SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2303.05376 · 2023

PC-JeDi: Diffusion for Particle Cloud Generation in High Energy Physics

arXiv · PDF · Open in the Atlas

Code that ran

15 of 17 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 14 have no contract verdict; 1 failure of our instrument; 2 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/2303.05376.json · badge: /api/ran/2303.05376.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
rodem-hep/pc-jedi canonical (classifier label)
classifier basis not recorded on the link
15 ran · 2 no recorded run
FunctionStatusWhere it lives
attention Ran
ran on a synthesized input
rodem-hep/pc-jedi/src/models/transformers.py
code served (permissive licence) · get_code("37f26eb1c05ac461")
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, 3, 4, 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 math

torch.manual_seed(0)

# Batch size: 2, heads: 3, sequence length: 4, feature dim: 5
batch_size = 2
num_heads = 3
seq_length = 4
feature_dim = 5

# Create query, key, value tensors with shape (batch, heads, seq_length, feature_dim)
query = torch.randn(batch_size, num_heads, seq_length, feature_dim)
key = torch.randn(batch_size, num_heads, seq_length, feature_dim)
value = torch.randn(batch_size, num_heads, seq_length, feature_dim)

# Dimension of key features (used for scaling)
dim_key = feature_dim

# Attention mask: shape (batch, seq_length, seq_length) - boolean tensor
# True means attend, False means mask out
attn_mask = torch.ones(batch_size, seq_length, seq_length, dtype=torch.bool)

# Attention bias: shape (batch, seq_length, seq_length, heads)
# Extra weights to add to attention scores
attn_bias = torch.randn(batch_size, seq_length, seq_length, num_heads)

# Dropout probability
drp = 0.1

# Training mode
training = True

ARGS = {
    'query': query,
    'key': key,
    'value': value,
    'dim_key': dim_key,
    'attn_mask': attn_mask,
    'attn_bias': attn_bias,
    'drp': drp,
    'training': training,
}
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 4, dtype family float, shape (2, 3, 4, 5) — shape (2, 3, 4, 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
cosine_beta_shedule Ran
ran on a synthesized input
rodem-hep/pc-jedi/src/models/diffusion.py
code served (permissive licence) · get_code("1504ad575ab6e48d")
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 (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
import math

torch.manual_seed(0)

# The function expects:
# - diff_time: T.Tensor (used in arithmetic operations)
# - max_sr: float (default 1)
# - min_sr: float (default 0.01)

# Create a small tensor for diff_time
# The function uses it in: diffusion_angles = start_angle + diff_time * (end_angle - start_angle)
# So it should be a tensor of reasonable shape
diff_time = torch.linspace(0, 1, 4)

ARGS = {
    "diff_time": diff_time,
    "max_sr": 0.9,
    "min_sr": 0.02
}
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 (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
cosine_diffusion_shedule Ran
ran on a synthesized input
rodem-hep/pc-jedi/src/models/diffusion.py
code served (permissive licence) · get_code("2995fa6ac54da40c")
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 3, 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
import math

torch.manual_seed(0)

# Create a small tensor with diffusion times between 0 and 1
# Shape: (2, 4) - batch of 2, with 4 time values
diff_time = torch.linspace(0, 1, 8).reshape(2, 4)

# Use default values for max_sr and min_sr
max_sr = 1.0
min_sr = 0.01

ARGS = {
    "diff_time": diff_time,
    "max_sr": max_sr,
    "min_sr": min_sr,
}
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 3, 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
cosine_encoding Ran
ran; our contract draft was wrong, not the code
rodem-hep/pc-jedi/src/models/modules.py
code served (permissive licence) · get_code("6a728d7d4c676cdb")
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 (2, 32) 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 math

torch.manual_seed(0)

# Create a small input tensor
# The function expects the final dimension to be 1, or it will unsqueeze it
# Let's create a 2D tensor with shape (2, 1) - batch of 2 values
x = torch.tensor([[0.3], [0.7]], dtype=torch.float32)

