SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2410.02162 · 2024

Planning in Strawberry Fields: Evaluating and Improving the Planning and Scheduling Capabilities of LRM o1

arXiv · PDF · Open in the Atlas

Code that ran

8 of 9 samples (functions and classes) lifted from this paper's linked repositories executed on a synthesized input in Syntology's jail (0 of them constructed an object rather than computing a result; 0 honoured the contract we drafted, 0 violated it and 0 have no contract verdict; 8 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/2410.02162.json · badge: /api/ran/2410.02162.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
karthikv792/gpt-plan-benchmark canonical (classifier label)
classifier basis: census
7 ran · 0 no recorded run
karthikv792/llms-planning canonical (classifier label)
classifier basis: census
1 ran · 1 no recorded run
FunctionStatusWhere it lives
fill_template Ran
ran; our contract draft was wrong, not the code
karthikv792/gpt-plan-benchmark/plan-bench/utils/pddl_to_text.py
code served (permissive licence) · get_code("8221b37cc068b36a")
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 numpy as np

# Examine the function to understand what parameters it needs
# fill_template(INIT, GOAL, PLAN, data, instruction=False)

# INIT: string - initial conditions
# GOAL: string - goal statement
# PLAN: string - plan text
# data: dict - must have 'domain_name' key (checked with 'in' operator)
# instruction: bool - optional flag for instruction mode

ARGS = {
    'INIT': 'block A is on block B',
    'GOAL': 'block A is on the table',
    'PLAN': '\n1. Move block A from block B to the table',
    'data': {
        'domain_name': 'blocksworld_v1',
        'other_field': 'some_value'
    },
    'instruction': 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 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
generate_from_bloom Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
karthikv792/gpt-plan-benchmark/plan-bench/utils/llm_utils.py
code served (permissive licence) · get_code("52b78255f2b9a159")
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. The code mentions cuda, and the harness aliases cuda to the CPU for every sample; whether this run needed that is 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 transformers import StoppingCriteriaList, StoppingCriteria

torch.manual_seed(0)

class MockTokenizer:
    def __init__(self):
        self.vocab_size = 50257
        self.pad_token_id = 0
        self.eos_token_id = 2
    
    def __call__(self, text, return_tensors=None):
        # Return mock tokenized input
        input_ids = torch.tensor([[1, 2, 3, 4]])
        attention_mask = torch.tensor([[1, 1, 1, 1]])
        result = {
            'input_ids': input_ids,
            'attention_mask': attention_mask
        }
        return result
    
    def decode(self, token_ids, skip_special_tokes=False):
        return "This is a generated response [PLAN END]"

class MockBloomModel(nn.Module):
    def __init__(self):
        super().__init__()
        self.embedding = nn.Embedding(50257, 8)
        self.linear = nn.Linear(8, 50257)
    
    def forward(self, input_ids, **kwargs):
        x = self.embedding(input_ids)
        return self.linear(x)
    
    def generate(self, input_ids, max_new_tokens, temperature=0, top_p=1):
        # Mock generate method that returns token sequences
        batch_size = input_ids.shape[0]
        output_length = input_ids.shape[1] + max_new_tokens
        output_sequences = torch.randint(1, 100, (batch_size, output_length))
        return output_sequences

model = MockBloomModel()
tokenizer = MockTokenizer()
query = "What is the plan?"
max_tokens = 10

ARGS = {
    'model': model,
    'tokenizer': tokenizer,
    'query': query,
    'max_tokens': max_tokens
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch, transformers — 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_data Ran
ran; our contract draft was wrong, not the code
karthikv792/llms-planning/llm_planning_analysis/stats_generation.py
code served (permissive licence) · get_code("5a1377db1f3d97a5")
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 json
import tempfile
import os

# Create temporary directory and mock result files
tmpdir = tempfile.mkdtemp()
results_dir = os.path.join(tmpdir, 'results')
os.makedirs(results_dir, exist_ok=True)

# Create mock domain/model/task directories
domain = 'test_domain'
model = 'test_model'
task = 'test_task'

domain_path = os.path.join(results_dir, domain, model)
os.makedirs(domain_path, exist_ok=True)

domain_3_path = os.path.join(results_dir, f'{domain}_3', model)
os.makedirs(domain_3_path, exist_ok=True)

