SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2602.03264 · 2026

HypCBC: Domain-Invariant Hyperbolic Cross-Branch Consistency for Generalizable Medical Image Analysis

Sebastian Doerrich, Francesco Salvo, Christian Ledig, Jonas Alle

arXiv · PDF · Open in the Atlas

Code that ran

14 of 18 samples (functions and classes) lifted from this paper's linked repositories executed on a synthesized input in Syntology's jail (3 of them constructed an object rather than computing a result; 6 honoured the contract we drafted, 0 violated it and 3 have no contract verdict; 5 failures of our instrument; 4 no recorded run), measured 2026-11-22; 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/2602.03264.json · badge: /api/ran/2602.03264.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
francescodisalvo05/hyperbolic-cross-branch-consistency community
Papers with Code archive
14 ran · 4 no recorded run
FunctionStatusWhere it lives
ModelConfig Ran
ran on a synthesized input · constructed an object; no result was computed
francescodisalvo05/hyperbolic-cross-branch-consistency/hypcbc/model/model.py
code served (permissive licence) · get_code("908199f880ef1d73")
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, ModelConfig: 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 numpy as np
from pathlib import Path
from typing import Optional

torch.manual_seed(0)

ARGS = {
    'backbone_id': 'dinov2_small',
    'freeze_backbone': True,
    'device': None,
    'manifold': 'hyp',
    'branch1_dim': 128,
    'branch2_dim': 256,
    'num_classes': 10,
    'hyp_curvature': 1.0,
    'hyp_clip_r': 1.0,
    'extract_projections': False,
}
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 ModelConfig — 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 ModelConfig 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
RiemannianGradient Ran
ran on a synthesized input · constructed an object; no result was computed
francescodisalvo05/hyperbolic-cross-branch-consistency/hypcbc/model/model.py
code served (permissive licence) · get_code("11123b6c215df399")
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, RiemannianGradient: 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)

# The RiemannianGradient.forward expects:
# - ctx: autograd context (automatically provided by autograd)
# - x: a tensor
#
# The backward uses:
# - grad_output: gradient tensor (automatically provided)
# - x: saved tensor from forward
# - Operations: pow(2), sum(-1, keepdim=True), pow(2), multiplication
#
# So x should be a small tensor with values that won't cause numerical issues

x = torch.randn(2, 4, requires_grad=True)

ARGS = {
    "x": x,
}
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 RiemannianGradient — 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 RiemannianGradient 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
ToPoincare Ran
ran on a synthesized input · constructed an object; no result was computed
francescodisalvo05/hyperbolic-cross-branch-consistency/hypcbc/model/model.py
code served (permissive licence) · get_code("d5e88944d10f2c9b")
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, ToPoincare: 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 = {
    "c": 1.0,
    "train_c": False,
    "train_x": False,
    "ball_dim": None,
    "riemannian": True,
    "clip_r": None,
}
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 ToPoincare — 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 ToPoincare 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
_expmap Ran
ran, honoured the contract we drafted
francescodisalvo05/hyperbolic-cross-branch-consistency/hypcbc/model/model.py
code served (permissive licence) · get_code("fdedecef0f50c5a5")
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)

# Create small test tensors
# x: base point in hyperbolic space, shape (batch=2, dim=4)
x = torch.randn(2, 4)

# u: tangent vector at x, shape (batch=2, dim=4)
u = torch.randn(2, 4)

# c: curvature parameter (scalar)
c = torch.tensor(0.5)

ARGS = {
    "x": x,
    "u": u,
    "c": c,
}
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
_expmap0 Ran
ran; our contract draft was wrong, not the code
francescodisalvo05/hyperbolic-cross-branch-consistency/hypcbc/model/model.py
code served (permissive licence) · get_code("fb6894c383135715")
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)

# u is a tensor that will have .norm() called on it with dim=-1, p=2, keepdim=True
# It should be a small tensor with batch dimension and feature dimension
u = torch.randn(2, 4)

