SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2402.10885 · 2024

3D Diffuser Actor: Policy Diffusion with 3D Scene Representations

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 7 have no contract verdict; 1 failure 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/2402.10885.json · badge: /api/ran/2402.10885.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
nickgkan/3d_diffuser_actor community (classifier label)
classifier basis not recorded on the link
8 ran · 1 no recorded run
FunctionStatusWhere it lives
all_gather Ran
ran; our contract draft was wrong, not the code
nickgkan/3d_diffuser_actor/engine.py
code served (permissive licence) · get_code("82c73aecaae88536")
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. 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 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 pickle
import torch
import torch.distributed as dist

torch.manual_seed(0)

# Since the function uses dist.is_available() and dist.is_initialized(),
# which will return False in a non-distributed environment,
# the function will take the early return path: return [data]
# So we just need to pass any picklable object as data

data = {
    "key1": "value1",
    "key2": [1, 2, 3],
    "key3": {"nested": "dict"},
    "tensor": torch.randn(2, 4)
}

ARGS = {
    "data": data
}
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 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
convert_efficient_fpn Ran
ran on a synthesized input
nickgkan/3d_diffuser_actor/diffuser_actor/utils/converter.py
code served (permissive licence) · get_code("caeb119cf55a786d")
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 MockModel. 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
from torchvision.ops import FeaturePyramidNetwork
from collections import OrderedDict

torch.manual_seed(0)

# Build a minimal model that the convert_efficient_fpn function expects
class PredictionHead(nn.Module):
    def __init__(self):
        super().__init__()
        # feat_scales determines which level to stop at
        self.feat_scales = 2
        # feature_map_pyramid is indexed by feat_scales - 1
        self.feature_map_pyramid = ['res2', 'res3', 'res4', 'res5']

class MockModel(nn.Module):
    def __init__(self):
        super().__init__()
        self.prediction_head = PredictionHead()
        
        # Build a FeaturePyramidNetwork as a child module
        # in_channels_list should have 4 channels for each of 3 levels
        in_channels_list = [64, 128, 256]
        out_channels = 256
        
        self.fpn = FeaturePyramidNetwork(
            in_channels_list=in_channels_list,
            out_channels=out_channels
        )

model = MockModel()

ARGS = {
    'model': model
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch, torchvision — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type MockModel — a MockModel output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a MockModel 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_resnet18 Ran
ran on a synthesized input
nickgkan/3d_diffuser_actor/diffuser_actor/utils/resnet.py
code served (permissive licence) · get_code("5122b84a1c2fb599")
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 torch
from torchvision import transforms
from torchvision.models.resnet import BasicBlock, Bottleneck, ResNet

torch.manual_seed(0)

ARGS = {
    'pretrained': False,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch, torchvision — 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
load_resnet50 Ran
ran on a synthesized input
nickgkan/3d_diffuser_actor/diffuser_actor/utils/resnet.py
code served (permissive licence) · get_code("592a9914f3d643b4")
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 torch
from torchvision import transforms
from torchvision.models.resnet import BasicBlock, Bottleneck, ResNet

torch.manual_seed(0)

# Set up ARGS dict for load_resnet50 function
ARGS = {
    'pretrained': False,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch, torchvision — 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
multi_head_attention_forward Ran
ran on a synthesized input
nickgkan/3d_diffuser_actor/diffuser_actor/utils/multihead_custom_attention.py
code served (permissive licence) · get_code("616805ec8f4f95fb")
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.functional as F

torch.manual_seed(0)

# Set up dimensions
embed_dim = 8
num_heads = 2
head_dim = embed_dim // num_heads
tgt_len = 3
src_len = 4
bsz = 2

# Create query, key, value tensors
query = torch.randn(tgt_len, bsz, embed_dim)
key = torch.randn(src_len, bsz, embed_dim)
value = torch.randn(src_len, bsz, embed_dim)

# Create projection weights and biases
in_proj_weight = torch.randn(3 * embed_dim, embed_dim)
in_proj_bias = torch.randn(3 * embed_dim)

# Output projection weights and biases
out_proj_weight = torch.randn(embed_dim, embed_dim)
out_proj_bias = torch.randn(embed_dim)

# Separate projection weights (for use_separate_proj_weight=True case)
q_proj_weight = torch.randn(embed_dim, embed_dim)
k_proj_weight = torch.randn(embed_dim, embed_dim)
v_proj_weight = torch.randn(embed_dim, embed_dim)