# Create mock data files
mock_data = {
    'instances': [
        {
            'instance_id': 'inst_1',
            'ground_truth_plan': 'action1\naction2\naction3',
            'extracted_llm_plan': 'action1\naction2\naction3',
            'llm_correct': 1
        },
        {
            'instance_id': 'inst_2',
            'ground_truth_plan': ['step1', 'step2'],
            'extracted_llm_plan': 'step1\nstep2\nextra',
            'llm_correct': 0
        },
        {
            'instance_id': 'inst_3',
            'ground_truth_plan': 'move\ngrab',
            'extracted_llm_plan': 'move\ngrab',
            'llm_correct': 1
        }
    ]
}

# Write to both files
file1 = os.path.join(domain_path, task)
file2 = os.path.join(domain_3_path, task)

with open(file1, 'w') as f:
    json.dump(mock_data, f)

with open(file2, 'w') as f:
    json.dump(mock_data, f)

# Change to temp directory so relative paths work
original_dir = os.getcwd()
os.chdir(tmpdir)

ARGS = {
    'domain': 'test_domain',
    'model': 'test_model',
    'task': 'test_task'
}
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_sorted Ran
ran; our contract draft was wrong, not the code
karthikv792/gpt-plan-benchmark/plan-bench/utils/pddl_to_text.py
code served (permissive licence) · get_code("b088e7075b491add")
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.

from dataclasses import dataclass
from typing import List

@dataclass
class Subterm:
    name: str

@dataclass
class Atom:
    symbol: 'Symbol'
    subterms: List[Subterm]

@dataclass
class Symbol:
    name: str

# Create sample atoms with different symbols and subterms
atom1 = Atom(
    symbol=Symbol(name='oxygen'),
    subterms=[Subterm(name='electron'), Subterm(name='nucleus')]
)

atom2 = Atom(
    symbol=Symbol(name='hydrogen'),
    subterms=[Subterm(name='proton')]
)

atom3 = Atom(
    symbol=Symbol(name='carbon'),
    subterms=[Subterm(name='electron'), Subterm(name='electron'), Subterm(name='nucleus')]
)

init_atoms = [atom3, atom1, atom2]

ARGS = {
    'init_atoms': init_atoms
}
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 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
parse_problem Ran
ran; our contract draft was wrong, not the code
karthikv792/gpt-plan-benchmark/plan-bench/utils/pddl_to_text.py
code served (permissive licence) · get_code("799968ced0b94957")
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 random

class Term:
    def __init__(self, name):
        self.name = name

class Symbol:
    def __init__(self, name):
        self.name = name

class Atom:
    def __init__(self, symbol_name, subterm_names):
        self.symbol = Symbol(symbol_name)
        self.subterms = [Term(name) for name in subterm_names]

class Formula:
    def __init__(self, subformulas):
        self.subformulas = subformulas

class Init:
    def __init__(self, atoms):
        self.atoms = atoms
    
    def as_atoms(self):
        return self.atoms

class Problem:
    def __init__(self, init_atoms, goal_formulas):
        self.init = Init(init_atoms)
        self.goal = Formula(goal_formulas)

problem = Problem(
    init_atoms=[
        Atom("on", ["block_a", "block_b"]),
        Atom("clear", ["block_c"]),
        Atom("ontable", ["block_b"]),
    ],
    goal_formulas=[
        Atom("on", ["block_a", "block_c"]),
        Atom("ontable", ["block_b"]),
    ]
)

data = {
    'domain_name': 'blocksworld',
    'predicates': {
        'on': '{} is on {}',
        'clear': '{} is clear',
        'ontable': '{} is on the table',
    },
    'encoded_objects': {
        'block_a': 'Block A',
        'block_b': 'Block B',
        'block_c': 'Block C',
    }
}

ARGS = {
    'problem': problem,
    'data': data,
    'shuffle': False,
}
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
random_mapping Ran
ran; our contract draft was wrong, not the code
karthikv792/gpt-plan-benchmark/plan-bench/obfuscator.py
code served (permissive licence) · get_code("232066a31b3b39f1")
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 random

# Set seed for reproducibility
random.seed(0)

# Construct config dict with actions and predicates
config = {
    'actions': ['move', 'grab', 'drop'],
    'predicates': ['at', 'holding', 'near']
}

# Construct new_words list with enough words for mapping
new_words = ['word1', 'word2', 'word3', 'word4', 'word5', 'word6']

ARGS = {
    'config': config,
    'new_words': new_words
}
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
send_query Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
karthikv792/gpt-plan-benchmark/plan-bench/utils/llm_utils.py
code served (permissive licence) · get_code("e56a33cbda477581")
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. The code mentions cuda, and the harness aliases cuda to the CPU for every sample; whether this run needed that is 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 transformers import StoppingCriteriaList, StoppingCriteria

torch.manual_seed(0)

