SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2403.14715 · ICLR · 2025

Towards Understanding Why Label Smoothing Degrades Selective Classification and How to Fix It

Christos-Savvas Bouganis, Guoxuan Xia, Olivier Laurent, Gianni Franchi

arXiv · PDF · Open in the Atlas

Code that ran

9 of 10 samples (functions and classes) lifted from this paper's linked repositories executed on a synthesized input in Syntology's jail (2 of them constructed an object rather than computing a result; 1 honoured the contract we drafted, 0 violated it and 2 have no contract verdict; 6 failures of our instrument; 1 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/2403.14715.json · badge: /api/ran/2403.14715.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
tensorflow/tpu canonical (classifier label)
classifier basis: census
6 ran · 1 no recorded run
facebookresearch/deit unrelated (classifier label)
classifier basis: classifier_2025
2 ran · 0 no recorded run
ENSTA-U2IS-AI/Label-smoothing-Selective-classification-Code marked official in Papers with Code
Papers with Code archive
1 ran · 0 no recorded run
FunctionStatusWhere it lives
Block Ran
ran on a synthesized input · constructed an object; no result was computed
facebookresearch/deit/models_v2.py
code served (permissive licence) · get_code("b02d29318271a1fa")
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.vision_transformer import Mlp
from timm.models.layers import DropPath

torch.manual_seed(0)

class Attention(nn.Module):
    def __init__(self, dim, num_heads=8, qkv_bias=False, qk_scale=None, attn_drop=0.0, proj_drop=0.0):
        super().__init__()
        self.num_heads = num_heads
        head_dim = dim // num_heads
        self.scale = qk_scale or head_dim ** (-0.5)
        self.qkv = nn.Linear(dim, dim * 3, bias=qkv_bias)
        self.attn_drop = nn.Dropout(attn_drop)
        self.proj = nn.Linear(dim, dim)
        self.proj_drop = nn.Dropout(proj_drop)

    def forward(self, x):
        B, N, C = x.shape
        qkv = self.qkv(x).reshape(B, N, 3, self.num_heads, C // self.num_heads).permute(2, 0, 3, 1, 4)
        q, k, v = (qkv[0], qkv[1], qkv[2])
        q = q * self.scale
        attn = q @ k.transpose(-2, -1)
        attn = attn.softmax(dim=-1)
        attn = self.attn_drop(attn)
        x = (attn @ v).transpose(1, 2).reshape(B, N, C)
        x = self.proj(x)
        x = self.proj_drop(x)
        return x

ARGS = {
    'dim': 4,
    'num_heads': 2,
    'mlp_ratio': 4.0,
    'qkv_bias': False,
    'qk_scale': None,
    'drop': 0.0,
    'attn_drop': 0.0,
    'drop_path': 0.0,
    'act_layer': nn.GELU,
    'norm_layer': nn.LayerNorm,
    'Attention_block': Attention,
    'Mlp_block': Mlp,
    'init_values': 0.0001,
}
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
cross_replica_average Ran
ran; our contract draft was wrong, not the code
tensorflow/tpu/models/official/mask_rcnn/tpu_normalization.py
code served (permissive licence) · get_code("1ab5d66bd8c493da")
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 SymbolicTensor. 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 tensorflow.compat.v1 as tf
import numpy as np

tf.disable_v2_behavior()

# Create a minimal mock TPU context
class MockTPUContext:
    def __init__(self, number_of_shards):
        self.number_of_shards = number_of_shards

class MockTPUFunction:
    _tpu_context = MockTPUContext(number_of_shards=4)
    
    @staticmethod
    def get_tpu_context():
        return MockTPUFunction._tpu_context

# Monkey-patch the tpu_function module
from tensorflow.python.tpu import tpu_function
tpu_function.get_tpu_context = MockTPUFunction.get_tpu_context

# Create a small tensor for testing
tensor_value = tf.constant([[1.0, 2.0, 3.0, 4.0],
                            [5.0, 6.0, 7.0, 8.0]], dtype=tf.float32)

ARGS = {
    't': tensor_value,
    'num_groups': 2
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, tensorflow — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type SymbolicTensor — a SymbolicTensor 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
dataset Ran
ran; our contract draft was wrong, not the code
tensorflow/tpu/models/official/mnist/mnist_tpu.py
code served (permissive licence) · get_code("2e4c38986c252157")
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 DatasetV1Adapter. 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 tensorflow.compat.v1 as tf
import tempfile
import os

# Disable TensorFlow v2 behavior
tf.disable_v2_behavior()

# Create a minimal valid TFRecords file with MNIST-like data
temp_dir = tempfile.mkdtemp()
records_file = os.path.join(temp_dir, "test.tfrecords")

def create_tfrecord_file(filepath):
    """Create a minimal TFRecords file with 2 MNIST-like examples."""
    writer = tf.io.TFRecordWriter(filepath)
    
