SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 1910.02425 · 2019

Structured Object-Aware Physics Prediction for Video Modeling and Planning

arXiv · PDF · Open in the Atlas

Code that ran

7 of 8 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 7 have no contract verdict; 0 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/1910.02425.json · badge: /api/ran/1910.02425.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
jlko/STOVE canonical (classifier label)
classifier basis: census
7 ran · 1 no recorded run
FunctionStatusWhere it lives
encode_img Ran
ran on a synthesized input
jlko/STOVE/model/mcts/mcts_stove.py
code served (permissive licence) · get_code("93fcda1b3ffc3f86")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned Tensor, ndim 5, dtype family float, shape (2, 2, 5, 4, 3) 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)

# Create a small numpy array with shape (batch, batch, width, height, channels)
# Using distinct sizes: 2, 2, 4, 3, 5 for each dimension
img = np.random.randn(2, 2, 4, 3, 5).astype(np.float32)

ARGS = {
    "img": img
}
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 Tensor, ndim 5, dtype family float, shape (2, 2, 5, 4, 3) — shape (2, 2, 5, 4, 3) 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
generate_fitting_run Ran
ran on a synthesized input
jlko/STOVE/model/envs/envs.py
code served (permissive licence) · get_code("29f7fbb68f8a1858")
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

class MockEnv:
    """Mock environment that satisfies the interface used by generate_fitting_run."""
    
    def __init__(self, n=2, r=1.0, m=1.0, hw=10, granularity=10, res=50, t=0.01, 
                 init_v_factor=None, friction_coefficient=0.3, seed=None, 
                 sprites=False, use_colors=None, drift=False):
        self.n = n
        self.r = np.array([r] * n) if isinstance(r, (int, float)) else r
        self.m = m
        self.hw = hw
        self.granularity = granularity
        self.res = res
        self.t = t
        self.init_v_factor = init_v_factor
        self.friction_coefficient = friction_coefficient
        self.seed = seed
        self.sprites = sprites
        self.use_colors = use_colors
        self.drift = drift
        self.step_count = 0
    
    def get_obs_shape(self):
        """Return observation shape (res, res, 3) for a simple RGB image."""
        return (self.res, self.res, 3)
    
    def step(self):
        """Step the environment and return obs, state, and info."""
        # Return image (res, res, 3), state (n, 4), and empty info dict
        img = np.random.rand(self.res, self.res, 3)
        # State: n particles, each with 4 values (x, y, vx, vy)
        # Position between 0 and hw to pass quality check
        state = np.random.rand(self.n, 4) * self.hw
        state[:, 2:] = np.random.rand(self.n, 2) * 0.1  # Small velocities
        info = {}
        self.step_count += 1
        return img, state, info

ARGS = {
    'env_class': MockEnv,
    'run_len': 5,
    'run_num': 2,
    'max_tries': 10,
    'res': 32,
    'n': 2,
    'r': 1.0,
    'dt': 0.01,
    'granularity': 10,
    'fc': 0.3,
    'hw': 10,
    'm': 1.0,
    'seed': None,
    'init_v_factor': None,
    'check_overlap': False,
    'sprites': False,
    'use_colors': None,
    'drift': False,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, tqdm — 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
load_and_encode Ran
ran on a synthesized input
jlko/STOVE/model/rl/main_rl.py
code served (permissive licence) · get_code("e40725e62306ffd3")
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
import torch.nn as nn
import pickle
import tempfile
import os

torch.manual_seed(0)

# Create a minimal state_model that matches the function's usage
class MockStateModel(nn.Module):
    def __init__(self):
        super().__init__()
        self.linear = nn.Linear(1, 1)
    
    def forward(self, x, step, action=None, pretrain=False, future=None):
        batch_size = x.shape[0]
        seq_len = x.shape[1]
        
        # Return format: (something, init_state, something)
        # init_state needs 'obj_appearances' and 'z'
        init_state = {
            'obj_appearances': torch.randn(batch_size, seq_len, 3, 3),
            'z': torch.randn(batch_size, seq_len, 3, 18)
        }
        
        return None, init_state, None

# Create temporary pickle file with mock data
temp_dir = tempfile.mkdtemp()
pickle_path = os.path.join(temp_dir, 'data.pkl')

