SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2402.12535 · ICML · 2024

Locality-Sensitive Hashing-Based Efficient Point Transformer with Applications in High-Energy Physics

Pan Li, Javier Duarte, Mia Liu, Siqi Miao, Zhiyuan Lu

arXiv · PDF · Open in the Atlas

Code that ran

12 of 15 samples (functions and classes) lifted from this paper's linked repositories executed on a synthesized input in Syntology's jail (1 of them constructed an object rather than computing a result; 0 honoured the contract we drafted, 0 violated it and 1 has no contract verdict; 11 failures of our instrument; 3 no recorded run), measured 2026-11-22; repository links from the Papers with Code archive attributed under CC BY-SA, no affiliation; bodies inline only under a permissive licence as GitHub detects it. Not a reproduction of the paper's results.

"Ran" means the function executed on a synthesized input and returned a value. It is not a reproduction of the paper's results. This record, dated, as JSON: /api/ran/2402.12535.json · badge: /api/ran/2402.12535.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
Graph-COM/HEPT canonical (classifier label)
classifier basis: classifier_2025
5 ran · 1 no recorded run
graph-com/hept community
Papers with Code archive
7 ran · 2 no recorded run
FunctionStatusWhere it lives
E2LSH Ran
ran on a synthesized input · constructed an object; no result was computed
graph-com/hept/src/models/attention/hept.py
code served (permissive licence) · get_code("9647970031bdeb28")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned an instance of its own class, E2LSH: the run constructed the object and did not call it, so no result was computed. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
import torch.nn as nn

torch.manual_seed(0)

ARGS = {
    'n_hashes': 3,
    'n_heads': 2,
    'dim': 4,
    'r': 1
}
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 E2LSH — the run constructed an instance of its own class and did not call it; no shape is stated — the instrument coerces any output to an array; for a E2LSH 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
batched_index_select Ran
ran; our contract draft was wrong, not the code
graph-com/hept/src/models/attention/hept.py
code served (permissive licence) · get_code("171f7eb1327d57bd")
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
from einops import rearrange

torch.manual_seed(0)

# Create valid tensors for the function
# values: (1 or n_hashes, batch, seqlen, dim)
# indices: (n_hashes, batch, seqlen)

n_hashes = 2
batch = 3
seqlen = 4
dim = 5

# Create values tensor with shape (n_hashes, batch, seqlen, dim)
values = torch.randn(n_hashes, batch, seqlen, dim)

# Create indices tensor with shape (n_hashes, batch, seqlen)
# Indices should be valid positions for gathering along dimension -2 (seqlen dimension)
indices = torch.randint(0, seqlen, (n_hashes, batch, seqlen))

ARGS = {
    "values": values,
    "indices": indices,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
einops, 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
eval_one_batch Ran
ran; our contract draft was wrong, not the code
Graph-COM/HEPT/src/pileup_trainer.py
code served (permissive licence) · get_code("6bd0deb90b29ffea")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1) and returned 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

torch.manual_seed(0)

# Build a minimal model that returns embeddings
class SimpleModel(nn.Module):
    def __init__(self):
        super().__init__()
        self.linear = nn.Linear(4, 3)
    
    def forward(self, data):
        # Return embeddings of shape (batch_size, embedding_dim)
        return self.linear(data.x)

# Build a minimal data object with required attributes
class Data:
    def __init__(self):
        self.x = torch.randn(2, 4)  # batch_size=2, input_dim=4
        self.y = torch.randn(2, 1)  # batch_size=2, output_dim=1
        self.is_neu = torch.tensor([True, False])  # mask for neutral samples

# Build a minimal criterion (loss function)
criterion = nn.MSELoss()

# Build a minimal optimizer
model = SimpleModel()
optimizer = torch.optim.SGD(model.parameters(), lr=0.01)

# Build data object
data = Data()

# lr_s parameter (learning rate schedule or similar)
lr_s = 0.01

ARGS = {
    "model": model,
    "optimizer": optimizer,
    "criterion": criterion,
    "data": data,
    "lr_s": lr_s
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1
Imports
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
eval_one_batch Ran
ran; our contract draft was wrong, not the code
Graph-COM/HEPT/src/tracking_trainer.py
code served (permissive licence) · get_code("d2d3eabc93e4e2cd")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1) and returned 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

torch.manual_seed(0)

