SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2110.06192 · 2021

Beyond Pick-and-Place: Tackling Robotic Stacking of Diverse Shapes

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; 0 honoured the contract we drafted, 0 violated it and 4 have no contract verdict; 0 failures 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/2110.06192.json · badge: /api/ran/2110.06192.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
deepmind/rgb_stacking canonical (classifier label)
classifier basis not recorded on the link
4 ran · 0 no recorded run
FunctionStatusWhere it lives
has_ground_collision Ran
ran on a synthesized input
deepmind/rgb_stacking/rgb_stacking/physics_utils.py
code served (permissive licence) · get_code("e6ddb2ae0030cf34")
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 itertools
from unittest.mock import Mock, MagicMock

DEFAULT_COLLISION_MARGIN = 1e-08
GROUND_GEOM_PREFIXES = ['work_surface', 'ground']

# Create a mock physics object with the required structure
physics = Mock()

# Create mock contact objects
contact1 = Mock()
contact1.dist = 0.5e-08  # Within margin (< 1e-08)
contact1.geom1 = 0
contact1.geom2 = 1

contact2 = Mock()
contact2.dist = 5e-09  # Within margin
contact2.geom1 = 2
contact2.geom2 = 3

# Set up physics.data.contact
physics.data = Mock()
physics.data.contact = [contact1, contact2]

# Set up physics.model.id2name to return geom names
def mock_id2name(geom_id, geom_type):
    geom_names = {
        0: 'work_surface_left',
        1: 'gripper_finger',
        2: 'ground_plane',
        3: 'object_body'
    }
    return geom_names.get(geom_id, 'unknown')

physics.model = Mock()
physics.model.id2name = mock_id2name

# Define ARGS for has_ground_collision function
ARGS = {
    'physics': physics,
    'collision_geom_prefix': 'gripper',
    'margin': DEFAULT_COLLISION_MARGIN
}
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
has_object_collision Ran
ran on a synthesized input
deepmind/rgb_stacking/rgb_stacking/physics_utils.py
code served (permissive licence) · get_code("7c9d0777b902c3e4")
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 itertools
from unittest.mock import MagicMock, Mock

# Create a mock physics object with the required structure
physics = MagicMock()

# Create mock contact objects
contact1 = MagicMock()
contact1.dist = 1e-9  # Very small distance (collision)
contact1.geom1 = 0
contact1.geom2 = 1

contact2 = MagicMock()
contact2.dist = 1e-6  # Larger distance (no collision with default margin)
contact2.geom1 = 2
contact2.geom2 = 3

# Set up the contacts list
physics.data.contact = [contact1, contact2]

# Set up the model with id2name method
physics.model.id2name = MagicMock(side_effect=lambda geom_id, geom_type: {
    0: 'rgb_geom_1',
    1: 'rgb_object_1',
    2: 'other_geom',
    3: 'other_object'
}.get(geom_id, f'geom_{geom_id}'))

# Define arguments for has_object_collision
ARGS = {
    'physics': physics,
    'collision_geom_prefix': 'rgb_geom',
    'margin': 1e-08
}
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
has_robot_collision Ran
ran on a synthesized input
deepmind/rgb_stacking/rgb_stacking/physics_utils.py
code served (permissive licence) · get_code("fafb1fbec95a7e97")
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 itertools
from unittest.mock import Mock, MagicMock
import numpy as np

DEFAULT_COLLISION_MARGIN = 1e-08
ROBOT_GEOM_PREFIXES = ['robot']

def has_collision(physics, collision_geom_prefix_1, collision_geom_prefix_2, margin=DEFAULT_COLLISION_MARGIN):
    """Check for collisions between geoms."""
    for contact in physics.data.contact:
        if contact.dist > margin:
            continue
        geom1_name = physics.model.id2name(contact.geom1, 'geom')
        geom2_name = physics.model.id2name(contact.geom2, 'geom')
        for pair in itertools.product(collision_geom_prefix_1, collision_geom_prefix_2):
            if geom1_name.startswith(pair[0]) and geom2_name.startswith(pair[1]) or (geom2_name.startswith(pair[0]) and geom1_name.startswith(pair[1])):
                return True
    return False

def has_robot_collision(physics, collision_geom_prefix, margin=DEFAULT_COLLISION_MARGIN):
    """Check for collisions between geoms and the robot."""
    return has_collision(physics=physics, collision_geom_prefix_1=[collision_geom_prefix], collision_geom_prefix_2=ROBOT_GEOM_PREFIXES, margin=margin)


class MockContact:
    def __init__(self, geom1, geom2, dist):
        self.geom1 = geom1
        self.geom2 = geom2
        self.dist = dist


class MockModel:
    def __init__(self):
        self.geom_names = {
            0: 'obstacle_1',
            1: 'robot_link_1',
            2: 'obstacle_2',
            3: 'robot_link_2'
        }
    
    def id2name(self, geom_id, geom_type):
        return self.geom_names.get(geom_id, f'{geom_type}_{geom_id}')


class MockData:
    def __init__(self):
        self.contact = [
            MockContact(geom1=0, geom2=1, dist=0.5e-08),
            MockContact(geom1=2, geom2=3, dist=0.1e-08)
        ]


class MockPhysics:
    def __init__(self):
        self.model = MockModel()
        self.data = MockData()


physics = MockPhysics()
collision_geom_prefix = 'obstacle'
margin = DEFAULT_COLLISION_MARGIN

ARGS = {
    'physics': physics,
    'collision_geom_prefix': collision_geom_prefix,
    'margin': margin
}
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 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
tanh_squared Ran
ran on a synthesized input
deepmind/rgb_stacking/rgb_stacking/reward_functions.py
code served (permissive licence) · get_code("5d32c1f0bb8e70d0")
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 tanh_squared function
x = np.array([0.5, -0.3, 0.2, 0.1])  # A small numpy array representing error
margin = 1.5  # A positive float
loss_at_margin = 0.8  # A float between 0 and 1

ARGS = {
    "x": x,
    "margin": margin,
    "loss_at_margin": loss_at_margin,
}
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

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

We study the problem of robotic stacking with objects of complex geometry. We propose a challenging and diverse set of such objects that was carefully designed to require strategies beyond a simple "pick-and-place" solution. Our method is a reinforcement learning (RL) approach combined with vision-based interactive policy distillation and simulation-to-reality transfer. Our learned policies can efficiently handle multiple object combinations in the real world and exhibit a large variety of stacking skills. In a large experimental study, we investigate what choices matter for learning such general vision-based agents in simulation, and what affects optimal transfer to the real robot. We then leverage data collected by such policies and improve upon them with offline RL. A video and a blog post of our work are provided as supplementary material.

For agents

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

get_harvested_code_for_paper("2110.06192")
get_code_for_paper("2110.06192")
have("2110.06192")

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

curl https://syntology.ai/api/ran/2110.06192.json

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

[![Syntology run record](https://syntology.ai/api/ran/2110.06192.svg)](https://syntology.ai/paper/2110.06192)

Connect an agent — have() is free.