# Parameters for the function
outp_dim = 32
min_value = 0.0
max_value = 1.0
frequency_scaling = 'exponential'

ARGS = {
    'x': x,
    'outp_dim': outp_dim,
    'min_value': min_value,
    'max_value': max_value,
    'frequency_scaling': frequency_scaling
}
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 (2, 32) — shape (2, 32) 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
ddim_predict Ran
ran on a synthesized input
rodem-hep/pc-jedi/src/models/diffusion.py
code served (permissive licence) · get_code("2b57a32088530b36")
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 (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 torch as T

T.manual_seed(0)

# Create small test tensors with distinct dimensions
batch_size = 2
feature_dim = 4

# noisy_data: the input noisy tensor
noisy_data = T.randn(batch_size, feature_dim)

# pred_noises: predicted noise from model
pred_noises = T.randn(batch_size, feature_dim)

# signal_rates: scaling factor for signal component
signal_rates = T.ones(batch_size, 1) * 0.7

# noise_rates: scaling factor for noise component
noise_rates = T.ones(batch_size, 1) * 0.3

ARGS = {
    'noisy_data': noisy_data,
    'pred_noises': pred_noises,
    'signal_rates': signal_rates,
    'noise_rates': noise_rates,
}
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 (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
get_act Ran
ran on a synthesized input
rodem-hep/pc-jedi/src/models/modules.py
code served (permissive licence) · get_code("4e36f432e6fda919")
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 ReLU. 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)

# The function get_act takes a single parameter: name (str)
# It returns an nn.Module activation function based on the name
# Valid names are: 'relu', 'lrlu', 'silu', 'swish', 'selu', 'softmax', 'gelu', 'tanh', 'sigmoid'

ARGS = {
    'name': '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 ReLU — a ReLU output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a ReLU 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_loss_fn Ran
ran on a synthesized input
rodem-hep/pc-jedi/src/torch_utils.py
code served (permissive licence) · get_code("03b2ee87958648c6")
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 MSELoss. 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)

# The function get_loss_fn takes a 'name' parameter and optional kwargs
# Looking at the function body, it:
# - Checks the name against specific strings ('none', 'huber', 'mse', 'mae')
# - Returns None or an nn.Module loss function
# - The kwargs are passed to the loss constructors but the function doesn't use them directly

# We'll test with a valid loss name that exercises the main path
ARGS = {
    'name': 'mse',
}
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 MSELoss — a MSELoss output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a MSELoss 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_nrm Ran
ran on a synthesized input
rodem-hep/pc-jedi/src/models/modules.py
code served (permissive licence) · get_code("e8e415ee3dbd27cb")
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 BatchNorm1d. 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 = {
    'name': 'batch',
    'outp_dim': 64,
}
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 BatchNorm1d — a BatchNorm1d output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a BatchNorm1d 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
locals_to_mass_and_pt Ran
ran on a synthesized input
rodem-hep/pc-jedi/src/physics.py
code served (permissive licence) · get_code("40c1c42984bb85ec")
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 (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 torch as T

T.manual_seed(0)

# csts should be a tensor of shape (batch, num_constituents, 3)
# where the last dimension contains [del_eta, del_phi, log_pt]
# Using batch_size=2, num_constituents=4
csts = T.randn(2, 4, 3)

# mask should be a BoolTensor of shape (batch, num_constituents)
# to select which constituents are valid
mask = T.tensor([
    [True, True, True, False],
    [True, True, False, False]
])

ARGS = {
    "csts": csts,
    "mask": mask
}
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 (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
log_squash Ran
ran on a synthesized input
rodem-hep/pc-jedi/src/numpy_utils.py
code served (permissive licence) · get_code("7ba8326426ab16c1")
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 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 numpy as np

# Create a simple numpy array with mixed positive and negative values
# This will test the sign preservation and log squashing behavior
data = np.array([
    [-10.5, 2.3, -0.1, 100.0],
    [0.0, -50.2, 3.14, -0.001]
], dtype=np.float32)