    # Create 2 sample records
    for i in range(2):
        # Create a 28x28x1 image (784 bytes for uint8)
        image_data = bytes([i % 256] * 784)
        label = i % 10
        
        feature = {
            'image_raw': tf.train.Feature(bytes_list=tf.train.BytesList(value=[image_data])),
            'label': tf.train.Feature(int64_list=tf.train.Int64List(value=[label]))
        }
        example = tf.train.Example(features=tf.train.Features(feature=feature))
        writer.write(example.SerializeToString())
    
    writer.close()

create_tfrecord_file(records_file)

ARGS = {
    'records_file': records_file
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
tensorflow — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type DatasetV1Adapter — a DatasetV1Adapter output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a DatasetV1Adapter 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
efficientnet_x_params Ran
ran, honoured the contract we drafted
tensorflow/tpu/models/official/efficientnet/tpu/efficientnet_x_builder.py
code served (permissive licence) · get_code("45dc547f9acc69d7")
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 torch

torch.manual_seed(0)

ARGS = {
    "model_name": "efficientnet-x-b0"
}
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 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
eval_input_fn Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
tensorflow/tpu/models/samples/core/get_started/iris_data_tpu.py
code served (permissive licence) · get_code("78b2922704a5ab6c")
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 DatasetV1Adapter. 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 tensorflow.compat.v1 as tf
import numpy as np

tf.disable_v2_behavior()

# Create sample features as a dict
features = {
    'feature_1': np.array([[1.0, 2.0, 3.0, 4.0], [5.0, 6.0, 7.0, 8.0]]),
    'feature_2': np.array([[0.1, 0.2, 0.3, 0.4], [0.5, 0.6, 0.7, 0.8]])
}

# Create sample labels
labels = np.array([0, 1])

# Batch size
batch_size = 2

ARGS = {
    'features': features,
    'labels': labels,
    'batch_size': batch_size
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, tensorflow — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type DatasetV1Adapter — a DatasetV1Adapter output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a DatasetV1Adapter 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
metric_fn Ran
ran; our contract draft was wrong, not the code
tensorflow/tpu/models/official/mnist/mnist_tpu.py
code served (permissive licence) · get_code("868a20fd4649650a")
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 tensorflow.compat.v1 as tf
import numpy as np

tf.disable_v2_behavior()

# Create small sample tensors
batch_size = 2
num_classes = 4

# labels: shape (batch_size,) with class indices
labels = tf.constant([0, 2], dtype=tf.int32)

# logits: shape (batch_size, num_classes) with raw model outputs
logits = tf.constant([
    [2.1, 0.5, -1.2, 0.3],
    [-0.5, 1.1, 2.8, -0.2]
], dtype=tf.float32)

ARGS = {
    'labels': labels,
    'logits': logits
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, tensorflow — 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
norm_logits Ran
ran; our contract draft was wrong, not the code
ENSTA-U2IS-AI/Label-smoothing-Selective-classification-Code/torch_uncertainty_ls/utils.py
code served (permissive licence) · get_code("2fd3af75ac5f143b")
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

torch.manual_seed(0)

# Create a small logits tensor (batch_size=2, num_classes=4)
logits = torch.randn(4, 2, dtype=torch.float32)

# Define arguments for norm_logits
ARGS = {
    'logits': logits,
    'p': 2,
    'add_unit': 0.0,
}
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 (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
train_input_fn Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
tensorflow/tpu/models/samples/core/get_started/iris_data_tpu.py
code served (permissive licence) · get_code("9e48fa6fe2385a5c")
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 DatasetV1Adapter. 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 tensorflow.compat.v1 as tf
import numpy as np

tf.disable_v2_behavior()

# Create small sample data
num_samples = 8
num_features = 4
num_classes = 3

# Create features as a dict (as expected by the function)
features = {
    'feature_1': np.random.randn(num_samples, num_features).astype(np.float32),
    'feature_2': np.random.randn(num_samples, 2).astype(np.float32),
}

# Create labels
labels = np.random.randint(0, num_classes, size=(num_samples,)).astype(np.int32)

# Define batch size
batch_size = 2

ARGS = {
    'features': features,
    'labels': labels,
    'batch_size': batch_size,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, tensorflow — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type DatasetV1Adapter — a DatasetV1Adapter output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a DatasetV1Adapter 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
vit_models Ran
ran on a synthesized input · constructed an object; no result was computed
facebookresearch/deit/models_v2.py
code served (permissive licence) · get_code("067d430bb76eafdd")
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, vit_models: 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 Mlp, PatchEmbed
from timm.models.layers import DropPath, trunc_normal_

torch.manual_seed(0)

class Attention(nn.Module):
    def __init__(self, dim, num_heads=8, qkv_bias=False, qk_scale=None, attn_drop=0.0, proj_drop=0.0):
        super().__init__()
        self.num_heads = num_heads
        head_dim = dim // num_heads
        self.scale = qk_scale or head_dim ** (-0.5)
        self.qkv = nn.Linear(dim, dim * 3, bias=qkv_bias)
        self.attn_drop = nn.Dropout(attn_drop)
        self.proj = nn.Linear(dim, dim)
        self.proj_drop = nn.Dropout(proj_drop)

