SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2403.19904 · CVPR · 2024

Fully Geometric Panoramic Localization

Young Kim, Junho Kim, Jiwon Jeong

arXiv · PDF · Open in the Atlas

Code that ran

10 of 18 samples (functions and classes) lifted from this paper's linked repositories executed on a synthesized input in Syntology's jail (0 of them constructed an object rather than computing a result; 0 honoured the contract we drafted, 0 violated it and 10 have no contract verdict; 0 failures of our instrument; 8 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/2403.19904.json · badge: /api/ran/2403.19904.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
82magnolia/panoramic-localization community (classifier label)
classifier basis not recorded on the link
10 ran · 8 no recorded run
FunctionStatusWhere it lives
cloud2idx Ran
ran on a synthesized input
82magnolia/panoramic-localization/utils.py
code served (permissive licence) · get_code("4ee13181a173c0c1")
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, 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)

# Non-batched case: (N, 3) tensor
xyz_non_batched = torch.randn(4, 3)

# Batched case: (B, N, 3) tensor
xyz_batched = torch.randn(2, 4, 3)

ARGS = {
    "xyz": xyz_non_batched,
    "batched": 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 Tensor, ndim 2, dtype family float, shape (4, 2) — shape (4, 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
color_mod Ran
ran on a synthesized input
82magnolia/panoramic-localization/color_utils.py
code served (permissive licence) · get_code("8d32a7070c6109dd")
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 numpy as np

torch.manual_seed(0)

# Create a small image tensor (H, W, 3) with RGB values in [0, 1]
# Using small dimensions: 4x5x3
img = torch.rand(4, 5, 3)

# Create a small point cloud RGB tensor (N, 3) with RGB values in [0, 1]
# Using N=6 points
rgb = torch.rand(6, 3)

# Number of bins for histogram
num_bins = 32

ARGS = {
    "img": img,
    "rgb": rgb,
    "num_bins": num_bins,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
cv2, numpy, 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
get_init_dict_cpo Ran
ran on a synthesized input
82magnolia/panoramic-localization/dict_utils.py
code served (permissive licence) · get_code("4cbc70d6112dc53f")
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 dict. 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 typing import NamedTuple
import numpy as np

class Config(NamedTuple):
    xy_only: bool = False
    num_trans: int = 50
    yaw_only: bool = False
    num_yaw: int = 8
    num_pitch: int = 8
    num_roll: int = 8
    max_yaw: float = 2 * np.pi
    min_yaw: float = 0
    max_pitch: float = 2 * np.pi
    min_pitch: float = 0
    max_roll: float = 2 * np.pi
    min_roll: float = 0
    x_max: float = 100.0
    x_min: float = -100.0
    y_max: float = 100.0
    y_min: float = -100.0
    z_max: float = 50.0
    z_min: float = 0.0
    z_prior: float = 25.0
    dataset: str = "test_dataset"
    sample_rate_for_init: float = 0.5
    trans_init_mode: str = "quantile"
    num_split_h: int = 2
    num_split_w: int = 4
    hist_stat: str = "mean"
    sin_hist: bool = False
    benchmark_grid: bool = False

cfg = Config()

ARGS = {
    "cfg": cfg
}
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 dict — a dict output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a dict output the recorded ndim 0 and dtype family float describe that coercion, not the output, and are not stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
get_init_dict_piccolo Ran
ran on a synthesized input
82magnolia/panoramic-localization/dict_utils.py
code served (permissive licence) · get_code("7babdd5582135a76")
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 dict. 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 typing import NamedTuple
import numpy as np

class Config(NamedTuple):
    xy_only: bool = True
    num_trans: int = 50
    yaw_only: bool = True
    num_yaw: int = 4
    num_pitch: int = 0
    num_roll: int = 0
    max_yaw: float = 2 * np.pi
    min_yaw: float = 0
    max_pitch: float = 2 * np.pi
    min_pitch: float = 0
    max_roll: float = 2 * np.pi
    min_roll: float = 0
    x_max: float = None
    x_min: float = None
    y_max: float = None
    y_min: float = None
    z_max: float = None
    z_min: float = None
    z_prior: float = None
    dataset: str = "test_dataset"
    sample_rate_for_init: float = None
    trans_init_mode: str = "quantile"
    num_split_h: int = 2
    num_split_w: int = 4

cfg = Config(
    xy_only=True,
    num_trans=50,
    yaw_only=True,
    num_yaw=4,
    num_pitch=0,
    num_roll=0,
    max_yaw=2 * np.pi,
    min_yaw=0,
    max_pitch=2 * np.pi,
    min_pitch=0,
    max_roll=2 * np.pi,
    min_roll=0,
    x_max=100.0,
    x_min=-100.0,
    y_max=100.0,
    y_min=-100.0,
    z_max=50.0,
    z_min=0.0,
    z_prior=25.0,
    dataset="synthetic_dataset",
    sample_rate_for_init=0.5,
    trans_init_mode="quantile",
    num_split_h=2,
    num_split_w=4
)

