SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2602.12105 · 2026

Iskra: A System for Inverse Geometry Processing

Justin Solomon, Ana Dodik, Ahmed Mahmoud

arXiv · PDF · Open in the Atlas

Code that ran

10 of 12 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; 1 honoured the contract we drafted, 1 violated it and 0 have no contract verdict; 8 failures of our instrument; 2 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/2602.12105.json · badge: /api/ran/2602.12105.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
anadodik/iskra canonical (classifier label)
classifier basis: census
10 ran · 2 no recorded run
FunctionStatusWhere it lives
complex_nth_root Ran
ran; our contract draft was wrong, not the code
anadodik/iskra/iskra/directional.py
code served (permissive licence) · get_code("6f697048bb231ea1")
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 (4, 3) as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch

torch.manual_seed(0)

# Create a complex tensor with shape [...] 
# The function will add a dimension of size n to the output
z = torch.tensor([1.0 + 2.0j, 3.0 + 4.0j, -1.0 + 1.0j, 2.0 - 3.0j])

n = 3

ARGS = {
    "z": z,
    "n": n,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Tensor, ndim 2, dtype family float, shape (4, 3) — shape (4, 3) as a later run recorded it
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
hermite_cutoff Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
anadodik/iskra/iskra/mollifiers.py
code served (permissive licence) · get_code("116d7a7007a06158")
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 with values between 0 and 1 for testing the hermite_cutoff function
# The function uses torch operations and torch.where, so we need a valid torch.Tensor
x = torch.tensor([[-0.5, 0.2, 0.5, 0.8],
                   [1.2, 0.1, 0.9, 0.3]], dtype=torch.float32)

ARGS = {
    'x': x,
    'degree': 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 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
print_grad Ran
ran, honoured the contract we drafted
anadodik/iskra/iskra/signed_svd.py
code served (permissive licence) · get_code("cb3df4d661410a67")
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 print_grad function takes a single parameter x
# Looking at the implementation, x is passed to PrintGrad.apply()
# which expects a tensor that will be used in forward and backward passes
# x should be a tensor that requires gradients for backward to be called

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 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
print_identity Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
anadodik/iskra/iskra/_adjoint_debug_helpers.py
code served (permissive licence) · get_code("cdeba4fd2eaa69ab")
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 simple tensor for testing
x = torch.randn(2, 4, requires_grad=True)

# Create a string name for the tensor
name = "test_tensor"

# Construct the ARGS dictionary
ARGS = {
    "x": x,
    "name": name
}
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
remove_orange_nodes_from_dot Ran
ran; our contract draft was wrong, not the code
anadodik/iskra/iskra/_adjoint_debug_helpers.py
code served (permissive licence) · get_code("e081c7fa448561dc")
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 Digraph. 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.

from graphviz import Digraph

torch_manual_seed = __import__('torch').manual_seed(0)

# Create a minimal Digraph object with some nodes and edges
# that simulate a torchviz output
dot = Digraph(comment='test graph')

# Add some nodes with various attributes
dot.body.append('node1 [color="orange", fillcolor="orange"];')
dot.body.append('node2 [color="blue", fillcolor="white"];')
dot.body.append('node3 [color="orange"];')
dot.body.append('node4 [label="test"];')

# Add some edges
dot.body.append('node1 -> node2 [label="edge1"];')
dot.body.append('node2 -> node3;')
dot.body.append('node3 -> node4 [color="red"];')
dot.body.append('// this is a comment')
dot.body.append('node4 -> node1 [style="dotted"];')

ARGS = {
    'dot': dot
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
graphviz — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Digraph — a Digraph output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a Digraph 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
sk Ran
ran; our contract draft was wrong, not the code
anadodik/iskra/iskra/signed_svd.py
code served (permissive licence) · get_code("4a803fdda1dcaf47")
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, 4, 3, 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)

# The function sk takes a tensor v and extracts its last 3 dimensions as x, y, z
# Then constructs a skew-symmetric matrix from these components
# v should be a tensor where the last dimension has size 3

# Create a small tensor with shape (2, 4, 3) - batch of 2, sequence of 4, vector components of 3
v = torch.randn(2, 4, 3)