# c is a scalar curvature parameter used in computations
c = 0.5

ARGS = {
    "u": u,
    "c": c,
}
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
_lambda_x Ran
ran; our contract draft was wrong, not the code
francescodisalvo05/hyperbolic-cross-branch-consistency/hypcbc/model/model.py
code served (permissive licence) · get_code("ece8d3c8f5b02f62")
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
import numpy as np

torch.manual_seed(0)

# Analyze the function:
# - x: a tensor that supports .pow(2), .sum(-1, keepdim=...), and arithmetic
# - c: a scalar or tensor used in multiplication with x.pow(2).sum(...)
# - keepdim: boolean flag (default False)

# The function computes: 2 / (1 - c * x.pow(2).sum(-1, keepdim=keepdim))
# x.pow(2).sum(-1, ...) sums over the last dimension

# Create a small tensor with batch size 2 and feature dim 4
x = torch.randn(2, 4)

# c should be a scalar that keeps the denominator non-zero
# Let's use c = 0.5 so that (1 - c * sum) stays positive
c = 0.5

# keepdim is a boolean
keepdim = False

ARGS = {
    "x": x,
    "c": c,
    "keepdim": keepdim,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Tensor, ndim 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
_mobius_add Ran
ran, honoured the contract we drafted
francescodisalvo05/hyperbolic-cross-branch-consistency/hypcbc/model/model.py
code served (permissive licence) · get_code("c1ccd028f0589c2b")
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
import torch.nn as nn

torch.manual_seed(0)

# Create small tensors for x and y
# Shape: (batch=2, dim=4)
x = torch.randn(2, 4)
y = torch.randn(2, 4)

# c is a curvature parameter (scalar)
c = torch.tensor(0.5)

ARGS = {
    "x": x,
    "y": y,
    "c": c,
}
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
_mobius_addition_batch Ran
ran; our contract draft was wrong, not the code
francescodisalvo05/hyperbolic-cross-branch-consistency/hypcbc/model/model.py
code served (permissive licence) · get_code("589cc8bd3ab7016a")
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, 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)

# x: input tensor, used for:
#   - _tensor_dot(x, y) -> einsum 'ij,kj->ik'
#   - x.pow(2).sum(-1, keepdim=True) -> x2
#   - x.unsqueeze(1) and x.unsqueeze(2)
# Shape should be (batch_size, feature_dim)
x = torch.randn(2, 4)

# y: input tensor, used for:
#   - _tensor_dot(x, y) -> einsum 'ij,kj->ik'
#   - y.pow(2).sum(-1, keepdim=True) -> y2
#   - y used directly in computation
# Shape should be (batch_size, feature_dim) - using different batch size for distinction
y = torch.randn(3, 4)

# c: scalar curvature parameter, used for:
#   - multiplication with xy, x2, y2
#   - exponentiation c ** 2
c = 0.5

ARGS = {
    'x': x,
    'y': y,
    'c': c,
}
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, 4) — shape (2, 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
_project Ran
ran; our contract draft was wrong, not the code
francescodisalvo05/hyperbolic-cross-branch-consistency/hypcbc/model/model.py
code served (permissive licence) · get_code("1721ac2e42c9e61d")
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)

# x: a tensor that will have norm computed along last dimension
# The function calls x.norm(dim=-1, keepdim=True, p=2) so x needs to be at least 2D
x = torch.randn(2, 4)

# c: a scalar value used in the computation
# The function computes (1 - 0.001) / c ** 0.5, so c should be a positive number
c = 2.0

ARGS = {
    "x": x,
    "c": c,
}
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
_tensor_dot Ran
ran, honoured the contract we drafted
francescodisalvo05/hyperbolic-cross-branch-consistency/hypcbc/model/model.py
code served (permissive licence) · get_code("e553a87d2b4d64d1")
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, 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)

