SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2412.12151 · EMNLP · 2024

SMARTCAL: An Approach to Self-Aware Tool-Use Evaluation and Calibration

Lei Chen, Xiaodan Zhu, Yuanhao Shen, Robert Clary

arXiv · PDF · Open in the Atlas

Code that ran

11 of 14 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; 4 honoured the contract we drafted, 3 violated it and 0 have no contract verdict; 4 failures of our instrument; 3 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/2412.12151.json · badge: /api/ran/2412.12151.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
Henrysyh2000/SMARTCAL canonical (classifier label)
classifier basis: census
11 ran · 3 no recorded run
FunctionStatusWhere it lives
calculate_average Ran
ran, honoured the contract we drafted
Henrysyh2000/SMARTCAL/ART_SM_CAL/create_sample.py
code served (permissive licence) · get_code("da6cb4abeef41a45")
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 float64, ndim 0, 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

# Construct valid arguments for calculate_average
x_value = [1.5, 2.5, 3.5, None, 4.5]

ARGS = {
    "x": x_value
}
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 float64, ndim 0, dtype family float — no output shape is stated: observe_signature.py recorded float ndim 0, which disagrees with the fixture run
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
check_answer Ran
ran, violated the contract we drafted
Henrysyh2000/SMARTCAL/ART_SM_CAL/metric.py
code served (permissive licence) · get_code("219b8d0fcf748400")
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 bool, 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 pandas as pd

class Row:
    def __init__(self, answer, prediction):
        self.answer = answer
        self.prediction = prediction

# Create test row objects with different scenarios
row_match = Row(answer="The quick brown fox", prediction="quick brown")
row_no_match = Row(answer="hello world", prediction="goodbye universe")
row_empty_prediction = Row(answer="test answer", prediction="")

# Use the first row as the main test case
test_row = row_match

ARGS = {
    "row": test_row
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
pandas — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type bool, 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
check_answer_new Ran
ran, violated the contract we drafted
Henrysyh2000/SMARTCAL/ART_SM_CAL/metric.py
code served (permissive licence) · get_code("9de0005083fe612d")
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 bool, 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 re

class Row:
    def __init__(self, answer, prediction):
        self.answer = answer
        self.prediction = prediction

# Create test row objects
row_with_match = Row(answer="Hello, World!", prediction="helloworld")
row_with_no_match = Row(answer="Python", prediction="JavaScript")
row_with_empty = Row(answer="Test", prediction="")

# Test with the matching case
ARGS = {
    "row": row_with_match
}
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 bool, 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
cleanhtml Ran
ran; our contract draft was wrong, not the code
Henrysyh2000/SMARTCAL/ART_SM_CAL/sequential_interpreter.py
code served (permissive licence) · get_code("5bcaf0a56a3a5f6a")
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 str. 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 re

CLEANR = re.compile('<.*?>')

def cleanhtml(raw_html):
    cleantext = re.sub(CLEANR, '', raw_html)
    return cleantext

ARGS = {
    'raw_html': '<p>Hello <b>world</b>!</p><div>This is a test</div>'
}
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 str — a str output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a str 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
exact_match Ran
ran, honoured the contract we drafted
Henrysyh2000/SMARTCAL/ART_SM_CAL/art_uw.py
code served (permissive licence) · get_code("a442b0bd917515d1")
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 torch
import numpy as np

torch.manual_seed(0)

# The function takes labels and predictions, which are iterables of strings
# It calls .lower() on each label and prediction, so they must be strings
# It compares them for equality

labels = ["Hello", "World", "Test", "Python"]
predictions = ["hello", "WORLD", "test", "rust"]

ARGS = {
    "labels": labels,
    "predictions": predictions
}
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 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
finqa_equal Ran
ran, violated the contract we drafted
Henrysyh2000/SMARTCAL/ART_SM_CAL/pot_tools.py
code served (permissive licence) · get_code("efe7932c4c2edafd")
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 bool, 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 numpy as np
from typing import Union

# Test case 1: numeric comparison with include_percentage=False, is_close=False
prediction_1 = 42.5
reference_1 = 42.50001

