SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2303.06457 · IJCAI · 2023

Active Visual Exploration Based on Attention-Map Entropy

Tomasz Trzci Ński, Bartosz Zieli Ński, Grzegorz Rypeść, Adam Pardyl, Grzegorz Kurzejamski

arXiv · PDF · Open in the Atlas

Code that ran

10 of 15 samples (functions and classes) lifted from this paper's linked repositories executed on a synthesized input in Syntology's jail (8 of them constructed an object rather than computing a result; 0 honoured the contract we drafted, 0 violated it and 8 have no contract verdict; 2 failures of our instrument; 5 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/2303.06457.json · badge: /api/ran/2303.06457.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
apardyl/AME community
Papers with Code archive
10 ran · 5 no recorded run
FunctionStatusWhere it lives
Attention Ran
ran on a synthesized input · constructed an object; no result was computed
apardyl/AME/architectures/selectors.py
code served (permissive licence) · get_code("dfb4c043722a654f")
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, Attention: 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)

# Create a small input tensor
# Shape: (batch_size=2, sequence_length=3, dim=4)
x = torch.randn(2, 3, 4)

# Create a pad_mask tensor
# Shape: (batch_size=2, sequence_length=3)
# 0 = keep, 1 = ignore
pad_mask = torch.tensor([[0, 0, 1], [0, 1, 1]])

ARGS = {
    "dim": 4,
    "num_heads": 2,
    "qkv_bias": False,
    "qk_scale": None,
    "attn_drop": 0.1,
    "proj_drop": 0.1,
}
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 Attention — 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 Attention 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
BaseArchitecture Ran
ran on a synthesized input · constructed an object; no result was computed
apardyl/AME/architectures/selectors.py
code served (permissive licence) · get_code("58f4f511ca98ef5e")
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, BaseArchitecture: 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
import math
from argparse import Namespace

torch.manual_seed(0)

class DummyDataModule:
    def __init__(self):
        self.data_dir = "/tmp/data"
        self.train_batch_size = 2
        self.eval_batch_size = 4
        self.num_workers = 0
        self.num_samples = None
        self.image_size = (128, 256)
        self.train_dataset = [(torch.randn(3, 128, 256), 0) for _ in range(5)]
        self.test_dataset = [(torch.randn(3, 128, 256), 0) for _ in range(3)]
        self.val_dataset = [(torch.randn(3, 128, 256), 0) for _ in range(3)]

class DummyArgs:
    def __init__(self):
        self.lr = 1e-3
        self.min_lr = 1e-5
        self.warmup_epochs = 5
        self.weight_decay = 0.05
        self.epochs = 100
        self.data_dir = "/tmp/data"
        self.train_batch_size = 2
        self.eval_batch_size = 4
        self.num_workers = 0
        self.num_samples = None
        self.image_size = (128, 256)

args = DummyArgs()
datamodule = DummyDataModule()

ARGS = {
    'args': args,
    'datamodule': datamodule
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
pytorch_lightning, torch, torchmetrics — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type BaseArchitecture — 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 BaseArchitecture 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
BaseDataModule Ran
ran on a synthesized input · constructed an object; no result was computed
apardyl/AME/architectures/selectors.py
code served (permissive licence) · get_code("9e7fee3e5fb1d70f")
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, BaseDataModule: 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
from argparse import ArgumentParser
from torch.utils.data import Dataset

torch.manual_seed(0)

# Create a minimal dataset class
class MinimalDataset(Dataset):
    def __init__(self, size=10):
        self.size = size
    
    def __len__(self):
        return self.size
    
    def __getitem__(self, idx):
        # Return dummy image and label
        return torch.randn(3, 128, 256), torch.randint(0, 10, (1,)).item()

# Create a minimal args object with all required attributes
class Args:
    def __init__(self):
        self.data_dir = "/tmp/data"
        self.train_batch_size = 16
        self.eval_batch_size = 64
        self.num_workers = 4
        self.num_samples = None
        self.image_size = (128, 256)

# Instantiate args
args = Args()

ARGS = {
    "args": args,
    "no_aug": False
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
pytorch_lightning, torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type BaseDataModule — 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 BaseDataModule 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
BaseSegmentationDataModule Ran
ran on a synthesized input · constructed an object; no result was computed
apardyl/AME/architectures/selectors.py
code served (permissive licence) · get_code("7b2d49217fdd340e")
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, BaseSegmentationDataModule: 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
from torch.utils.data import Dataset
from argparse import Namespace
import sys

torch.manual_seed(0)

class DummyDataset(Dataset):
    def __init__(self, size=10):
        self.size = size
    
    def __len__(self):
        return self.size
    
    def __getitem__(self, idx):
        # Return a simple (input, target) tuple
        x = torch.randn(3, 4, 5)  # 3 channels, 4x5 spatial dims
        y = torch.randint(0, 2, (4, 5))  # segmentation mask
        return x, y

