SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2405.19793 · 2024

PDDLEGO: Iterative Planning in Textual Environments

arXiv · PDF · Open in the Atlas

Code that ran

11 of 12 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; 2 honoured the contract we drafted, 2 violated it and 0 have no contract verdict; 7 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/2405.19793.json · badge: /api/ran/2405.19793.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
zharry29/nl-to-pddl canonical (classifier label)
classifier basis: census
11 ran · 1 no recorded run
FunctionStatusWhere it lives
action_dist Ran
ran, honoured the contract we drafted
zharry29/nl-to-pddl/source/pddl-parser/pddl_parser/action.py
code served (permissive licence) · get_code("5b6ff36d5b5bdb27")
How it ran

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

class Action:
    def __init__(self, parameters, positive_preconditions, negative_preconditions, add_effects, del_effects):
        self.parameters = parameters
        self.positive_preconditions = positive_preconditions
        self.negative_preconditions = negative_preconditions
        self.add_effects = add_effects
        self.del_effects = del_effects

# Create two Action instances with matching structure
Action1 = Action(
    parameters=[("param_a", "type_1"), ("param_b", "type_2")],
    positive_preconditions=[("predicate_1", "param_a"), ("predicate_2", "param_b")],
    negative_preconditions=[("predicate_3", "param_a")],
    add_effects=[("effect_1", "param_a"), ("effect_2", "param_b")],
    del_effects=[("effect_3", "param_a")]
)

Action2 = Action(
    parameters=[("param_x", "type_1"), ("param_y", "type_2"), ("param_z", "type_3")],
    positive_preconditions=[("predicate_1", "param_x"), ("predicate_4", "param_y")],
    negative_preconditions=[("predicate_5", "param_x")],
    add_effects=[("effect_1", "param_x")],
    del_effects=[("effect_4", "param_y")]
)

ARGS = {
    "Action1": Action1,
    "Action2": Action2
}
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 tuple, ndim 1, dtype family int — a tuple output has no array shape of its own; none is stated
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
action_equal Ran
ran, violated the contract we drafted
zharry29/nl-to-pddl/source/pddl-parser/pddl_parser/action.py
code served (permissive licence) · get_code("34f579fa6aeca2a2")
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.

class Action:
    def __init__(self, parameters, positive_preconditions, negative_preconditions, add_effects, del_effects):
        self.parameters = parameters
        self.positive_preconditions = positive_preconditions
        self.negative_preconditions = negative_preconditions
        self.add_effects = add_effects
        self.del_effects = del_effects

# Create two identical Action objects for a successful comparison
parameters1 = [("x", "location"), ("y", "object")]
parameters2 = [("x", "location"), ("y", "object")]

positive_preconditions1 = [("at", "x"), ("holding", "y")]
positive_preconditions2 = [("at", "x"), ("holding", "y")]

negative_preconditions1 = [("moving", "x")]
negative_preconditions2 = [("moving", "x")]

add_effects1 = [("carried", "y")]
add_effects2 = [("carried", "y")]

del_effects1 = [("free", "y")]
del_effects2 = [("free", "y")]

Action1 = Action(
    parameters=parameters1,
    positive_preconditions=positive_preconditions1,
    negative_preconditions=negative_preconditions1,
    add_effects=add_effects1,
    del_effects=del_effects1
)

Action2 = Action(
    parameters=parameters2,
    positive_preconditions=positive_preconditions2,
    negative_preconditions=negative_preconditions2,
    add_effects=add_effects2,
    del_effects=del_effects2
)

ARGS = {
    "Action1": Action1,
    "Action2": Action2
}
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
action_instance_comp Ran
ran, violated the contract we drafted
zharry29/nl-to-pddl/source/pddl_evaluation/src/compare_plan.py
code served (permissive licence) · get_code("a339438fceb905a1")
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 collections

def setup_action_instances():
    # action_instance format: (element1, list_or_iterable)
    # The function compares:
    # 1. First elements for equality
    # 2. Counters of second elements for equality
    
    action_instance1 = ("action_type_A", ["item1", "item2", "item1"])
    action_instance2 = ("action_type_A", ["item1", "item1", "item2"])
    
    return action_instance1, action_instance2

action_instance1, action_instance2 = setup_action_instances()