# Test case 2: boolean prediction
prediction_2 = True
reference_2 = 'yes'

# Test case 3: string comparison
prediction_3 = "answer"
reference_3 = "answer"

# Test case 4: with include_percentage=True
prediction_4 = 50.0
reference_4 = 0.5

# Test case 5: with is_close=True
prediction_5 = 1.0001
reference_5 = 1.0

# Test case 6: None prediction
prediction_6 = None
reference_6 = 42.0

# Define ARGS with all parameter combinations
ARGS = {
    "prediction": prediction_1,
    "reference": reference_1,
    "include_percentage": False,
    "is_close": 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 bool, 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
format_answers Ran
ran; our contract draft was wrong, not the code
Henrysyh2000/SMARTCAL/ART_SM_CAL/create_sample.py
code served (permissive licence) · get_code("189de55b9f8a0629")
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 list, 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
import numpy as np

# Construct valid arguments for format_answers

# Test case 1: entity type with kg mode and answer present
answer_entity_kg = {
    'answerType': 'entity',
    'answer': [
        {'name': 'Albert Einstein', 'label': {'en': 'Albert Einstein', 'de': 'Albert Einstein'}},
        {'name': 'Marie Curie', 'label': {'en': 'Marie Curie', 'de': 'Marie Curie'}}
    ],
    'mention': 'some mention'
}

# Test case 2: entity type with text mode and answer present
answer_entity_text = {
    'answerType': 'entity',
    'answer': [
        {'name': 'Isaac Newton', 'label': {'en': 'Isaac Newton', 'de': 'Isaac Newton'}},
    ],
    'mention': 'physicist'
}

# Test case 3: entity type with text mode and answer None
answer_entity_text_none = {
    'answerType': 'entity',
    'answer': None,
    'mention': 'unknown scientist'
}

# Test case 4: entity type with kg mode and answer None
answer_entity_kg_none = {
    'answerType': 'entity',
    'answer': None,
    'mention': 'some mention'
}

# Test case 5: non-entity type (e.g., 'string') with text mode
answer_string_text = {
    'answerType': 'string',
    'answer': ['42', 'other'],
    'mention': 'irrelevant'
}

# Test case 6: non-entity type with kg mode
answer_string_kg = {
    'answerType': 'string',
    'answer': ['2020', 'other'],
    'mention': 'irrelevant'
}

ARGS = {
    'answer': answer_entity_kg,
    'mode': 'kg',
    'lang': 'en'
}
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 list, ndim 1, dtype family float — a list 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
get_precision Ran
ran, honoured the contract we drafted
Henrysyh2000/SMARTCAL/ART_SM_CAL/pot_tools.py
code served (permissive licence) · get_code("5032f10ada38d45d")
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

torch.manual_seed(0)

# The function get_precision takes a single parameter gt_ans which is a float
# It converts it to string and checks if it contains a decimal point
# Then counts the digits after the decimal point

ARGS = {
    "gt_ans": 3.14159
}
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 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
get_subset Ran
ran; our contract draft was wrong, not the code
Henrysyh2000/SMARTCAL/ART_SM_CAL/utils.py
code served (permissive licence) · get_code("28c9120c7dd08f39")
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 numpy as np
import torch

torch.manual_seed(0)

# Construct inputs: a list of data items
inputs = [
    np.array([1.0, 2.0, 3.0, 4.0]),
    np.array([5.0, 6.0, 7.0, 8.0]),
    np.array([9.0, 10.0, 11.0, 12.0]),
    np.array([13.0, 14.0, 15.0, 16.0]),
    np.array([17.0, 18.0, 19.0, 20.0]),
    np.array([21.0, 22.0, 23.0, 24.0]),
    np.array([25.0, 26.0, 27.0, 28.0]),
    np.array([29.0, 30.0, 31.0, 32.0]),
    np.array([33.0, 34.0, 35.0, 36.0]),
    np.array([37.0, 38.0, 39.0, 40.0]),
    np.array([41.0, 42.0, 43.0, 44.0]),
    np.array([45.0, 46.0, 47.0, 48.0]),
]

# Construct labels: a list of integer labels
labels = [0, 1, 0, 1, 0, 1, 0, 1, 0, 1, 0, 1]

# n parameter: number of samples to choose
n = 5

ARGS = {
    "inputs": inputs,
    "labels": labels,
    "n": n,
}
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 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
simplify_ans Ran
ran; our contract draft was wrong, not the code
Henrysyh2000/SMARTCAL/ART_SM_CAL/pot_tools.py
code served (permissive licence) · get_code("674b9427b4e0d492")
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 str. 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 numpy as np
import torch