ARGS = {
    'args': Namespace(
        data_dir='/tmp/data',
        train_batch_size=2,
        eval_batch_size=4,
        num_workers=0,
        num_samples=None,
        image_size=(128, 256)
    ),
    'no_aug': False
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
pytorch_lightning, torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type BaseSegmentationDataModule — 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 BaseSegmentationDataModule 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
Block Ran
ran on a synthesized input · constructed an object; no result was computed
apardyl/AME/architectures/selectors.py
code served (permissive licence) · get_code("d3c8bd812a7774f9")
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, Block: 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
from timm.models.layers import DropPath, Mlp

torch.manual_seed(0)

# Define the Block components
dim = 4
num_heads = 2
mlp_ratio = 4.0
qkv_bias = False
qk_scale = None
drop = 0.1
attn_drop = 0.1
drop_path = 0.1
act_layer = nn.GELU
norm_layer = nn.LayerNorm

ARGS = {
    "dim": dim,
    "num_heads": num_heads,
    "mlp_ratio": mlp_ratio,
    "qkv_bias": qkv_bias,
    "qk_scale": qk_scale,
    "drop": drop,
    "attn_drop": attn_drop,
    "drop_path": drop_path,
    "act_layer": act_layer,
    "norm_layer": norm_layer,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
timm, torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Block — 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 Block 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
MaeScheduler Ran
ran on a synthesized input · constructed an object; no result was computed
apardyl/AME/architectures/selectors.py
code served (permissive licence) · get_code("e5a019fc9ff41b6f")
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, MaeScheduler: 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
import torch.optim as optim

torch.manual_seed(0)

# Create a minimal model to optimize
model = nn.Linear(4, 3)

# Create an optimizer
optimizer = optim.SGD(model.parameters(), lr=0.1)

# Define arguments for MaeScheduler
ARGS = {
    'optimizer': optimizer,
    'lr': 0.1,
    'warmup_epochs': 5,
    'min_lr': 0.001,
    'epochs': 100
}
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 MaeScheduler — 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 MaeScheduler 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
MaskedAutoencoderViT Ran
ran on a synthesized input · constructed an object; no result was computed
apardyl/AME/architectures/selectors.py
code served (permissive licence) · get_code("d3f43c1a65fc5f44")
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, MaskedAutoencoderViT: 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
from timm.models.vision_transformer import PatchEmbed
from timm.models.layers import DropPath, Mlp
import numpy as np

torch.manual_seed(0)

ARGS = {
    'img_size': 224,
    'patch_size': 16,
    'in_chans': 3,
    'out_chans': 3,
    'embed_dim': 128,
    'depth': 2,
    'num_heads': 4,
    'decoder_embed_dim': 64,
    'decoder_depth': 2,
    'decoder_num_heads': 2,
    'mlp_ratio': 4.0,
    'norm_layer': nn.LayerNorm,
    'norm_pix_loss': False,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, timm, torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type MaskedAutoencoderViT — 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 MaskedAutoencoderViT 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
Retinalizer Ran
ran on a synthesized input · constructed an object; no result was computed
apardyl/AME/architectures/selectors.py
code served (permissive licence) · get_code("b98f3360f10e46f2")
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, Retinalizer: 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)

# Build a minimal model that satisfies the attributes used in Retinalizer
class MinimalMAE(nn.Module):
    def __init__(self):
        super().__init__()
        # patch_embed with patch_size attribute
        self.patch_embed = nn.Module()
        self.patch_embed.patch_size = (16,)  # tuple with single element
        # grid_size attribute
        self.grid_size = (8, 8)  # (height, width)

class MinimalModel(nn.Module):
    def __init__(self):
        super().__init__()
        self.num_glimpses = 3
        self.mae = MinimalMAE()

# Build args object
class ArgsConfig:
    def __init__(self):
        self.glimpse_size = 3  # in patches; 3 * 16 = 48 pixels

model = MinimalModel()
args = ArgsConfig()