# Create two 2D tensors with distinct dimensions
# x: shape (2, 4) - batch size 2, feature dim 4
# y: shape (3, 4) - batch size 3, feature dim 4
# The einsum 'ij,kj->ik' will produce output of shape (2, 3)

x = torch.randn(2, 4)
y = torch.randn(3, 4)

ARGS = {
    'x': x,
    'y': y
}
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, 3) — shape (2, 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
expmap Ran
ran, honoured the contract we drafted
francescodisalvo05/hyperbolic-cross-branch-consistency/hypcbc/model/model.py
code served (permissive licence) · get_code("0b29f3a43a5cfdbc")
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)

# Create small tensors for the Poincaré ball model
# x: starting point on poincare ball (batch_size=2, dim=4)
x = torch.randn(2, 4) * 0.1  # Keep small to stay in valid poincare ball region

# u: speed/tangent vector on poincare ball (batch_size=2, dim=4)
u = torch.randn(2, 4) * 0.1

# c: curvature parameter (scalar)
c = 1.0

ARGS = {
    "x": x,
    "u": u,
    "c": c,
}
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
expmap0 Ran
ran, honoured the contract we drafted
francescodisalvo05/hyperbolic-cross-branch-consistency/hypcbc/model/model.py
code served (permissive licence) · get_code("e1a6df60b9ec7a8d")
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)

# The function expmap0 takes:
# - u: a tensor (speed vector on poincare ball)
# - c: float|tensor (ball negative curvature, keyword-only argument with default 1.0)

# Create a small tensor for u with shape (batch=2, dim=4)
u = torch.randn(2, 4, dtype=torch.float32)

# c can be a float or tensor - use a float with default value
c = 1.0

ARGS = {
    "u": u,
    "c": c,
}
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
project Ran
ran, honoured the contract we drafted
francescodisalvo05/hyperbolic-cross-branch-consistency/hypcbc/model/model.py
code served (permissive licence) · get_code("85d8ee3dda139d04")
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)

# Create a small tensor on the Poincaré ball
# Shape: (batch=2, dim=4) - a point in 4D space
x = torch.randn(2, 4, dtype=torch.float32)

# Curvature parameter - can be float or tensor
c = 1.0

ARGS = {
    "x": x,
    "c": c,
}
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
tanh Ran
ran; our contract draft was wrong, not the code
francescodisalvo05/hyperbolic-cross-branch-consistency/hypcbc/model/model.py
code served (permissive licence) · get_code("be5af45dd86c2cc1")
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
import torch.nn as nn

torch.manual_seed(0)

# The function takes x (a tensor) and clamp (an int)
# It calls x.clamp(-clamp, clamp).tanh()
# So x must be a tensor with clamp and tanh methods

x = torch.randn(2, 4)
clamp_value = 15

ARGS = {
    "x": x,
    "clamp": clamp_value
}
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
BaseConfig Not yet run
unverified (harvested, no recorded run)
francescodisalvo05/hyperbolic-cross-branch-consistency/hypcbc/model/model.py
code served (permissive licence) · get_code("68b719d12af8320b")
How it ran

no invocation recorded.

HyperbolicMLR Not yet run
unverified (harvested, no recorded run)
francescodisalvo05/hyperbolic-cross-branch-consistency/hypcbc/model/model.py
code served (permissive licence) · get_code("34ae3c5207520dda")
How it ran

no invocation recorded.

ModelModule Not yet run
unverified (harvested, no recorded run)
francescodisalvo05/hyperbolic-cross-branch-consistency/hypcbc/model/model.py
code served (permissive licence) · get_code("6b9b964f6e1edbfd")
How it ran

no invocation recorded.

hyperbolic_softmax Not yet run
unverified (harvested, no recorded run)
francescodisalvo05/hyperbolic-cross-branch-consistency/hypcbc/model/model.py
code served (permissive licence) · get_code("cebc11a24161781c")
How it ran

no invocation recorded.