# Build a minimal model that returns embeddings
class MinimalModel(nn.Module):
    def __init__(self):
        super().__init__()
        self.linear = nn.Linear(4, 3)
    
    def forward(self, data):
        # Return embeddings of shape (batch_size, embedding_dim)
        return self.linear(data.features)

# Build a minimal criterion that accepts the signature used in the function
class MinimalCriterion(nn.Module):
    def forward(self, embeddings, point_pairs_index, particle_id, reconstructable, pt):
        # Return a scalar loss
        return torch.tensor(0.5, requires_grad=True)

# Build a minimal data object with required attributes
class MinimalData:
    def __init__(self):
        self.features = torch.randn(2, 4)  # batch_size=2, feature_dim=4
        self.point_pairs_index = torch.randint(0, 2, (2, 3))
        self.particle_id = torch.randint(0, 5, (2,))
        self.reconstructable = torch.randint(0, 2, (2,))
        self.pt = torch.randn(2, 1)

model = MinimalModel()
optimizer = torch.optim.SGD(model.parameters(), lr=0.01)
criterion = MinimalCriterion()
data = MinimalData()
lr_s = 0.001

ARGS = {
    "model": model,
    "optimizer": optimizer,
    "criterion": criterion,
    "data": data,
    "lr_s": lr_s
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1
Imports
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
invert_permutation Ran
ran; our contract draft was wrong, not the code
graph-com/hept/src/models/attention/hept.py
code served (permissive licence) · get_code("40497e2f00847422")
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 int, 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

torch.manual_seed(0)

# Create a valid permutation tensor
# A permutation is a tensor containing each index from 0 to n-1 exactly once
# Shape: (batch, n) where batch=2, n=4
perm = torch.tensor([
    [0, 2, 1, 3],
    [3, 1, 0, 2]
], dtype=torch.long)

ARGS = {
    "perm": perm
}
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 int, 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
lsh_mapping Ran
ran; our contract draft was wrong, not the code
graph-com/hept/src/models/attention/hept.py
code served (permissive licence) · get_code("871d392728fe3f2f")
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

torch.manual_seed(0)

class SimpleLSH(nn.Module):
    """Minimal LSH module that hashes input tensors to integer values"""
    def __init__(self, input_dim, num_hashes):
        super().__init__()
        self.num_hashes = num_hashes
        self.register_buffer('hash_matrix', torch.randn(input_dim, num_hashes))
    
    def forward(self, x):
        # x shape: (batch, seq_len, input_dim)
        # output shape: (batch, seq_len, num_hashes)
        hashed = torch.matmul(x, self.hash_matrix)
        return hashed.round()

# Create the LSH module
e2lsh = SimpleLSH(input_dim=4, num_hashes=3)

# Create queries and keys tensors
# Shape: (batch_size=2, seq_len=5, embedding_dim=4)
queries = torch.randn(2, 5, 4)
keys = torch.randn(2, 5, 4)

ARGS = {
    'e2lsh': e2lsh,
    'queries': queries,
    'keys': keys
}
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 — 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
qkv_res Ran
ran; our contract draft was wrong, not the code
Graph-COM/HEPT/example/hept.py
code served (permissive licence) · get_code("190ac0f8a58e671b")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1) and returned 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

torch.manual_seed(0)

# s_query: shape (..., i, d) - query sequence
# s_key: shape (..., j, d) - key sequence
# s_value: shape (..., j, d) - value sequence
# The function performs attention-like operations with RBF kernels

# Create small tensors with distinct dimensions
batch_size = 2
query_len = 3
key_len = 4
d_model = 5

s_query = torch.randn(batch_size, query_len, d_model)
s_key = torch.randn(batch_size, key_len, d_model)
s_value = torch.randn(batch_size, key_len, d_model)

ARGS = {
    "s_query": s_query,
    "s_key": s_key,
    "s_value": s_value,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1
Imports
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
sort_to_buckets Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
graph-com/hept/src/models/attention/hept.py
code served (permissive licence) · get_code("61ca2f5e8315764c")
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 5, dtype family float, shape (2, 2, 2, 4, 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
from einops import rearrange

torch.manual_seed(0)