ARGS = {
    "action_instance1": action_instance1,
    "action_instance2": action_instance2,
}
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
captialize_each_word Ran
ran; our contract draft was wrong, not the code
zharry29/nl-to-pddl/source/pddl_evaluation/src/util.py
code served (permissive licence) · get_code("c7d2ee1f8f892828")
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 torch

ARGS = {
    'input': 'hello world python code'
}
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 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
compare_actions_f1 Ran
ran, honoured the contract we drafted
zharry29/nl-to-pddl/source/pddl_evaluation/src/compare_plan.py
code served (permissive licence) · get_code("0589936b54490eaa")
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 collections

def parse_action(action_str):
    action_name = action_str.split(" ('")[0]
    arguments = action_str.split(" ('")[1].split("')")[0].split("', '")
    return (action_name, arguments)

def action_instance_comp(action_instance1, action_instance2):
    return action_instance1[0] == action_instance2[0] and collections.Counter(action_instance1[1]) == collections.Counter(action_instance2[1])

def compare_actions_f1(pred_actions, gold_actions):
    pred_actions = [parse_action(action) for action in pred_actions if action != '']
    gold_actions = [parse_action(action) for action in gold_actions if action != '']
    correct_pred, correct_gold = (0, 0)
    all_pred, all_gold = (len(pred_actions), len(gold_actions))
    for pred_action in pred_actions:
        for gold_action in gold_actions:
            if action_instance_comp(pred_action, gold_action):
                correct_pred += 1
                break
    for gold_action in gold_actions:
        for pred_action in pred_actions:
            if action_instance_comp(pred_action, gold_action):
                correct_gold += 1
                break
    precision = correct_pred / all_pred
    recall = correct_gold / all_gold
    f1 = 2 * precision * recall / (precision + recall)
    return (precision, recall, f1)

# Setup test data
# Action strings follow format: "action_name ('arg1', 'arg2', ...)"
pred_actions = [
    "click ('button1', 'submit')",
    "type ('field1', 'hello')",
    "navigate ('page2')"
]

gold_actions = [
    "click ('button1', 'submit')",
    "type ('field1', 'hello')",
    "scroll ('down')"
]

ARGS = {
    "pred_actions": pred_actions,
    "gold_actions": gold_actions
}
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 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
get_action_names Ran
ran; our contract draft was wrong, not the code
zharry29/nl-to-pddl/source/pddl_evaluation/src/make_gold.py
code served (permissive licence) · get_code("011b75c04a7466f1")
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 torch

# Analyze the function: get_action_names(domain)
# The function takes a 'domain' parameter and:
# 1. Splits it by '\n'
# 2. Strips whitespace from each line
# 3. Filters lines that start with '(:action'
# 4. Joins them with '\n'
# So 'domain' must be a string containing PDDL domain definition text

domain_text = """(define (domain logistics)
  (:requirements :typing :action-costs)
  (:types truck location obj)
  (:action load
    :parameters (?o - obj ?t - truck ?l - location)
    :precondition (and (at ?o ?l) (at ?t ?l))
    :effect (and (not (at ?o ?l)) (in ?o ?t)))
  (:action unload
    :parameters (?o - obj ?t - truck ?l - location)
    :precondition (and (in ?o ?t) (at ?t ?l))
    :effect (and (at ?o ?l) (not (in ?o ?t))))
  (:action drive
    :parameters (?t - truck ?from ?to - location)
    :precondition (at ?t ?from)
    :effect (and (not (at ?t ?from)) (at ?t ?to)))
)
"""

ARGS = {
    "domain": domain_text
}
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 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
get_actions Ran
ran; our contract draft was wrong, not the code
zharry29/nl-to-pddl/source/pddl_evaluation/src/make_gold.py
code served (permissive licence) · get_code("4bf89936621523df")
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 torch
import torch.nn as nn

# The function takes a 'domain' parameter which is a string
# Looking at the function body, it:
# 1. Splits the string by '(:action'
# 2. Joins parts back with '(:action'
# 3. Splits by '\n' and processes lines
# 4. Checks if lines start with '   ' (3 spaces)
# 5. Returns a reformatted string

# Create a sample domain string with action definitions
domain = """(:domain test-domain
  (:requirements :strips :typing)
  (:types location robot)
  (:action move
    :parameters (?r - robot ?from - location ?to - location)
    :precondition (at ?r ?from)
    :effect (and (at ?r ?to) (not (at ?r ?from))))
  (:action pick
    :parameters (?r - robot ?obj - object)
    :precondition (at ?r ?obj)
    :effect (holding ?r ?obj)))"""

