SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2402.02500 · NeurIPS · 2024

Point Cloud Matters: Rethinking the Impact of Different Observation Spaces on Robot Learning

Wanli Ouyang, Di Huang, Haoyi Zhu, Tong He, Yating Wang, Weicai Ye, Resnet Vit, Multivit Spunet, Pointnet Multimae

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; 2 honoured the contract we drafted, 0 violated it and 0 have no contract verdict; 6 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/2402.02500.json · badge: /api/ran/2402.02500.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
HaoyiZhu/PointCloudMatters canonical (classifier label)
classifier basis: census
8 ran · 1 no recorded run
FunctionStatusWhere it lives
check_state_dict Ran
ran; our contract draft was wrong, not the code
HaoyiZhu/PointCloudMatters/src/utils/common_utils.py
code served (permissive licence) · get_code("a488db5b59ecea96")
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 torch
import logging

torch.manual_seed(0)

# Create a logger
logger = logging.getLogger("test_logger")
logger.setLevel(logging.INFO)
handler = logging.StreamHandler()
logger.addHandler(handler)

# Create state_dict (model's current state)
state_dict = {
    'layer1.weight': torch.randn(4, 3),
    'layer1.bias': torch.randn(4),
    'layer2.weight': torch.randn(5, 4),
    'layer2.bias': torch.randn(5),
}

# Create state_dict_disk (weights from disk)
# Some keys match, some don't; some shapes match, some don't
state_dict_disk = {
    'layer1.weight': torch.randn(4, 3),  # matches in key and shape
    'layer1.bias': torch.randn(3),  # matches in key but different shape
    'layer2.weight': torch.randn(5, 4),  # matches in key and shape
    'layer3.weight': torch.randn(6, 5),  # key not in state_dict
}

ARGS = {
    'state_dict': state_dict,
    'state_dict_disk': state_dict_disk,
    'logger': logger,
}
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 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
import_modules_from_strings Ran
ran; our contract draft was wrong, not the code
HaoyiZhu/PointCloudMatters/src/utils/misc.py
code served (permissive licence) · get_code("0bccf0e06e6688aa")
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.

import warnings
from importlib import import_module

def import_modules_from_strings(imports, allow_failed_imports=False):
    """Import modules from the given list of strings.

    Args:
        imports (list | str | None): The given module names to be imported.
        allow_failed_imports (bool): If True, the failed imports will return
            None. Otherwise, an ImportError is raise. Default: False.

    Returns:
        list[module] | module | None: The imported modules.

    Examples:
        >>> osp, sys = import_modules_from_strings(
        ...     ['os.path', 'sys'])
        >>> import os.path as osp_
        >>> import sys as sys_
        >>> assert osp == osp_
        >>> assert sys == sys_
    """
    if not imports:
        return
    single_import = False
    if isinstance(imports, str):
        single_import = True
        imports = [imports]
    if not isinstance(imports, list):
        raise TypeError(f'custom_imports must be a list but got type {type(imports)}')
    imported = []
    for imp in imports:
        if not isinstance(imp, str):
            raise TypeError(f'{imp} is of type {type(imp)} and cannot be imported.')
        try:
            imported_tmp = import_module(imp)
        except ImportError:
            if allow_failed_imports:
                warnings.warn(f'{imp} failed to import and is ignored.', UserWarning)
                imported_tmp = None
            else:
                raise ImportError
        imported.append(imported_tmp)
    if single_import:
        imported = imported[0]
    return imported


ARGS = {
    "imports": ["os.path", "sys"],
    "allow_failed_imports": 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 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
interpolate_linear Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
HaoyiZhu/PointCloudMatters/src/utils/misc.py
code served (permissive licence) · get_code("129f75ac28d5d99f")
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 1, dtype family float, shape (4,) 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

# Set random seed for reproducibility
np.random.seed(0)

# Define arguments for interpolate_linear
target_t = 5
t1 = 0
t2 = 10
x1 = np.array([1.0, 2.0, 3.0, 4.0])
x2 = np.array([5.0, 6.0, 7.0, 8.0])

ARGS = {
    "target_t": target_t,
    "t1": t1,
    "t2": t2,
    "x1": x1,
    "x2": x2,
}
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 1, dtype family float, shape (4,) — shape (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
is_seq_of Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
HaoyiZhu/PointCloudMatters/src/utils/misc.py
code served (permissive licence) · get_code("3551b62f3276001a")
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 torch
import numpy as np
from collections import abc

torch.manual_seed(0)

# Test case 1: A valid sequence of integers
seq1 = [1, 2, 3, 4, 5]
expected_type1 = int
seq_type1 = list

# Test case 2: A valid sequence of strings with tuple type
seq2 = ("hello", "world", "test")
expected_type2 = str
seq_type2 = tuple

# Test case 3: Using default seq_type (abc.Sequence)
seq3 = [1.5, 2.5, 3.5]
expected_type3 = float
seq_type3 = None

# Test case 4: A valid sequence of torch tensors
seq4 = [torch.randn(2, 4), torch.randn(2, 4)]
expected_type4 = torch.Tensor
seq_type4 = list

ARGS = {
    "seq": seq1,
    "expected_type": expected_type1,
    "seq_type": seq_type1
}
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 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
load_bytes Ran
ran; our contract draft was wrong, not the code
HaoyiZhu/PointCloudMatters/src/utils/io.py
code served (permissive licence) · get_code("d9118aea6a0811f4")
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 bytes. 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 file with some binary content
temp_file = tempfile.NamedTemporaryFile(delete=False, suffix='.bin')
temp_filename = temp_file.name
temp_file.write(b'test binary content here')
temp_file.close()

try:
    ARGS = {
        'filename': temp_filename
    }
finally:
    # Clean up will happen after test, but we keep the file around for the function to read
    pass
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 bytes — a bytes output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a bytes 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_numpy_text Ran
ran, honoured the contract we drafted
HaoyiZhu/PointCloudMatters/src/utils/io.py
code served (permissive licence) · get_code("7843f915d988dd5e")
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, 4) 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
import tempfile
import os

