SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2102.11137 · 2021

Program Synthesis Guided Reinforcement Learning for Partially Observed Environments

arXiv · PDF · Open in the Atlas

Code that ran

4 of 4 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; 3 honoured the contract we drafted, 0 violated it and 0 have no contract verdict; 1 failure of our instrument; 0 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/2102.11137.json · badge: /api/ran/2102.11137.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
yycdavid/program-synthesis-guided-rl canonical (classifier label)
classifier basis: classifier_2025
4 ran · 0 no recorded run
FunctionStatusWhere it lives
avg_stats_against_iter Ran
ran, honoured the contract we drafted
yycdavid/program-synthesis-guided-rl/analysis/plots.py
code served (permissive licence) · get_code("8fd516571a673e35")
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) and returned tuple, ndim 2, 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 a temporary directory and stats file
temp_dir = tempfile.mkdtemp()
results_dir = os.path.join(temp_dir, 'results')
os.makedirs(results_dir, exist_ok=True)

# Create sample stats data
stats_data = [
    {
        'iter': 1000,
        'result': [
            {
                'task': 'task_a',
                'reward': [0.5, 0.6, 0.7],
                'finish_time': [10, 12, 11]
            },
            {
                'task': 'task_b',
                'reward': [0.4, 0.5],
                'finish_time': [15, 14]
            }
        ]
    },
    {
        'iter': 2000,
        'result': [
            {
                'task': 'task_a',
                'reward': [0.7, 0.8, 0.75],
                'finish_time': [9, 10, 9]
            },
            {
                'task': 'task_b',
                'reward': [0.6, 0.65],
                'finish_time': [12, 11]
            }
        ]
    },
    {
        'iter': 3000,
        'result': [
            {
                'task': 'task_a',
                'reward': [0.8, 0.85, 0.82],
                'finish_time': [8, 8, 9]
            },
            {
                'task': 'task_b',
                'reward': [0.7, 0.72],
                'finish_time': [10, 10]
            }
        ]
    }
]

# Write stats to file
stats_path = os.path.join(results_dir, 'test_exp', 'stats.json')
os.makedirs(os.path.dirname(stats_path), exist_ok=True)
with open(stats_path, 'w') as f:
    json.dump(stats_data, f)

# Create args object
class Args:
    def __init__(self):
        self.max_iter = 400000
        self.exp = 'test_exp'
        self.file = 'stats.json'
        self.task = 'all'
        self.iter = -1

args = Args()

# Define ARGS for avg_stats_against_iter function
ARGS = {
    'stats': stats_data,
    'target_stats': 'reward',
    'target_task': 'all',
    'max_iter': 400000
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1
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 2, 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
final_iter_avg_stats Ran
ran, honoured the contract we drafted
yycdavid/program-synthesis-guided-rl/analysis/plots.py
code served (permissive licence) · get_code("7d7cd230b0330b12")
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) 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 os
import json
import tempfile

# Create a temporary directory and stats file for testing
temp_dir = tempfile.mkdtemp()
stats_file = os.path.join(temp_dir, 'stats.json')

# Create sample stats data matching the expected structure
stats_data = [
    {
        'iter': 1000,
        'result': [
            {
                'task': 'task1',
                'reward': [0.5, 0.6, 0.7],
                'finish_time': [10.0, 11.0, 12.0]
            },
            {
                'task': 'task2',
                'reward': [0.8, 0.9],
                'finish_time': [5.0, 6.0]
            }
        ]
    },
    {
        'iter': 2000,
        'result': [
            {
                'task': 'task1',
                'reward': [0.6, 0.7, 0.8],
                'finish_time': [9.0, 10.0, 11.0]
            },
            {
                'task': 'task2',
                'reward': [0.9, 1.0],
                'finish_time': [4.0, 5.0]
            }
        ]
    },
    {
        'iter': 3000,
        'result': [
            {
                'task': 'task1',
                'reward': [0.7, 0.8, 0.9],
                'finish_time': [8.0, 9.0, 10.0]
            },
            {
                'task': 'task2',
                'reward': [1.0, 1.1],
                'finish_time': [3.0, 4.0]
            }
        ]
    }
]