ARGS = {
    "data": data
}
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 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
merge_masks Ran
ran on a synthesized input
rodem-hep/pc-jedi/src/models/transformers.py
code served (permissive licence) · get_code("13ad605957a293d0")
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 3, dtype family float, shape (2, 3, 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 as T

T.manual_seed(0)

device = T.device('cpu')

# q_mask: Union[T.BoolTensor, None] - boolean mask for queries
q_mask = T.tensor([[True, True, False], [True, True, True]], dtype=T.bool, device=device)

# kv_mask: Union[T.BoolTensor, None] - boolean mask for key/values
kv_mask = T.tensor([[True, False], [True, True]], dtype=T.bool, device=device)

# attn_mask: Union[T.BoolTensor, None] - attention mask
attn_mask = T.tensor([[[True, True], [True, False], [False, False]], 
                       [[True, True], [True, True], [True, False]]], dtype=T.bool, device=device)

# q_shape: T.Size - shape of query tensor (batch=2, seq_len=3, hidden=4)
q_shape = T.Size([2, 3, 4])

# k_shape: T.Size - shape of key tensor (batch=2, seq_len=2, hidden=4)
k_shape = T.Size([2, 2, 4])

ARGS = {
    'q_mask': q_mask,
    'kv_mask': kv_mask,
    'attn_mask': attn_mask,
    'q_shape': q_shape,
    'k_shape': k_shape,
    '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 3, dtype family float, shape (2, 3, 2) — shape (2, 3, 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
numpy_locals_to_mass_and_pt Ran
ran on a synthesized input
rodem-hep/pc-jedi/src/physics.py
code served (permissive licence) · get_code("9ab34bf4ce285384")
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 2, dtype family float, 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 numpy as np

# Create small sample data for constituents
# Shape: (batch_size, num_constituents, num_features)
# Features are: del_eta, del_phi, log_pt (or pt)
batch_size = 2
num_constituents = 4
num_features = 3

csts = np.random.randn(batch_size, num_constituents, num_features).astype(np.float32)
# Make pt values reasonable (del_eta and del_phi should be small, pt should be positive when logged)
csts[..., 2] = np.random.uniform(0, 3, (batch_size, num_constituents))

# Create mask to select valid constituents
# Shape: (batch_size, num_constituents)
mask = np.ones((batch_size, num_constituents), dtype=np.float32)
mask[0, 3] = 0  # mask out one constituent in first batch

ARGS = {
    "csts": csts,
    "mask": mask,
    "pt_logged": True,
}
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 2, dtype family float, 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
reload_original_config Ran
ran on a synthesized input
rodem-hep/pc-jedi/src/hydra_utils.py
code served (permissive licence) · get_code("d690d626ae8c473f")
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 DictConfig. 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 os
import tempfile
from pathlib import Path
from omegaconf import DictConfig, OmegaConf
import torch

torch.manual_seed(0)

# Create a temporary directory to act as the working directory
tmpdir = tempfile.mkdtemp()
original_cwd = os.getcwd()
os.chdir(tmpdir)

# Create a minimal full_config.yaml file
config_yaml_path = Path(tmpdir) / 'full_config.yaml'
config_yaml_content = """
ckpt_path: null
loggers:
  wandb:
    resume: false
"""
config_yaml_path.write_text(config_yaml_content)

# Create checkpoints directory with dummy checkpoint files
checkpoints_dir = Path(tmpdir) / 'checkpoints'
checkpoints_dir.mkdir(exist_ok=True)

# Create last and best checkpoint files with different modification times
last_ckpt = checkpoints_dir / 'last_v0.ckpt'
best_ckpt = checkpoints_dir / 'best_v0.ckpt'
last_ckpt.write_text('dummy checkpoint content')
best_ckpt.write_text('dummy checkpoint content')

# Ensure last is more recent
import time
time.sleep(0.01)
last_ckpt.touch()

# Create the input config
input_cfg = OmegaConf.create({
    'ckpt_path': None,
    'loggers': {
        'wandb': {
            'resume': False
        }
    }
})