# Create test data for __call__
# images: batch_size x channels x height x width
# For a grid of 8x8 patches with patch_size=16, we need 8*16=128 pixels per dimension
batch_size = 2
channels = 3
height = 128  # 8 patches * 16 px/patch
width = 128   # 8 patches * 16 px/patch
images = torch.randn(batch_size, channels, height, width)

# glimpses: list of tensors, each with shape (batch_size,)
# containing patch indices in the grid
glimpses = [
    torch.tensor([10, 20], dtype=torch.long),  # batch indices for first glimpse
    torch.tensor([25, 30], dtype=torch.long),  # batch indices for second glimpse
    torch.tensor([40, 50], dtype=torch.long),  # batch indices for third glimpse
]

ARGS = {
    "model": model,
    "args": args,
}
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 Retinalizer — 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 Retinalizer 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
dict_to_cpu Ran
ran; our contract draft was wrong, not the code
apardyl/AME/architectures/selectors.py
code served (permissive licence) · get_code("afe739802249075b")
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.

import torch

torch.manual_seed(0)

# Construct a dictionary of tensors for the dict_to_cpu function
# The function expects Dict[str, torch.Tensor]
tensor_dict = {
    "tensor_1": torch.randn(2, 4),
    "tensor_2": torch.randn(2, 3),
    "tensor_3": torch.randn(4, 5),
}

ARGS = {
    "tensor_dict": tensor_dict,
}
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 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_2d_sincos_pos_embed Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
apardyl/AME/architectures/selectors.py
code served (permissive licence) · get_code("f10b0368daa03927")
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 ndarray, ndim 2, dtype family float, shape (12, 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 numpy as np

ARGS = {
    'embed_dim': 8,
    'grid_size': (4, 3),
    'cls_token': False
}
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 ndarray, ndim 2, dtype family float, shape (12, 8) — shape (12, 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
AttentionGlimpseSelector Not yet run
unverified (harvested, no recorded run)
apardyl/AME/architectures/selectors.py
code served (permissive licence) · get_code("87ceaab158726213")
How it ran

no invocation recorded.

BaseGlimpseMae Not yet run
unverified (harvested, no recorded run)
apardyl/AME/architectures/selectors.py
code served (permissive licence) · get_code("4acaf185025d6cc9")
How it ran

no invocation recorded.

BaseGlimpseSelector Not yet run
unverified (harvested, no recorded run)
apardyl/AME/architectures/selectors.py
code served (permissive licence) · get_code("526634f146a3e767")
How it ran

no invocation recorded.

MetricMixin Not yet run
unverified (harvested, no recorded run)
apardyl/AME/architectures/selectors.py
code served (permissive licence) · get_code("9c8b6524d1d9c312")
How it ran

no invocation recorded.

mae_vit_large_patch16_dec512d8b Not yet run
unverified (harvested, no recorded run)
apardyl/AME/architectures/selectors.py
code served (permissive licence) · get_code("287ba7a4489fb320")
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, 2 are also present in other papers whose code we hold (0 with at least one paper dated earlier, 0 the same month, 2 later) and 13 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
get_2d_sincos_pos_embed
brick of this paper's assembly: helper
2 other papers
2 later · 1 only as an assembly brick
2404.03482 (2024-04)
AdaGlimpse: Active Visual Exploration with Arbitrary Glimpse Position and Scale
2405.05841 (2024-05)
Self-Supervised Pre-training with Symmetric Superimposition Modeling for Scene Text Recognition
dict_to_cpu
brick of this paper's assembly: helper
1 other paper
1 later
2404.03482 (2024-04)
AdaGlimpse: Active Visual Exploration with Arbitrary Glimpse Position and Scale

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

Active visual exploration addresses the issue of limited sensor capabilities in real-world scenarios, where successive observations are actively chosen based on the environment. To tackle this problem, we introduce a new technique called Attention-Map Entropy (AME). It leverages the internal uncertainty of the transformer-based model to determine the most informative observations. In contrast to existing solutions, it does not require additional loss components, which simplifies the training. Through experiments, which also mimic retina-like sensors, we show that such simplified training significantly improves the performance of reconstruction, segmentation and classification on publicly available datasets.

For agents

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

get_harvested_code_for_paper("2303.06457")
get_code_for_paper("2303.06457")
have("2303.06457")

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

curl https://syntology.ai/api/ran/2303.06457.json

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

[![Syntology run record](https://syntology.ai/api/ran/2303.06457.svg)](https://syntology.ai/paper/2303.06457)

Connect an agent — have() is free.