# Optional bias tensors
bias_k = None
bias_v = None

# Masks
key_padding_mask = torch.zeros(bsz, src_len, dtype=torch.bool)
attn_mask = None

# Static k, v (for static attention)
static_k = None
static_v = None

# Additional parameters
k_mem = None
v_mem = None
gate_attn = None
mem_mask = None
rotary_pe = None

ARGS = {
    'query': query,
    'key': key,
    'value': value,
    'embed_dim_to_check': embed_dim,
    'num_heads': num_heads,
    'in_proj_weight': in_proj_weight,
    'in_proj_bias': in_proj_bias,
    'bias_k': bias_k,
    'bias_v': bias_v,
    'add_zero_attn': False,
    'dropout_p': 0.1,
    'out_proj_weight': out_proj_weight,
    'out_proj_bias': out_proj_bias,
    'training': True,
    'key_padding_mask': key_padding_mask,
    'need_weights': True,
    'attn_mask': attn_mask,
    'use_separate_proj_weight': False,
    'q_proj_weight': q_proj_weight,
    'k_proj_weight': k_proj_weight,
    'v_proj_weight': v_proj_weight,
    'static_k': static_k,
    'static_v': static_v,
    'slot_competition': False,
    'rotary_pe': rotary_pe,
    'return_kv': False,
    'k_mem': k_mem,
    'v_mem': v_mem,
    'gate_attn': gate_attn,
    'mem_mask': mem_mask,
}
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
norm_tensor Ran
ran on a synthesized input
nickgkan/3d_diffuser_actor/diffuser_actor/utils/utils.py
code served (permissive licence) · get_code("bad211cd427d59bf")
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

torch.manual_seed(0)

# Create a small tensor with batch size 2 and dimension 4
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
normalise_quat Ran
ran on a synthesized input
nickgkan/3d_diffuser_actor/diffuser_actor/utils/utils.py
code served (permissive licence) · get_code("6a887764a925a2e4")
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

torch.manual_seed(0)

# The function normalise_quat takes a tensor x and normalizes quaternions
# It expects x to have shape (..., 4) where the last dimension represents quaternion components
# It computes the norm along the last dimension and divides x by it

# Create a small tensor with shape (2, 4) - batch of 2 quaternions
x = torch.randn(2, 4)

ARGS = {
    "x": x
}
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
sample_ghost_points_grid Ran
ran on a synthesized input
nickgkan/3d_diffuser_actor/diffuser_actor/utils/utils.py
code served (permissive licence) · get_code("5ee893d8fc620de5")
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 (125, 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 numpy as np
import einops

def sample_ghost_points_grid(bounds, num_points_per_dim=10):
    x_ = np.linspace(bounds[0][0], bounds[1][0], num_points_per_dim)
    y_ = np.linspace(bounds[0][1], bounds[1][1], num_points_per_dim)
    z_ = np.linspace(bounds[0][2], bounds[1][2], num_points_per_dim)
    x, y, z = np.meshgrid(x_, y_, z_, indexing='ij')
    ghost_points = einops.rearrange(np.stack([x, y, z]), 'n x y z -> (x y z) n')
    return ghost_points

# Set up bounds as a 2x3 array defining min and max corners of a 3D box
bounds = np.array([
    [0.0, 0.0, 0.0],      # min corner (x_min, y_min, z_min)
    [10.0, 20.0, 30.0]    # max corner (x_max, y_max, z_max)
])

ARGS = {
    'bounds': bounds,
    'num_points_per_dim': 5
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
einops, 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 (125, 3) — shape (125, 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
convert_multihead_flash_attention Not yet run
unverified (harvested, no recorded run)
nickgkan/3d_diffuser_actor/diffuser_actor/utils/converter.py
code served (permissive licence) · get_code("c921e2f2f2ff9812")
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, 2 are also present in other papers whose code we hold (2 with at least one paper dated earlier, 1 the same month, 1 later) and 7 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
all_gather 30 other papers
22 dated earlier, 1 the same month, 6 later, 1 same year, order not determinable
1811.07628 (2018-11)
ATOM: Accurate Tracking by Overlap Maximization
1811.10200 (2018-11)
IDD: A Dataset for Exploring Problems of Autonomous Navigation in Unconstrained Environments
2007.02419 (2020-07)
Attention-based Joint Detection of Object and Semantic Part
and 27 more not named here
norm_tensor 1 other paper
1 dated earlier
2209.04899 (2022-09)
Instruction-driven history-aware policies for robotic manipulations

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

Diffusion policies are conditional diffusion models that learn robot action distributions conditioned on the robot and environment state. They have recently shown to outperform both deterministic and alternative action distribution learning formulations. 3D robot policies use 3D scene feature representations aggregated from a single or multiple camera views using sensed depth. They have shown to generalize better than their 2D counterparts across camera viewpoints. We unify these two lines of work and present 3D Diffuser Actor, a neural policy equipped with a novel 3D denoising transformer that fuses information from the 3D visual scene, a language instruction and proprioception to predict the noise in noised 3D robot pose trajectories. 3D Diffuser Actor sets a new state-of-the-art on RLBench with an absolute performance gain of 18.1% over the current SOTA on a multi-view setup and an absolute gain of 13.1% on a single-view setup. On the CALVIN benchmark, it improves over the current SOTA by a 9% relative increase. It also learns to control a robot manipulator in the real world from a handful of demonstrations. Through thorough comparisons with the current SOTA policies and ablations of our model, we show 3D Diffuser Actor's design choices dramatically outperform 2D representations, regression and classification objectives, absolute attentions, and holistic non-tokenized 3D scene embeddings.

For agents

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

get_harvested_code_for_paper("2402.10885")
get_code_for_paper("2402.10885")
have("2402.10885")

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

curl https://syntology.ai/api/ran/2402.10885.json

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

[![Syntology run record](https://syntology.ai/api/ran/2402.10885.svg)](https://syntology.ai/paper/2402.10885)

Connect an agent — have() is free.