SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2604.20744 · 2026

AAC: Admissible-by-Architecture Differentiable Landmark Compression for ALT

An Le, Vien Ngo

arXiv · PDF · Open in the Atlas

Code that ran

14 of 15 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; 5 honoured the contract we drafted, 0 violated it and 0 have no contract verdict; 9 failures of our instrument; 1 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/2604.20744.json · badge: /api/ran/2604.20744.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
anindex/aac canonical (classifier label)
classifier basis: census
14 ran · 1 no recorded run
FunctionStatusWhere it lives
alt_memory_bytes Ran
ran, honoured the contract we drafted
anindex/aac/src/aac/baselines/alt.py
code served (permissive licence) · get_code("61c68458c64ac988")
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 int, ndim 0, dtype family int, shape () as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
import numpy as np

torch.manual_seed(0)

ARGS = {
    "num_landmarks": 100,
    "dtype_size": 4,
    "is_directed": True,
}
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 int, ndim 0, dtype family int, shape () — shape () as a later run recorded it
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
build_grid_edge_index Ran
ran, honoured the contract we drafted
anindex/aac/src/aac/contextual/encoders.py
code served (permissive licence) · get_code("a70a9e72b16248de")
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 2, 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)

# The function build_grid_edge_index takes a single parameter: grid_size (int)
# It builds edge indices for an 8-connected grid graph of size grid_size x grid_size
# The function doesn't actually use any complex objects, just an integer

ARGS = {
    "grid_size": 3,
}
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 2, 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
cdh_memory_bytes Ran
ran, honoured the contract we drafted
anindex/aac/src/aac/baselines/cdh.py
code served (permissive licence) · get_code("529f752c5193f82d")
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 int, ndim 0, dtype family int, shape () as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import math

def _pivot_index_bytes(num_pivots: int) -> int:
    """Smallest integer width (in bytes) that can index ``num_pivots``."""
    if num_pivots <= 1:
        return 1
    return max(1, math.ceil(math.log2(num_pivots) / 8))

def cdh_memory_bytes(num_pivots: int, num_stored: int, *, dtype_size: int=4, is_directed: bool=False) -> int:
    """Per-vertex deployed bytes: r distances plus r pivot indices.

    Directed graphs store a second r distances for the backward table; the
    index array is shared between the two.
    """
    index_size = _pivot_index_bytes(num_pivots)
    dist_factor = 2 if is_directed else 1
    return num_stored * (dist_factor * dtype_size + index_size)


ARGS = {
    "num_pivots": 16,
    "num_stored": 100,
    "dtype_size": 4,
    "is_directed": False,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
none outside the standard library — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type int, ndim 0, dtype family int, shape () — shape () as a later run recorded it
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
cell_costs_to_edge_weights Ran
ran; our contract draft was wrong, not the code
anindex/aac/src/aac/contextual/encoders.py
code served (permissive licence) · get_code("6cb32e29ad415487")
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, 84) 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)

# Parameters for cell_costs_to_edge_weights
# grid_size: int - size of the grid (e.g., 4x4 grid)
# cell_costs: torch.Tensor - shape (B, H, W) where B is batch size, H and W are grid dimensions

grid_size = 4
batch_size = 2

# Create cell_costs tensor with shape (batch_size, grid_size, grid_size)
# Values should be positive to match the docstring expectation
cell_costs = torch.rand(batch_size, grid_size, grid_size, dtype=torch.float32) + 0.1

ARGS = {
    "cell_costs": cell_costs,
    "grid_size": grid_size,
}
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, 84) — shape (2, 84) 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
combine_across_seeds Ran
ran; our contract draft was wrong, not the code
anindex/aac/src/aac/stats.py
code served (permissive licence) · get_code("dfd6c453a464bbb0")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned tuple, ndim 1, dtype family 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 numpy as np
from collections.abc import Sequence

ARGS = {
    'pvalues': [0.01, 0.05, 0.001, 0.1],
    'method': 'stouffer',
    'independent_queries': True,
    'effects': None,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, scipy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type tuple, ndim 1, dtype family 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
contextual_loss Ran
ran; our contract draft was wrong, not the code
anindex/aac/src/aac/contextual/loss.py
code served (permissive licence) · get_code("18b2e15a34e376f7")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned Tensor, ndim 0, dtype family float, shape () as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
import torch.nn as nn

torch.manual_seed(0)