# X shape: (batch_size, seq_len, height, width, channels)
# We need seq_len > 8 so that range(len(X[0]) - 8) has at least one iteration
X = torch.randn(2, 12, 4, 5, 3).numpy()
action = torch.randn(2, 12, 2).numpy()

data = {
    'X': X,
    'action': action
}

with open(pickle_path, 'wb') as f:
    pickle.dump(data, f)

ARGS = {
    'path': pickle_path,
    'state_model': MockStateModel(),
    'device': 'cpu'
}
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
multi_one_hot Ran
ran on a synthesized input
jlko/STOVE/model/mcts/mcts_stove.py
code served (permissive licence) · get_code("4414a482ef6ea681")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned Tensor, ndim 3, dtype family float, shape (1, 4, 5) 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)

# actions: list of action indices
# action_shape: int representing the maximum number of entries in the one-hot vector
actions = [0, 2, 1, 3]  # list of 4 action indices
action_shape = 5  # maximum action index is 5

ARGS = {
    "actions": actions,
    "action_shape": action_shape,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Tensor, ndim 3, dtype family float, shape (1, 4, 5) — shape (1, 4, 5) 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
norm Ran
ran on a synthesized input
jlko/STOVE/model/envs/envs.py
code served (permissive licence) · get_code("66d9b15aaa1af8ef")
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 ndarray, ndim 2, dtype family float, shape (2, 1) 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 numpy as np

# Test case 1: 1D array (vector)
x_1d = np.array([1.0, 2.0, 3.0, 4.0])

# Test case 2: 2D array (matrix with 2 rows, 4 columns)
x_2d = np.array([[1.0, 2.0, 3.0, 4.0],
                 [5.0, 6.0, 7.0, 8.0]])

# Use the 2D case as the primary argument since it tests both branches
ARGS = {
    'x': x_2d
}
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 ndarray, ndim 2, dtype family float, shape (2, 1) — shape (2, 1) 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
parse_wrapper Ran
ran on a synthesized input
jlko/STOVE/model/envs/envs.py
code served (permissive licence) · get_code("17a350274b0cf4c1")
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 Namespace. 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 argparse
import torch

torch.manual_seed(0)

# The function parse_wrapper takes a list of command-line arguments (strings)
# and parses them using argparse. It expects script_args to be a list of strings
# that can be passed to parser.parse_args().

ARGS = {
    'script_args': ['--test-gen', '--no-save']
}
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 Namespace — a Namespace output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a Namespace 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
tile Ran
ran on a synthesized input
jlko/STOVE/model/mcts/mcts_stove.py
code served (permissive licence) · get_code("78aa7a12549439c7")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned Tensor, ndim 2, dtype family float, shape (6, 4) as a later run recorded it. 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, 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)

# Create a small test tensor
# Shape: (2, 4) - batch of 2, dimension of 4
a = torch.randn(2, 4)

# Dimension to tile along (0 or 1)
dim = 0

# Number of repeats per tile
n_tile = 3

ARGS = {
    'a': a,
    'dim': dim,
    'n_tile': n_tile
}
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 Tensor, ndim 2, dtype family float, shape (6, 4) — shape (6, 4) 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
parse_args Not yet run
unverified (harvested, no recorded run)
jlko/STOVE/model/rl/main_rl.py
code served (permissive licence) · get_code("6671b44fe3654fd2")
How it ran

no invocation recorded.

Also present in other papers' code

Of 8 samples lifted from this paper's linked repositories, 8 carry a function identity and 0 do not. Of the 8 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 8 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

When humans observe a physical system, they can easily locate objects, understand their interactions, and anticipate future behavior, even in settings with complicated and previously unseen interactions. For computers, however, learning such models from videos in an unsupervised fashion is an unsolved research problem. In this paper, we present STOVE, a novel state-space model for videos, which explicitly reasons about objects and their positions, velocities, and interactions. It is constructed by combining an image model and a dynamics model in compositional manner and improves on previous work by reusing the dynamics model for inference, accelerating and regularizing training. STOVE predicts videos with convincing physical behavior over hundreds of timesteps, outperforms previous unsupervised models, and even approaches the performance of supervised baselines. We further demonstrate the strength of our model as a simulator for sample efficient model-based control in a task with heavily interacting objects.

For agents

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

get_harvested_code_for_paper("1910.02425")
get_code_for_paper("1910.02425")
have("1910.02425")

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

curl https://syntology.ai/api/ran/1910.02425.json

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

[![Syntology run record](https://syntology.ai/api/ran/1910.02425.svg)](https://syntology.ai/paper/1910.02425)

Connect an agent — have() is free.