# The function sort_to_buckets takes:
# - x: a tensor that gets rearranged to (1, b, s, d)
# - perm: indices tensor used for batched_index_select, shape (n_hashes, batch, seqlen)
# - bucketsz: an integer for bucket size

# Let's construct valid arguments:
batch_size = 2
seq_len = 8
dim = 4
n_hashes = 2
bucket_sz = 4

# x: input tensor (b, s, d) - will be rearranged to (1, b, s, d) inside the function
x = torch.randn(batch_size, seq_len, dim)

# perm: indices tensor (n_hashes, batch, seqlen) with values in range [0, seqlen)
# These are permutation indices for selecting along sequence dimension
perm = torch.randint(0, seq_len, (n_hashes, batch_size, seq_len))

# bucketsz: integer specifying bucket size
# Must divide evenly into seqlen for the rearrange operation
bucketsz = bucket_sz

ARGS = {
    'x': x,
    'perm': perm,
    'bucketsz': bucketsz
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
einops, torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Tensor, ndim 5, dtype family float, shape (2, 2, 2, 4, 4) — shape (2, 2, 2, 4, 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
train_one_batch Ran
ran; our contract draft was wrong, not the code
Graph-COM/HEPT/src/tracking_trainer.py
code served (permissive licence) · get_code("79fe4ecdc280014a")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1) and returned 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
from torch.optim.lr_scheduler import LambdaLR

torch.manual_seed(0)

# Define a minimal model that takes data and returns embeddings
class SimpleEmbeddingModel(nn.Module):
    def __init__(self):
        super().__init__()
        self.linear = nn.Linear(4, 3)
    
    def forward(self, data):
        # Return embeddings of shape (batch_size, embedding_dim)
        return self.linear(data.features)

# Define a minimal criterion that matches the function's usage
class CustomLoss(nn.Module):
    def forward(self, embeddings, point_pairs_index, particle_id, reconstructable, pt):
        # Simple loss computation using the provided tensors
        return torch.mean(embeddings) + torch.mean(particle_id.float())

# Define a minimal data object with required attributes
class Data:
    def __init__(self):
        self.features = torch.randn(2, 4)  # batch_size=2, feature_dim=4
        self.point_pairs_index = torch.tensor([[0, 1], [1, 0]], dtype=torch.long)
        self.particle_id = torch.tensor([1, 2], dtype=torch.long)
        self.reconstructable = torch.tensor([True, False])
        self.pt = torch.tensor([0.5, 1.5])

# Create instances
model = SimpleEmbeddingModel()
optimizer = torch.optim.SGD(model.parameters(), lr=0.01)
criterion = CustomLoss()
data = Data()

# Create a LambdaLR scheduler
lr_scheduler = LambdaLR(optimizer, lr_lambda=lambda epoch: 0.95 ** epoch)

ARGS = {
    "model": model,
    "optimizer": optimizer,
    "criterion": criterion,
    "data": data,
    "lr_s": lr_scheduler,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1
Imports
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
train_one_batch Ran
ran; our contract draft was wrong, not the code
Graph-COM/HEPT/src/pileup_trainer.py
code served (permissive licence) · get_code("712687f42d50cadf")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1) and returned 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
from torch.optim.lr_scheduler import LambdaLR

torch.manual_seed(0)

# Define a minimal model that returns embeddings
class SimpleModel(nn.Module):
    def __init__(self):
        super().__init__()
        self.linear = nn.Linear(4, 3)
    
    def forward(self, data):
        # Return embeddings of shape (batch_size, 3)
        return self.linear(data.x)

# Define a minimal data object with required attributes
class Data:
    def __init__(self):
        self.x = torch.randn(2, 4)  # batch_size=2, input_dim=4
        self.y = torch.randn(2, 1)  # batch_size=2, output_dim=1
        self.is_neu = torch.tensor([True, False])  # mask for which samples to use
    
    def __call__(self, *args, **kwargs):
        # Allow model(data) to work by returning x
        return self.x

# Create instances
model = SimpleModel()
optimizer = torch.optim.SGD(model.parameters(), lr=0.01)
criterion = nn.MSELoss()
data = Data()