# Create a minimal compressor module that implements condition_regularization()
class MinimalCompressor(nn.Module):
    def __init__(self):
        super().__init__()
        self.linear = nn.Linear(4, 4)
    
    def condition_regularization(self):
        # Return a scalar tensor representing the condition regularization
        return torch.tensor(0.05)

# Construct the arguments
d_true = torch.randn(2)  # (B,) batch of 2
h_smooth = torch.randn(2)  # (B,) batch of 2
compressor = MinimalCompressor()
aux_loss_val = torch.tensor(0.1)
alpha_aux = 0.5
cond_lambda = 0.01

ARGS = {
    "d_true": d_true,
    "h_smooth": h_smooth,
    "compressor": compressor,
    "aux_loss_val": aux_loss_val,
    "alpha_aux": alpha_aux,
    "cond_lambda": cond_lambda,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Tensor, ndim 0, dtype family float, shape () — shape () as a later run recorded it
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
direction_counts Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
anindex/aac/src/aac/selection.py
code served (permissive licence) · get_code("bce434f4f78e968d")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned tuple, ndim 1, dtype family int; no output shape is stated. The image it ran in was not recorded. If your copy returns the same type and ndim on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
from collections.abc import Sequence

Item = tuple[int, int]

def direction_counts(items: Sequence[Item], chosen: Sequence[int]) -> tuple[int, int]:
    """``(forward, backward)`` slots a selection deploys; free split may zero one."""
    directions = [items[i][1] for i in chosen]
    return (directions.count(0), directions.count(1))

torch.manual_seed(0)

items = [
    (10, 0),
    (20, 1),
    (30, 0),
    (40, 1),
    (50, 0),
]

chosen = [0, 2, 4]

ARGS = {
    "items": items,
    "chosen": chosen,
}
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 1, dtype family int — a tuple output has no array shape of its own; none is stated
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
fastmap_memory_bytes Ran
ran, honoured the contract we drafted
anindex/aac/src/aac/baselines/fastmap.py
code served (permissive licence) · get_code("cb2135d7f129f35f")
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 int, ndim 0, dtype family int, shape () as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
import numpy as np

# The function fastmap_memory_bytes takes two parameters:
# - num_dims: int - number of dimensions
# - dtype_size: int - size of data type in bytes (default 4)
# The function simply multiplies these two values.

torch.manual_seed(0)

ARGS = {
    "num_dims": 8,
    "dtype_size": 4
}
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 int, ndim 0, dtype family int, shape () — shape () as a later run recorded it
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
fdr_reject Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
anindex/aac/src/aac/stats.py
code served (permissive licence) · get_code("d2b8aba77bbed8b8")
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 2, 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 numpy as np
from scipy import stats

# Construct test p-values
pvalues = [0.001, 0.008, 0.039, 0.041, 0.042, 0.06, 0.074, 0.205]

# Construct arguments dict
ARGS = {
    'pvalues': pvalues,
    'q': 0.05
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, scipy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type tuple, ndim 2, 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
make_fastmap_heuristic Ran
ran; our contract draft was wrong, not the code
anindex/aac/src/aac/baselines/fastmap.py
code served (permissive licence) · get_code("215328d2ffe01e69")
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 function. 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

torch.manual_seed(0)

# Create a small FastMap coordinates tensor
# Shape: (V=2 vertices, num_dims=4)
coords = torch.randn(2, 4, dtype=torch.float32)

# Set up the arguments
ARGS = {
    "coords": coords,
    "dtype": None,
    "conservative": None,
}
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 function — a function output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a function 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
objective Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
anindex/aac/src/aac/selection.py
code served (permissive licence) · get_code("0c8d1fc0f878dbab")
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 float, ndim 0, dtype family float, shape () as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import numpy as np
from collections.abc import Sequence