    def forward(self, x):
        B, N, C = x.shape
        qkv = self.qkv(x).reshape(B, N, 3, self.num_heads, C // self.num_heads).permute(2, 0, 3, 1, 4)
        q, k, v = (qkv[0], qkv[1], qkv[2])
        q = q * self.scale
        attn = q @ k.transpose(-2, -1)
        attn = attn.softmax(dim=-1)
        attn = self.attn_drop(attn)
        x = (attn @ v).transpose(1, 2).reshape(B, N, C)
        x = self.proj(x)
        x = self.proj_drop(x)
        return x

class Block(nn.Module):
    def __init__(self, dim, num_heads, mlp_ratio=4.0, qkv_bias=False, qk_scale=None, drop=0.0, attn_drop=0.0, drop_path=0.0, act_layer=nn.GELU, norm_layer=nn.LayerNorm, Attention_block=Attention, Mlp_block=Mlp, init_values=0.0001):
        super().__init__()
        self.norm1 = norm_layer(dim)
        self.attn = Attention_block(dim, num_heads=num_heads, qkv_bias=qkv_bias, qk_scale=qk_scale, attn_drop=attn_drop, proj_drop=drop)
        self.drop_path = DropPath(drop_path) if drop_path > 0.0 else nn.Identity()
        self.norm2 = norm_layer(dim)
        mlp_hidden_dim = int(dim * mlp_ratio)
        self.mlp = Mlp_block(in_features=dim, hidden_features=mlp_hidden_dim, act_layer=act_layer, drop=drop)

    def forward(self, x):
        x = x + self.drop_path(self.attn(self.norm1(x)))
        x = x + self.drop_path(self.mlp(self.norm2(x)))
        return x

ARGS = {
    'img_size': 224,
    'patch_size': 16,
    'in_chans': 3,
    'num_classes': 1000,
    'embed_dim': 4,
    'depth': 2,
    'num_heads': 2,
    'mlp_ratio': 4.0,
    'qkv_bias': False,
    'qk_scale': None,
    'drop_rate': 0.0,
    'attn_drop_rate': 0.0,
    'drop_path_rate': 0.0,
    'norm_layer': nn.LayerNorm,
    'global_pool': None,
    'block_layers': Block,
    'Patch_layer': PatchEmbed,
    'act_layer': nn.GELU,
    'Attention_block': Attention,
    'Mlp_block': Mlp,
    'dpr_constant': True,
    'init_scale': 0.0001,
    'mlp_ratio_clstk': 4.0,
}
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 vit_models — 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 vit_models 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
load_data Not yet run
unverified (harvested, no recorded run)
tensorflow/tpu/models/samples/core/get_started/iris_data_tpu.py
code served (permissive licence) · get_code("72bd61b1c3a52a02")
How it ran

no invocation recorded.

Also present in other papers' code

Of 10 samples lifted from this paper's linked repositories, 10 carry a function identity and 0 do not. Of the 10 distinct functions with an identity, 1 are also present in other papers whose code we hold (1 with at least one paper dated earlier, 0 the same month, 0 later) and 9 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
Block
brick of this paper's assembly: helper
1 other paper
1 dated earlier
2304.04952 (2023-04)
Data-Efficient Image Quality Assessment with Attention-Panel Decoder

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

Label smoothing (LS) is a popular regularisation method for training neural networks as it is effective in improving test accuracy and is simple to implement. "Hard" one-hot labels are "smoothed" by uniformly distributing probability mass to other classes, reducing overfitting. Prior work has suggested that in some cases LS can degrade selective classification (SC) -where the aim is to reject misclassifications using a model's uncertainty. In this work, we first demonstrate empirically across an extended range of large-scale tasks and architectures that LS consistently degrades SC. We then address a gap in existing knowledge, providing an explanation for this behaviour by analysing logit-level gradients: LS degrades the uncertainty rank ordering of correct vs incorrect predictions by suppressing the max logit more when a prediction is likely to be correct, and less when it is likely to be wrong. This elucidates previously reported experimental results where strong classifiers underperform in SC. We then demonstrate the empirical effectiveness of post-hoc logit normalisation for recovering lost SC performance caused by LS. Furthermore, linking back to our gradient analysis, we again provide an explanation for why such normalisation is effective. Project page: https://ensta-u2is-ai.github. io/Understanding-Label-smoothing-Selective-classification/ Recently Zhu et al. (2022; 2024) empirically observe that LS can lead to worse SC for convolutional neural networks (CNNs) performing image classification, yet it is not clear why this occurs. Concurrently, large-scale empirical evaluations of open-source pre-trained models have shown

For agents

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

get_harvested_code_for_paper("2403.14715")
get_code_for_paper("2403.14715")
have("2403.14715")

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

curl https://syntology.ai/api/ran/2403.14715.json

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

[![Syntology run record](https://syntology.ai/api/ran/2403.14715.svg)](https://syntology.ai/paper/2403.14715)

Connect an agent — have() is free.