Also present in other papers' code

Of 18 samples lifted from this paper's linked repositories, 18 carry a function identity and 0 do not. Of the 18 distinct functions with an identity, 6 are also present in other papers whose code we hold (6 with at least one paper dated earlier, 0 the same month, 0 later) and 12 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
tanh
brick of this paper's assembly: helper
4 other papers
4 dated earlier · 1 only as an assembly brick
1904.02239 (2019-04)
Hyperbolic Image Embeddings
2108.02390 (2021-08)
Fuzzy Logic Based Logical Query Answering on Knowledge Graphs
2203.10833 (2022-03)
Hyperbolic Vision Transformers: Combining Improvements in Metric Learning
and 1 more not named here
_expmap0
brick of this paper's assembly: helper
1 other paper
1 dated earlier · 1 only as an assembly brick
2203.10833 (2022-03)
Hyperbolic Vision Transformers: Combining Improvements in Metric Learning
_mobius_add
brick of this paper's assembly: helper
1 other paper
1 dated earlier · 1 only as an assembly brick
2203.10833 (2022-03)
Hyperbolic Vision Transformers: Combining Improvements in Metric Learning
_project
brick of this paper's assembly: helper
1 other paper
1 dated earlier · 1 only as an assembly brick
2203.10833 (2022-03)
Hyperbolic Vision Transformers: Combining Improvements in Metric Learning
expmap0
brick of this paper's assembly: helper
1 other paper
1 dated earlier · 1 only as an assembly brick
2203.10833 (2022-03)
Hyperbolic Vision Transformers: Combining Improvements in Metric Learning
project
brick of this paper's assembly: helper
1 other paper
1 dated earlier · 1 only as an assembly brick
2203.10833 (2022-03)
Hyperbolic Vision Transformers: Combining Improvements in Metric Learning

Presence, not direction: an identical function body in two papers' repositories does not say who wrote it, which came from which, or that either paper's method rests on it. Matching is exact after normalisation, so a renamed or edited function is not matched. Dates are when papers appeared, not when code was written. Only .py files the harvester lifted are covered; notebooks are not read. A new-style arXiv id YYMM.NNNN dates a paper to the month 20YY-MM; otherwise Paper.year dates it to the year; otherwise it is undated. Dates are when the paper appeared, not when the code was written. Read 2026-11-22.

Repositories linked to this paper

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

Abstract

Robust generalization beyond training distributions remains a critical challenge for deep neural networks. This is especially pronounced in medical image analysis, where data is often scarce and covariate shifts arise from different hardware devices, imaging protocols, and heterogeneous patient populations. These factors collectively hinder reliable performance and slow down clinical adoption. Despite recent progress, existing learning paradigms primarily rely on the Euclidean manifold, whose flat geometry fails to capture the complex, hierarchical structures present in clinical data. In this work, we exploit the advantages of hyperbolic manifolds to model complex data characteristics. We present the first comprehensive validation of hyperbolic representation learning for medical image analysis and demonstrate statistically significant gains across eleven in-distribution datasets and three ViT models. We further propose an unsupervised, domain-invariant hyperbolic cross-branch consistency constraint. Extensive experiments confirm that our proposed method promotes domain-invariant features and outperforms state-of-the-art Euclidean methods by an average of +2.1% AUC on three domain generalization benchmarks: Fitzpatrick17k, Camelyon17-WILDS, and a cross-dataset setup for retinal imaging. These datasets span different imaging modalities, data sizes, and label granularities, confirming generalization capabilities across substantially different conditions. The code is available at github.com/francescodisalvo05/hyperbolic-cross-branch-consistency.

For agents

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

get_harvested_code_for_paper("2602.03264")
get_code_for_paper("2602.03264")
have("2602.03264")

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

curl https://syntology.ai/api/ran/2602.03264.json

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

[![Syntology run record](https://syntology.ai/api/ran/2602.03264.svg)](https://syntology.ai/paper/2602.03264)

Connect an agent — have() is free.