ARGS = {
    "cfg": cfg
}
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 dict — a dict output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a dict 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
histogram Ran
ran on a synthesized input
82magnolia/panoramic-localization/color_utils.py
code served (permissive licence) · get_code("e4429b985acc2e33")
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 (32, 32, 32) as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch

torch.manual_seed(0)

# Case 1: 3D tensor (H, W, 3) with corresponding mask (H, W)
img_3d = torch.randint(0, 256, (4, 5, 3), dtype=torch.uint8).float()
mask_3d = torch.randint(0, 2, (4, 5), dtype=torch.uint8)

# Case 2: 4D tensor (B, H, W, 3) with corresponding mask (B, H, W)
img_4d = torch.randint(0, 256, (2, 4, 5, 3), dtype=torch.uint8).float()
mask_4d = torch.randint(0, 2, (2, 4, 5), dtype=torch.uint8)

channels_list = [32, 32, 32]

ARGS = {
    "img": img_3d,
    "mask": mask_3d,
    "channels": channels_list,
    "normalize": True
}
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 (32, 32, 32) — shape (32, 32, 32) as a later run recorded it
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
histogram_sphere Ran
ran on a synthesized input
82magnolia/panoramic-localization/color_utils.py
code served (permissive licence) · get_code("bbf76fc88a3f8910")
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 (8, 8, 8) 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 RGB color values (N, 3) - values in range [0, 255]
colors = torch.tensor([
    [128, 64, 32],
    [200, 150, 100],
    [50, 75, 200],
    [255, 0, 128]
], dtype=torch.float32)

# Create weights (N, 1) or (N,)
weights = torch.tensor([1.0, 2.0, 1.5, 0.5], dtype=torch.float32)

# Channel bins per dimension
channels = [8, 8, 8]

# Normalization flag
normalize = True

ARGS = {
    "colors": colors,
    "weights": weights,
    "channels": channels,
    "normalize": normalize
}
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 (8, 8, 8) — shape (8, 8, 8) 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
inv_cloud2idx Ran
ran on a synthesized input
82magnolia/panoramic-localization/utils.py
code served (permissive licence) · get_code("941e51ef730b7afb")
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
import numpy as np

torch.manual_seed(0)

# Create a small tensor with shape (2, 2) representing 2 points in 2D coordinate space
# The function expects coordinates in normalized range [-1, 1]
coord_arr = torch.randn(2, 2)

ARGS = {
    "coord_arr": coord_arr
}
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, 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
sample_from_img Ran
ran on a synthesized input
82magnolia/panoramic-localization/utils.py
code served (permissive licence) · get_code("b12260e406df53a9")
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
import torch.nn.functional as F

torch.manual_seed(0)

# Create a small image tensor (H, W, 3)
img = torch.rand(4, 5, 3)

# Create coordinate arrays for non-batched case
coord_arr_non_batched = torch.rand(2, 2) * 2 - 1  # (N, 2) in range [-1, 1]

# Create coordinate arrays for batched case
coord_arr_batched = torch.rand(3, 2, 2) * 2 - 1  # (batch_size, N, 2) in range [-1, 1]

ARGS = {
    'img': img,
    'coord_arr': coord_arr_non_batched,
    'padding': 'zeros',
    'mode': 'bilinear',
    'batched': False,
}
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
split_3d Ran
ran on a synthesized input
82magnolia/panoramic-localization/edge_utils.py
code served (permissive licence) · get_code("25ecc063202588cf")
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 sample 3D edge directions: (N_3D, 3) tensor
# Using N_3D=2 for tiny batch, 3 for 3D coordinates
dirs = torch.randn(2, 3)

# Create principal directions: (3, 3) tensor
# Three orthonormal principal directions in 3D space
principal_3d = torch.tensor([
    [1.0, 0.0, 0.0],
    [0.0, 1.0, 0.0],
    [0.0, 0.0, 1.0]
], dtype=torch.float32)