# Write stats to file
with open(stats_file, 'w') as f:
    json.dump(stats_data, f)

# Create a minimal args object
class Args:
    def __init__(self):
        self.max_iter = 400000

args = Args()

ARGS = {
    'stats_path': stats_file,
    'args': args,
    'target_task': 'all',
    'iter': 'final'
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1
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_target_task_avg Ran
ran, honoured the contract we drafted
yycdavid/program-synthesis-guided-rl/analysis/plots.py
code served (permissive licence) · get_code("90e045f67ec5ac65")
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) 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 a temporary directory and stats file
temp_dir = tempfile.mkdtemp()
stats_file = os.path.join(temp_dir, 'stats.json')

# Create sample stats data matching the expected structure
stats_data = [
    {
        'iter': 1000,
        'result': [
            {
                'task': 'task_a',
                'reward': [0.5, 0.6, 0.7],
                'finish_time': [10.0, 11.0, 12.0]
            },
            {
                'task': 'task_b',
                'reward': [0.4, 0.5],
                'finish_time': [15.0, 16.0]
            }
        ]
    },
    {
        'iter': 2000,
        'result': [
            {
                'task': 'task_a',
                'reward': [0.6, 0.7, 0.8],
                'finish_time': [9.0, 10.0, 11.0]
            },
            {
                'task': 'task_b',
                'reward': [0.5, 0.6],
                'finish_time': [14.0, 15.0]
            }
        ]
    }
]

# Write stats to file
with open(stats_file, 'w') as f:
    json.dump(stats_data, f)

# Define ARGS for get_target_task_avg function
ARGS = {
    'stats_path': stats_file,
    'target_tasks': 'task_a'
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1
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
loss_function Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
yycdavid/program-synthesis-guided-rl/train_cvae.py
code served (permissive licence) · get_code("0ba8b6945b3ef257")
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) and returned Tensor, 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 torch.nn.functional as F

torch.manual_seed(0)

# Create small tensors for the loss function
batch_size = 2
feature_dim = 4

# recon_x: reconstructed output from decoder, should be in [0, 1] for binary cross-entropy
recon_x = torch.sigmoid(torch.randn(batch_size, feature_dim))

# x: original input, should be in [0, 1] for binary cross-entropy
x = torch.rand(batch_size, feature_dim)

# mu: mean of the latent distribution
mu = torch.randn(batch_size, feature_dim)

# logvar: log variance of the latent distribution
logvar = torch.randn(batch_size, feature_dim)

ARGS = {
    "recon_x": recon_x,
    "x": x,
    "mu": mu,
    "logvar": logvar
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1
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 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

Also present in other papers' code

Of 4 samples lifted from this paper's linked repositories, 4 carry a function identity and 0 do not. Of the 4 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 4 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

A key challenge for reinforcement learning is solving long-horizon planning problems. Recent work has leveraged programs to guide reinforcement learning in these settings. However, these approaches impose a high manual burden on the user since they must provide a guiding program for every new task. Partially observed environments further complicate the programming task because the program must implement a strategy that correctly, and ideally optimally, handles every possible configuration of the hidden regions of the environment. We propose a new approach, model predictive program synthesis (MPPS), that uses program synthesis to automatically generate the guiding programs. It trains a generative model to predict the unobserved portions of the world, and then synthesizes a program based on samples from this model in a way that is robust to its uncertainty. In our experiments, we show that our approach significantly outperforms non-program-guided approaches on a set of challenging benchmarks, including a 2D Minecraft-inspired environment where the agent must complete a complex sequence of subtasks to achieve its goal, and achieves a similar performance as using handcrafted programs to guide the agent. Our results demonstrate that our approach can obtain the benefits of program-guided reinforcement learning without requiring the user to provide a new guiding program for every new task.

For agents

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

get_harvested_code_for_paper("2102.11137")
get_code_for_paper("2102.11137")
have("2102.11137")

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

curl https://syntology.ai/api/ran/2102.11137.json

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

[![Syntology run record](https://syntology.ai/api/ran/2102.11137.svg)](https://syntology.ai/paper/2102.11137)

Connect an agent — have() is free.