# Create a temporary numpy file for testing
temp_dir = tempfile.mkdtemp()
temp_file = os.path.join(temp_dir, "test_data.txt")

# Write sample data to the temporary file
sample_data = np.array([[1.0, 2.0, 3.0, 4.0],
                        [5.0, 6.0, 7.0, 8.0]])
np.savetxt(temp_file, sample_data)

ARGS = {
    "filename": temp_file
}
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, 4) — shape (2, 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
load_text Ran
ran; our contract draft was wrong, not the code
HaoyiZhu/PointCloudMatters/src/utils/io.py
code served (permissive licence) · get_code("76e8fea908592e80")
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.

import tempfile
import os

# Create a temporary file with some test content
temp_file = tempfile.NamedTemporaryFile(mode='w', delete=False, suffix='.txt')
temp_file.write("line one\nline two\nline three\n")
temp_file.close()

ARGS = {
    'filename': temp_file.name
}
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
reduce_mean Ran
ran, honoured the contract we drafted
HaoyiZhu/PointCloudMatters/src/utils/common_utils.py
code served (permissive licence) · get_code("0d86269cf8e78d35")
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 (2, 4) 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.distributed as dist

torch.manual_seed(0)

# Create a simple tensor argument
tensor = torch.randn(2, 4)

ARGS = {
    "tensor": tensor
}
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 2, dtype family float, shape (2, 4) — shape (2, 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
multi_apply Not yet run
unverified (harvested, no recorded run)
HaoyiZhu/PointCloudMatters/src/utils/common_utils.py
code served (permissive licence) · get_code("f7898a5762a4d3e4")
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, 8 are also present in other papers whose code we hold (4 with at least one paper dated earlier, 0 the same month, 7 later) and 1 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.

FunctionAlso present inEarliest-datedOthers named
multi_apply 10 other papers
9 dated earlier, 1 later
1911.09070 (2019-11)
EfficientDet: Scalable and Efficient Object Detection
2003.10152 (2020-03)
SOLOv2: Dynamic and Fast Instance Segmentation
2006.05683 (2020-06)
TubeTK: Adopting Tubes to Track Multi-Object in a One-Step Training Model
and 7 more not named here
reduce_mean 5 other papers
2 dated earlier, 3 later
2010.15831 (2020-10)
RelationNet++: Bridging Visual Representations for Object Detection via Transformer Decoder
2308.05581 (2023-08)
Category Feature Transformer for Semantic Segmentation
2410.08208 (2024-10)
SPA: 3D Spatial-Awareness Enables Effective Embodied Representation
and 2 more not named here
is_seq_of 3 other papers
1 dated earlier, 2 later
2306.13856 (2023-06)
Learning-to-Rank Meets Language: Boosting Language-Driven Ordering Alignment for Ordinal Classification
2503.06800 (2025-03)
VideoPhy-2: A Challenging Action-Centric Physical Commonsense Evaluation in Video Generation
2503.16429 (2025-03)
Sonata: Self-Supervised Learning of Reliable Point Representations
check_state_dict 1 other paper
1 later
2410.08208 (2024-10)
SPA: 3D Spatial-Awareness Enables Effective Embodied Representation
import_modules_from_strings 1 other paper
1 dated earlier
2304.12308 (2023-04)
A SUBMISSION TO IEEE TRANSACTION ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE 1 Segment Anything in 3D with Radiance Fields
load_bytes 1 other paper
1 later
2410.08208 (2024-10)
SPA: 3D Spatial-Awareness Enables Effective Embodied Representation
load_numpy_text 1 other paper
1 later
2410.08208 (2024-10)
SPA: 3D Spatial-Awareness Enables Effective Embodied Representation
load_text 1 other paper
1 later
2410.08208 (2024-10)
SPA: 3D Spatial-Awareness Enables Effective Embodied Representation

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

In robot learning, the observation space is crucial due to the distinct characteristics of different modalities, which can potentially become a bottleneck alongside policy design. In this study, we explore the influence of various observation spaces on robot learning, focusing on three predominant modalities: RGB, RGB-D, and point cloud. We introduce OBSBench, a benchmark comprising two simulators and 125 tasks, along with standardized pipelines for various encoders and policy baselines. Extensive experiments on diverse contact-rich manipulation tasks reveal a notable trend: point cloud-based methods, even those with the simplest designs, frequently outperform their RGB and RGB-D counterparts. This trend persists in both scenarios: training from scratch and utilizing pre-training. Furthermore, our findings demonstrate that point cloud observations often yield better policy performance and significantly stronger generalization capabilities across various geometric and visual conditions. These outcomes suggest that the 3D point cloud is a valuable observation modality for intricate robotic tasks. We also suggest that incorporating both appearance and coordinate information can enhance the performance of point cloud methods. We hope our work provides valuable insights and guidance for designing more generalizable and robust robotic models. 38th Conference on Neural Information Processing Systems (NeurIPS 2024) Track on Datasets and Benchmarks.

For agents

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

get_harvested_code_for_paper("2402.02500")
get_code_for_paper("2402.02500")
have("2402.02500")

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

curl https://syntology.ai/api/ran/2402.02500.json

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

[![Syntology run record](https://syntology.ai/api/ran/2402.02500.svg)](https://syntology.ai/paper/2402.02500)

Connect an agent — have() is free.