ARGS = {
    'cfg': input_cfg,
    'get_best': False
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
omegaconf, pytorch_lightning, torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type DictConfig — a DictConfig output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a DictConfig 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
to_np Ran
ran on a synthesized input
rodem-hep/pc-jedi/src/torch_utils.py
code served (permissive licence) · get_code("1042ba4270cff87d")
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 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 torch as T
import numpy as np

T.manual_seed(0)

# Test case 1: A single tensor
tensor_input = T.randn(2, 4)

# Test case 2: A tuple of tensors
tuple_input = (T.randn(2, 4), T.randn(3, 5))

# Test case 3: A list of tensors
list_input = [T.randn(2, 4), T.randn(3, 5)]

# Test case 4: A tensor with bfloat16 dtype
bfloat16_tensor = T.randn(2, 4).to(dtype=T.bfloat16)

# Test case 5: A tensor with gradient
tensor_with_grad = T.randn(2, 4, requires_grad=True)

ARGS = {
    "inpt": tensor_input
}
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 ndarray, 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
undo_log_squash Ran
ran on a synthesized input
rodem-hep/pc-jedi/src/numpy_utils.py
code served (permissive licence) · get_code("fb41e7055a7e1e1e")
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 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 numpy as np

# Create test data for undo_log_squash
# The function takes an np.ndarray and applies the inverse of a log squash operation
# It uses np.sign, np.exp, and np.abs which work element-wise on arrays

# Create a small test array with mixed positive and negative values
data = np.array([
    [0.5, -0.3, 1.2, -0.8],
    [0.1, -2.0, 0.0, 1.5]
], dtype=np.float32)

ARGS = {
    "data": data
}
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 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
instantiate_collection Not yet run
unverified (harvested, no recorded run)
rodem-hep/pc-jedi/src/hydra_utils.py
code served (permissive licence) · get_code("814934ff90988982")
How it ran

no invocation recorded.

plot_multi_hists Not yet run
unverified (harvested, no recorded run)
rodem-hep/pc-jedi/src/plotting.py
code served (permissive licence) · get_code("6466f297d91408a6")
How it ran

no invocation recorded.

Also present in other papers' code

Of 17 samples lifted from this paper's linked repositories, 17 carry a function identity and 0 do not. Of the 17 distinct functions with an identity, 2 are also present in other papers whose code we hold (0 with at least one paper dated earlier, 0 the same month, 2 later) and 15 are only here among the 52,977 papers whose code Syntology holds (as of 2026-09-26).

Only here means: among the 52,977 papers whose code Syntology holds (51,887 with lifted samples, 1,090 through their assemblies only), as of 2026-09-26, no other paper's lifted samples or current assembly bricks contain an identical function body after normalisation.

FunctionAlso present inEarliest-datedOthers named
cosine_encoding 1 other paper
1 later
2401.13537 (2024-01)
Masked Particle Modeling on Sets: Towards Self-Supervised High Energy Physics Foundation Models
instantiate_collection 1 other paper
1 later
2406.13074 (2024-06)
PIPPIN: Generating variable length full events from partons

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

In this paper, we present a new method to efficiently generate jets in High Energy Physics called PC-JeDi. This method utilises score-based diffusion models in conjunction with transformers which are well suited to the task of generating jets as particle clouds due to their permutation equivariance. PC-JeDi achieves competitive performance with current state-of-the-art methods across several metrics that evaluate the quality of the generated jets. Although slower than other models, due to the large number of forward passes required by diffusion models, it is still substantially faster than traditional detailed simulation. Furthermore, PC-JeDi uses conditional generation to produce jets with a desired mass and transverse momentum for two different particles, top quarks and gluons.

For agents

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

get_harvested_code_for_paper("2303.05376")
get_code_for_paper("2303.05376")
have("2303.05376")

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

curl https://syntology.ai/api/ran/2303.05376.json

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

[![Syntology run record](https://syntology.ai/api/ran/2303.05376.svg)](https://syntology.ai/paper/2303.05376)

Connect an agent — have() is free.