# Test case 1: numpy scalar
numpy_scalar = np.array(3.14159)

# Test case 2: numpy array
numpy_array = np.array([2.71828, 1.41421])

# Test case 3: empty/falsy value
empty_val = None

# Test case 4: list of sympy-like objects (we'll use floats as stand-ins)
list_of_nums = [1.5, 2.5, 3.5]

# Test case 5: single element list
single_elem_list = [42.0]

# Test case 6: relational type (simulated with a class)
class RelationalMock:
    def __str__(self):
        return "relational_result"

relational_obj = RelationalMock()
relational_obj.__class__.__module__ = 'relational'
relational_obj.__class__.__name__ = 'relational_type'

# Create test cases for different branches
test_cases = [
    numpy_scalar,
    numpy_array,
    list_of_nums,
    single_elem_list,
    relational_obj,
    empty_val,
    3.14,
    [1.0],
]

# Primary ARGS - demonstrate with a numpy array and convert_to_str=True
ARGS = {
    'ans': np.array(3.14159),
    'convert_to_str': 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 str — a str output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a str 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
token_match Ran
ran, honoured the contract we drafted
Henrysyh2000/SMARTCAL/ART_SM_CAL/art_uw.py
code served (permissive licence) · get_code("ed0f66014daff74a")
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 torch
import numpy as np

torch.manual_seed(0)

# Analyze the function:
# - labels: iterable of items that support .lower() method (strings)
# - predictions: iterable of iterables where each inner item supports .lower() method
# The function checks if label.lower() is in [p.lower() for p in predict]

labels = ["Hello", "World", "Python"]
predictions = [
    ["hello", "hi", "greetings"],
    ["earth", "world", "globe"],
    ["python", "code", "script"]
]

ARGS = {
    "labels": labels,
    "predictions": predictions
}
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 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
load_mintaka Not yet run
unverified (harvested, no recorded run)
Henrysyh2000/SMARTCAL/ART_SM_CAL/create_sample.py
code served (permissive licence) · get_code("cbe8ae02fd4f5c06")
How it ran

no invocation recorded.

random_tasks Not yet run
unverified (harvested, no recorded run)
Henrysyh2000/SMARTCAL/ART_SM_CAL/prompt_library.py
code served (permissive licence) · get_code("ca20cb9ad005aa47")
How it ran

no invocation recorded.

search Not yet run
unverified (harvested, no recorded run)
Henrysyh2000/SMARTCAL/ART_SM_CAL/sequential_interpreter.py
code served (permissive licence) · get_code("4d8b1a7898526649")
How it ran

no invocation recorded.

Also present in other papers' code

Of 14 samples lifted from this paper's linked repositories, 14 carry a function identity and 0 do not. Of the 14 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 14 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

The tool-use ability of Large Language Models (LLMs) has a profound impact on a wide range of industrial applications. However, LLMs' self-control and calibration capability in appropriately using tools remains understudied. The problem is consequential as it raises potential risks of degraded performance and poses a threat to the trustworthiness of the models. In this paper, we conduct a study on a family of state-of-the-art LLMs on three datasets with two mainstream tool-use frameworks. Our study reveals the tool-abuse behavior of LLMs, a tendency for models to misuse tools with overconfidence. We also find that this is a common issue regardless of model capability. Accordingly, we propose a novel approach, SMART-CAL, to mitigate the observed issues, and our results show an average of 8.6 percent increase in the QA performance and a 21.6 percent decrease in Expected Calibration Error (ECE) compared to baseline models. 1 Step 3: Augmented Reasoning (AR) Step 1: Self-Evaluation (SE) Q1: [search] Use a search engine to find the answer to the question.

For agents

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

get_harvested_code_for_paper("2412.12151")
get_code_for_paper("2412.12151")
have("2412.12151")

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

curl https://syntology.ai/api/ran/2412.12151.json

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

[![Syntology run record](https://syntology.ai/api/ran/2412.12151.svg)](https://syntology.ai/paper/2412.12151)

Connect an agent — have() is free.