# Construct bounds: a 2D numpy array where rows are different bound values
# and columns represent different queries/dimensions
bounds = np.array([
    [1.5, -0.5, 2.0, 1.2],
    [-1.0, 0.8, -2.5, 0.3],
    [0.5, 1.5, 0.8, -0.9]
], dtype=np.float32)

# Construct chosen: a sequence of indices to select from bounds
chosen = [0, 2]

ARGS = {
    "bounds": bounds,
    "chosen": chosen
}
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 float, ndim 0, dtype family float, shape () — shape () as a later run recorded it
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
parameter_shape Ran
ran, honoured the contract we drafted
anindex/aac/src/aac/compression/selectors.py
code served (permissive licence) · get_code("840f98c140562db7")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned tuple, ndim 1, dtype family int; no output shape is stated. The image it ran in was not recorded. If your copy returns the same type and ndim on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch

torch.manual_seed(0)

ARGS = {
    'name': 'row_gumbel',
    'num_select': 3,
    'num_pool': 5,
}
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 1, dtype family int — a tuple output has no array shape of its own; none is stated
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
row_gumbel Ran
ran; our contract draft was wrong, not the code
anindex/aac/src/aac/compression/selectors.py
code served (permissive licence) · get_code("cbb34500d0fb61ee")
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 logits tensor: shape (2, 4) meaning 2 rows, 4 classes
logits = torch.randn(2, 4, dtype=torch.float32)

# Create a generator for reproducibility
generator = torch.Generator()
generator.manual_seed(42)

ARGS = {
    "logits": logits,
    "tau": 1.0,
    "hard": True,
    "generator": generator,
}
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
successive_softmax Ran
ran; our contract draft was wrong, not the code
anindex/aac/src/aac/compression/selectors.py
code served (permissive licence) · get_code("502ef0916aa805e4")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned Tensor, ndim 2, dtype family float, shape (3, 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

torch.manual_seed(0)

ARGS = {
    "scores": torch.randn(5),
    "num_select": 3,
    "tau": 1.0,
    "hard": False,
    "generator": 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 Tensor, ndim 2, dtype family float, shape (3, 5) — shape (3, 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
signed_stouffer Not yet run
unverified (harvested, no recorded run)
anindex/aac/src/aac/stats.py
code served (permissive licence) · get_code("70395f80aa40c7f1")
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, 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 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.

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

We introduce AAC (Architecturally Admissible Compressor), a differentiable landmarkselection module for ALT (A*, Landmarks, and Triangle inequality) shortest-path heuristics whose outputs are admissible by construction: each forward pass is a row-stochastic mixture of triangle-inequality lower bounds, so the heuristic is admissible for every parameter setting without requiring convergence, calibration, or projection (Proposition 2). At deployment the module reduces to classical ALT on a learned subset (Proposition 4), composing endto-end with neural encoders while preserving the classical toolchain. The construction is the first differentiable instance of the compress-while-preserving-admissibility tradition in classical heuristic search. Under a matched per-vertex memory protocol we establish that ALT with farthest-pointsampling landmarks (FPS-ALT) has provably near-optimal coverage on metric graphs (Theorem 5, Corollary 6), leaving at most a few percentage points of headroom for any selector. AAC operates near this ceiling: the gap is 0.9-3.9 percentage points on 9 road networks and ≤1.3 percentage points on synthetic graphs, with zero admissibility violations across 1,500+ queries and all logged runs. At matched memory AAC is also 1.2-1.5× faster than FPS-ALT at the median query on DIMACS road networks, amortizing its offline cost within 170-1,924 queries. A controlled ablation isolates the binding constraint: training-objective drift under default initialization, not architectural capacity; identity-on-first-m initialization closes the expansion-count gap entirely. We release the module, a reusable matched-memory benchmarking protocol with paired two-one-sided-test (TOST) equivalence and pre-registration, and a reference compressed-differential-heuristics baseline.

For agents

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

get_harvested_code_for_paper("2604.20744")
get_code_for_paper("2604.20744")
have("2604.20744")

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

curl https://syntology.ai/api/ran/2604.20744.json

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

[![Syntology run record](https://syntology.ai/api/ran/2604.20744.svg)](https://syntology.ai/paper/2604.20744)

Connect an agent — have() is free.