# Inlier threshold for dot product
inlier_thres = 0.05

ARGS = {
    "dirs": dirs,
    "principal_3d": principal_3d,
    "inlier_thres": inlier_thres
}
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
split_by_axes Ran
ran on a synthesized input
82magnolia/panoramic-localization/edge_utils.py
code served (permissive licence) · get_code("841a7f72356b6c69")
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

torch.manual_seed(0)

# edge_2d: (N_2D, 9) tensor containing [normals start end]
# The function uses edge_2d[:, :3] which extracts the first 3 columns (normals)
N_2D = 2
edge_2d = torch.randn(N_2D, 9)

# dirs: (N_3D, 3) tensor containing 3D edge directions
N_3D = 4
dirs = torch.randn(N_3D, 3)

# principal_3d: (3, 3) tensor containing three principal directions in 3D
principal_3d = torch.randn(3, 3)

# rot_mtx: (3, 3) or (N_r, 3, 3) torch tensor containing rotation matrix
# Test with 2D case first
rot_mtx = torch.randn(3, 3)

# inlier_thres: Dot product threshold for extracting inliers along principal directions
inlier_thres = 0.05

ARGS = {
    "edge_2d": edge_2d,
    "dirs": dirs,
    "principal_3d": principal_3d,
    "rot_mtx": rot_mtx,
    "inlier_thres": inlier_thres,
}
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
get_init_dict_ldl Not yet run
unverified (harvested, no recorded run)
82magnolia/panoramic-localization/dict_utils.py
code served (permissive licence) · get_code("70ebeddff6217dd6")
How it ran

no invocation recorded.

load_logger Not yet run
unverified (harvested, no recorded run)
82magnolia/panoramic-localization/log_utils.py
code served (permissive licence) · get_code("b2711bf8551c89dc")
How it ran

no invocation recorded.

obtain_gt_stanford Not yet run
unverified (harvested, no recorded run)
82magnolia/panoramic-localization/data_utils.py
code served (permissive licence) · get_code("09931227dd57457f")
How it ran

no invocation recorded.

parse_ini Not yet run
unverified (harvested, no recorded run)
82magnolia/panoramic-localization/parse_utils.py
code served (permissive licence) · get_code("83f2b2da5fec9344")
How it ran

no invocation recorded.

parse_value Not yet run
unverified (harvested, no recorded run)
82magnolia/panoramic-localization/parse_utils.py
code served (permissive licence) · get_code("a6cb944a5fc971a0")
How it ran

no invocation recorded.

read_line Not yet run
unverified (harvested, no recorded run)
82magnolia/panoramic-localization/data_utils.py
code served (permissive licence) · get_code("9291ca2c1dfc92f1")
How it ran

no invocation recorded.

read_txt_pcd Not yet run
unverified (harvested, no recorded run)
82magnolia/panoramic-localization/data_utils.py
code served (permissive licence) · get_code("0f2e11318287210f")
How it ran

no invocation recorded.

split_2d Not yet run
unverified (harvested, no recorded run)
82magnolia/panoramic-localization/edge_utils.py
code served (permissive licence) · get_code("edde9e4433897830")
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, 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 18 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 a lightweight and accurate localization method that only utilizes the geometry of 2D-3D lines. Given a pre-captured 3D map, our approach localizes a panorama image, taking advantage of the holistic 360 • view. The system mitigates potential privacy breaches or domain discrepancies by avoiding trained or hand-crafted visual descriptors. However, as lines alone can be ambiguous, we express distinctive yet compact spatial contexts from relationships between lines, namely the dominant directions of parallel lines and the intersection between nonparallel lines. The resulting representations are efficient in processing time and memory compared to conventional visual descriptor-based methods. Given the groups of dominant line directions and their intersections, we accelerate the search process to test thousands of pose candidates in less than a millisecond without sacrificing accuracy. We empirically show that the proposed 2D-3D matching can localize panoramas for challenging scenes with similar structures, dramatic domain shifts or illumination changes. Our fully geometric approach does not involve extensive parameter tuning or neural network training, making it a practical algorithm that can be readily deployed in the real world. Project page including the code is available through this link: https://82magnolia.github.io/fgpl/.

For agents

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

get_harvested_code_for_paper("2403.19904")
get_code_for_paper("2403.19904")
have("2403.19904")

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

curl https://syntology.ai/api/ran/2403.19904.json

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

[![Syntology run record](https://syntology.ai/api/ran/2403.19904.svg)](https://syntology.ai/paper/2403.19904)

Connect an agent — have() is free.