# Create a LambdaLR scheduler
lr_scheduler = LambdaLR(optimizer, lr_lambda=lambda epoch: 0.95 ** epoch)

ARGS = {
    "model": model,
    "optimizer": optimizer,
    "criterion": criterion,
    "data": data,
    "lr_s": lr_scheduler
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1
Imports
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
uniform Ran
ran; our contract draft was wrong, not the code
graph-com/hept/src/models/attention/hept.py
code served (permissive licence) · get_code("4a3d3d2dcce1a90a")
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

torch.manual_seed(0)

ARGS = {
    'a': 0.0,
    'b': 1.0,
    'shape': (2, 4),
    'device': 'cpu'
}
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
unsort_from_buckets Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
graph-com/hept/src/models/attention/hept.py
code served (permissive licence) · get_code("7474f796cb45df30")
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, 20, 6) 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
from einops import rearrange

torch.manual_seed(0)

# Create s_x with shape (n_hashes, batch, nbuckets, bucketsz, dim)
# Using distinct sizes: h=2, b=3, nbuckets=4, bucketsz=5, d=6
n_hashes = 2
batch = 3
nbuckets = 4
bucketsz = 5
dim = 6

s_x = torch.randn(n_hashes, batch, nbuckets, bucketsz, dim)

# Create perm_inverse with shape (n_hashes, batch, seqlen)
# seqlen should equal nbuckets * bucketsz = 20
seqlen = nbuckets * bucketsz
perm_inverse = torch.randperm(seqlen).unsqueeze(0).unsqueeze(0).expand(n_hashes, batch, -1)

ARGS = {
    "s_x": s_x,
    "perm_inverse": perm_inverse,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
einops, 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, 20, 6) — shape (2, 3, 20, 6) 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
HEPTAttention Not yet run
unverified (harvested, no recorded run)
graph-com/hept/src/models/attention/hept.py
code served (permissive licence) · get_code("480184db76bc335c")
How it ran

no invocation recorded.

get_geo_shift Not yet run
unverified (harvested, no recorded run)
graph-com/hept/src/models/attention/hept.py
code served (permissive licence) · get_code("f2bf68b5d3153051")
How it ran

no invocation recorded.

prep_qk Not yet run
unverified (harvested, no recorded run)
Graph-COM/HEPT/example/hept.py
code served (permissive licence) · get_code("7524f0a2b7ec2b68")
How it ran

no invocation recorded.

Also present in other papers' code

Of 15 samples lifted from this paper's linked repositories, 15 carry a function identity and 0 do not. Of the 15 distinct functions with an identity, 1 are also present in other papers whose code we hold (1 with at least one paper dated earlier, 0 the same month, 0 later) and 14 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
uniform
brick of this paper's assembly: helper
1 other paper
1 dated earlier
2111.12840 (2021-11)
Explaining machine-learned particle-flow reconstruction

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

This study introduces a novel transformer model optimized for large-scale point cloud processing in scientific domains such as high-energy physics (HEP) and astrophysics. Addressing the limitations of graph neural networks and standard transformers, our model integrates local inductive bias and achieves near-linear complexity with hardware-friendly regular operations. One contribution of this work is the quantitative analysis of the error-complexity tradeoff of various sparsification techniques for building efficient transformers. Our findings highlight the superiority of using locality-sensitive hashing (LSH), especially OR & AND-construction LSH, in kernel approximation for large-scale point cloud data with local inductive bias. Based on this finding, we propose LSH-based Efficient Point Transformer (HEPT), which combines E 2 LSH with OR & AND constructions and is built upon regular computations. HEPT demonstrates remarkable performance on two critical yet time-consuming HEP tasks, significantly outperforming existing GNNs and transformers in accuracy and computational speed, marking a significant advancement in geometric deep learning and large-scale scientific data processing. Our code is available at https: //github.com/Graph-COM/HEPT.

For agents

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

get_harvested_code_for_paper("2402.12535")
get_code_for_paper("2402.12535")
have("2402.12535")

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

curl https://syntology.ai/api/ran/2402.12535.json

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

[![Syntology run record](https://syntology.ai/api/ran/2402.12535.svg)](https://syntology.ai/paper/2402.12535)

Connect an agent — have() is free.