ARGS = {
    'v': v
}
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, 4, 3, 3) — shape (2, 4, 3, 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
smoothstep Ran
ran; our contract draft was wrong, not the code
anadodik/iskra/iskra/mollifiers.py
code served (permissive licence) · get_code("7d66b53e8e921bc7")
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 = {
    "x": torch.randn(2, 4),
    "start": 0.0,
    "end": 1.0,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Tensor, ndim 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
soft_clamp Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
anadodik/iskra/iskra/mollifiers.py
code served (permissive licence) · get_code("ad719793a133ac0b")
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 numpy as np

torch.manual_seed(0)

# Create a small tensor with values in the range [0, 1]
x = torch.rand(2, 4)

# Parameters for soft_clamp
sigma = 0.05
left = True
right = True

ARGS = {
    "x": x,
    "sigma": sigma,
    "left": left,
    "right": right,
}
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 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
to_extrinsic Ran
ran, violated the contract we drafted
anadodik/iskra/iskra/directional.py
code served (permissive licence) · get_code("958d974a605b3afb")
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 batch size 2, with complex directional tensor
# u should be Complex type with shape [B] or [B, N]
u = torch.tensor([1.0 + 2.0j, 3.0 + 4.0j], dtype=torch.complex64)

# tangents: shape [B, 3] - first basis vector
tangents = torch.randn(2, 3, dtype=torch.float32)

# binormals: shape [B, 3] - second basis vector
binormals = torch.randn(2, 3, dtype=torch.float32)

ARGS = {
    "u": u,
    "tangents": tangents,
    "binormals": binormals,
}
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
to_intrinsic Ran
ran; our contract draft was wrong, not the code
anadodik/iskra/iskra/directional.py
code served (permissive licence) · get_code("ce25a23ee6883c5d")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned Tensor, ndim 1, dtype family float, shape (2,) as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch

torch.manual_seed(0)

# Create small batch of 3D vectors
batch_size = 2
v = torch.randn(batch_size, 3, dtype=torch.float32)
tangents = torch.randn(batch_size, 3, dtype=torch.float32)
binormals = torch.randn(batch_size, 3, dtype=torch.float32)

ARGS = {
    "v": v,
    "tangents": tangents,
    "binormals": binormals,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Tensor, ndim 1, dtype family float, shape (2,) — shape (2,) as a later run recorded it
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
sample_circle Not yet run
unverified (harvested, no recorded run)
anadodik/iskra/iskra/random.py
code served (permissive licence) · get_code("0f0e422e3505dfce")
How it ran

no invocation recorded.

sample_sphere Not yet run
unverified (harvested, no recorded run)
anadodik/iskra/iskra/random.py
code served (permissive licence) · get_code("1d1068147bd71301")
How it ran

no invocation recorded.

Also present in other papers' code

Of 12 samples lifted from this paper's linked repositories, 12 carry a function identity and 0 do not. Of the 12 distinct functions with an identity, 0 are also present in other papers whose code we hold (0 with at least one paper dated earlier, 0 the same month, 0 later) and 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.

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

Fig. 1. Our system embeds geometry processing algorithms within larger differentiable pipelines. Users can optimize vertex positions to run a fairing flow backwards, a mesh's intrinsic metric to reduce area distortion during parameterization, or control handles to match a deformed mesh to a target pose. We propose a system for differentiating through solutions to geometry processing problems. Our system differentiates a broad class of geometric algorithms, exploiting existing fast problem-specific schemes common to geometry processing, including local-global and ADMM solvers. It is compatible with machine learning frameworks, opening doors to new classes of inverse geometry processing applications. We marry the scatter-gather approach to mesh processing with tensor-based workflows and rely on the adjoint method applied to user-specified imperative code to generate an efficient backward pass behind the scenes. We demonstrate our approach by differentiating through mean curvature flow, spectral conformal parameterization, geodesic distance computation, and as-rigid-as-possible deformation, examining usability and performance on these applications. Our system allows practitioners to differentiate through existing geometry processing algorithms without needing to reformulate them, resulting in low implementation effort, fast runtimes, and lower memory requirements than differentiable optimization tools not tailored to geometry processing.

For agents

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

get_harvested_code_for_paper("2602.12105")
get_code_for_paper("2602.12105")
have("2602.12105")

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

curl https://syntology.ai/api/ran/2602.12105.json

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

[![Syntology run record](https://syntology.ai/api/ran/2602.12105.svg)](https://syntology.ai/paper/2602.12105)

Connect an agent — have() is free.