ARGS = {
    'domain': domain
}
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 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
get_domain_header Ran
ran; our contract draft was wrong, not the code
zharry29/nl-to-pddl/source/pddl_evaluation/src/make_gold.py
code served (permissive licence) · get_code("a23888f4b12c48f9")
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 torch

def get_domain_header(domain):
    return domain.split('(:action')[0].strip()

# The function takes a single parameter 'domain' which is a string
# It splits on '(:action' and returns the first part, stripped of whitespace

domain_string = "domain blocksworld (:action pick-up"

ARGS = {
    "domain": domain_string
}
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 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
load_json Ran
ran; our contract draft was wrong, not the code
zharry29/nl-to-pddl/source/pddl_evaluation/src/util.py
code served (permissive licence) · get_code("550dd4e385e290a6")
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 json
import tempfile
import os

# Create a temporary JSON file with test data
temp_dir = tempfile.mkdtemp()
test_json_file = os.path.join(temp_dir, "test.json")

test_data = {"key": "value", "number": 42, "list": [1, 2, 3]}
with open(test_json_file, 'w') as f:
    json.dump(test_data, f)

# Define ARGS with the filename parameter
ARGS = {
    "filename": test_json_file
}
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 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
load_txt_prompt Ran
ran; our contract draft was wrong, not the code
zharry29/nl-to-pddl/source/pddl_evaluation/src/util.py
code served (permissive licence) · get_code("d136846408593074")
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 tempfile
import os

# Create a temporary text file with sample content
temp_dir = tempfile.mkdtemp()
temp_filename = os.path.join(temp_dir, "prompt.txt")

with open(temp_filename, 'w') as f:
    f.write("This is a sample prompt.\n")
    f.write("It contains multiple lines.\n")
    f.write("And various content.")

ARGS = {
    "filename": temp_filename
}
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
parse_action Ran
ran; our contract draft was wrong, not the code
zharry29/nl-to-pddl/source/pddl_evaluation/src/compare_plan.py
code served (permissive licence) · get_code("9fc0305b3d6abb02")
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)

def parse_action(action_str):
    action_name = action_str.split(" ('")[0]
    arguments = action_str.split(" ('")[1].split("')")[0].split("', '")
    return (action_name, arguments)

# The function takes a single parameter: action_str
# It expects a string formatted as: "action_name ('arg1', 'arg2', ...)"
# The function parses this string to extract the action name and arguments list

action_str = "move ('left', 'forward')"

ARGS = {
    "action_str": action_str
}
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
action_dist Not yet run
unverified (harvested, no recorded run)
zharry29/nl-to-pddl/source/pddl-parser/pddl_parser/action.py
code served (permissive licence) · get_code("ddb36e302e1fe2dc")
How it ran

no invocation recorded.

Also present in other papers' code

Of 12 samples lifted from this paper's linked repositories, 12 carry a function identity and 0 do not. Of the 12 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 12 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

Planning in textual environments have been shown to be a long-standing challenge even for current models. A recent, promising line of work uses LLMs to generate a formal representation of the environment that can be solved by a symbolic planner. However, existing methods rely on a fully-observed environment where all entity states are initially known, so a one-off representation can be constructed, leading to a complete plan. In contrast, we tackle partially-observed environments where there is initially no sufficient information to plan for the end-goal. We propose PDDLEGO that iteratively construct a planning representation that can lead to a partial plan for a given sub-goal. By accomplishing the sub-goal, more information is acquired to augment the representation, eventually achieving the end-goal. We show that plans produced by few-shot PDDLEGO are 43% more efficient than generating plans end-to-end on the Coin Collector simulation, with strong performance (98%) on the more complex Cooking World simulation where end-to-end LLMs fail to generate coherent plans (4%).

For agents

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

get_harvested_code_for_paper("2405.19793")
get_code_for_paper("2405.19793")
have("2405.19793")

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

curl https://syntology.ai/api/ran/2405.19793.json

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

[![Syntology run record](https://syntology.ai/api/ran/2405.19793.svg)](https://syntology.ai/paper/2405.19793)

Connect an agent — have() is free.