# Stand-in Tokenizer class
class StandInTokenizer:
    def __call__(self, text, return_tensors=None):
        # Return a dict with input_ids
        if isinstance(text, str):
            # Simple tokenization: return dummy token ids
            token_ids = torch.tensor([[101, 2054, 2003, 102]], dtype=torch.long)
        else:
            token_ids = torch.tensor([[101, 102]], dtype=torch.long)
        
        result = {'input_ids': token_ids}
        if return_tensors == 'pt':
            return result
        return result
    
    def decode(self, token_ids, skip_special_tokes=False):
        return "This is a sample output from the model"

# Stand-in BLOOM model class
class StandInBloomModel(nn.Module):
    def __init__(self):
        super().__init__()
        self.linear = nn.Linear(4, 4)
    
    def generate(self, input_ids, max_new_tokens, temperature, top_p):
        # Return dummy generated sequence
        batch_size = input_ids.shape[0]
        generated = torch.cat([
            input_ids,
            torch.tensor([[102, 2003, 2054]], dtype=torch.long)
        ], dim=1)
        return generated

# Create the model dict for bloom engine
bloom_model = {
    'model': StandInBloomModel(),
    'tokenizer': StandInTokenizer()
}

# Create arguments for send_query
ARGS = {
    'query': 'What is the plan?',
    'engine': 'bloom',
    'max_tokens': 100,
    'model': bloom_model,
    'stop': '[STATEMENT]'
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
openai, torch, transformers — 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
store_predicates Ran
ran; our contract draft was wrong, not the code
karthikv792/gpt-plan-benchmark/plan-bench/model_parser/parser_new.py
code served (permissive licence) · get_code("57a2407f329edb96")
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. 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.

class MockSort:
    def __init__(self, name):
        self.name = name

class MockPredicate:
    def __init__(self, symbol, sorts):
        self.symbol = symbol
        self.sort = sorts

class MockLanguage:
    def __init__(self):
        self.predicates = [
            MockPredicate('on', [MockSort('block'), MockSort('surface')]),
            MockPredicate('clear', [MockSort('block')]),
            MockPredicate('=', [MockSort('obj'), MockSort('obj')]),
            MockPredicate('holding', [MockSort('gripper'), MockSort('block')]),
            MockPredicate('<', [MockSort('num'), MockSort('num')]),
        ]

class MockProblem:
    def __init__(self):
        self.language = MockLanguage()

class MockReader:
    def __init__(self):
        self.problem = MockProblem()

ARGS = {
    'reader': MockReader()
}
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 list — a list 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
random_mapping Not yet run
unverified (harvested, no recorded run)
karthikv792/llms-planning/llm_planning_analysis/obfuscator.py
code served (permissive licence) · get_code("51fd8b7713cc3071")
How it ran

no invocation recorded.

Also present in other papers' code

Of 9 samples lifted from this paper's linked repositories, 9 carry a function identity and 0 do not. Of the 9 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 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.

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 ability to plan a course of action that achieves a desired state of affairs has long been considered a core competence of intelligent agents and has been an integral part of AI research since its inception. With the advent of large language models (LLMs), there has been considerable interest in the question of whether or not they possess such planning abilities, but -- despite the slew of new private and open source LLMs since GPT3 -- progress has remained slow. OpenAI claims that their recent o1 (Strawberry) model has been specifically constructed and trained to escape the normal limitations of autoregressive LLMs -- making it a new kind of model: a Large Reasoning Model (LRM). In this paper, we evaluate the planning capabilities of two LRMs (o1-preview and o1-mini) on both planning and scheduling benchmarks. We see that while o1 does seem to offer significant improvements over autoregressive LLMs, this comes at a steep inference cost, while still failing to provide any guarantees over what it generates. We also show that combining o1 models with external verifiers -- in a so-called LRM-Modulo system -- guarantees the correctness of the combined system's output while further improving performance.

For agents

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

get_harvested_code_for_paper("2410.02162")
get_code_for_paper("2410.02162")
have("2410.02162")

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

curl https://syntology.ai/api/ran/2410.02162.json

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

[![Syntology run record](https://syntology.ai/api/ran/2410.02162.svg)](https://syntology.ai/paper/2410.02162)

Connect an agent — have() is free.