https://syntology.ai/api/ran/2208.01893.json

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

[![Syntology run record](https://syntology.ai/api/ran/2208.01893.svg)](https://syntology.ai/paper/2208.01893)

Connect an agent — have() is free.