SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2308.08855 · ICCV · 2023

Realistic Full-Body Tracking from Sparse Observations via Joint-Level Modeling

Xiaojie Jin, Xue Zhou, Zhuo Su, Wen Chao, Xiaozheng Zheng

arXiv · PDF · Open in the Atlas

Code that ran

14 of 14 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 0 have no contract verdict; 14 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/2308.08855.json · badge: /api/ran/2308.08855.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
zxz267/AvatarJLM canonical (classifier label)
classifier basis: census
14 ran · 0 no recorded run
FunctionStatusWhere it lives
bgs Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
zxz267/AvatarJLM/utils/utils_transform.py
code served (permissive licence) · get_code("b813a8c681ebd63f")
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 (2, 3, 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 torch.nn.functional as F

torch.manual_seed(0)

# The function bgs expects d6s to be reshaped to (-1, 2, 3)
# So the input should have 6 elements per batch item (2*3=6)
# Let's create a batch of 2 items with 6 values each
d6s = torch.randn(2, 6)

ARGS = {
    "d6s": d6s
}
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 (2, 3, 3) — shape (2, 3, 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
define_Model Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
zxz267/AvatarJLM/models/select_model.py
code served (permissive licence) · get_code("8aed4f90ba7e05fb")
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 ModelAvatarJLM. 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

torch.manual_seed(0)

# Create a minimal mock model class that satisfies ModelAvatarJLM interface
class ModelAvatarJLM(nn.Module):
    def __init__(self, opt):
        super().__init__()
        self.opt = opt
        # Store minimal required attributes based on function usage
        self.model_name = opt.get('model', 'avatarJLM')
        # Add a simple linear layer to make this a valid nn.Module
        self.fc = nn.Linear(4, 3)
    
    def forward(self, x):
        return self.fc(x)

# Patch the import path so our mock is used
import sys
import types
mock_model_module = types.ModuleType('models.model_jlm')
mock_model_module.ModelAvatarJLM = ModelAvatarJLM
sys.modules['models.model_jlm'] = mock_model_module
sys.modules['models'] = types.ModuleType('models')
sys.modules['models'].model_jlm = mock_model_module

# Define the opt dict with all required configuration
opt = {
    'model': 'avatarJLM',
    'input_dim': 4,
    'output_dim': 3,
    'hidden_dim': 8,
}

ARGS = {
    'opt': opt,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
models, torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type ModelAvatarJLM — a ModelAvatarJLM output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a ModelAvatarJLM 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
dict2str Ran
ran; our contract draft was wrong, not the code
zxz267/AvatarJLM/utils/utils_option.py
code served (permissive licence) · get_code("16c3e584beb5ce07")
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 str. 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 numpy as np

torch.manual_seed(0)

# Analyze the function:
# - dict2str takes 'opt' (a dict) and 'indent_l' (an int, default 1)
# - It recursively processes the dict, handling nested dicts and other values
# - It converts values to strings and formats them with indentation

# Create a test dictionary with nested structure
opt = {
    'learning_rate': 0.001,
    'batch_size': 32,
    'model': {
        'hidden_dim': 128,
        'num_layers': 3,
        'dropout': 0.5,
        'architecture': {
            'type': 'transformer',
            'heads': 8
        }
    },
    'epochs': 100,
    'optimizer': 'adam'
}

ARGS = {
    'opt': opt,
    'indent_l': 1
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1
Imports
numpy, torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type str — a str output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a str 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
find_last_checkpoint Ran
ran; our contract draft was wrong, not the code
zxz267/AvatarJLM/utils/utils_option.py
code served (permissive licence) · get_code("7f17c0641cd21670")
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 os
import re
import glob
import tempfile
import shutil

# Create a temporary directory for testing
temp_dir = tempfile.mkdtemp()

# Create some dummy checkpoint files to test the function
checkpoint_files = [
    os.path.join(temp_dir, '100_G.pth'),
    os.path.join(temp_dir, '200_G.pth'),
    os.path.join(temp_dir, '50_G.pth'),
]

for checkpoint_file in checkpoint_files:
    with open(checkpoint_file, 'w') as f:
        f.write('dummy checkpoint')

ARGS = {
    'save_dir': temp_dir,
    'net_type': 'G',
}
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 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
floating_error Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
zxz267/AvatarJLM/utils/utils_metric.py
code served (permissive licence) · get_code("0fd68a97e0fd9f90")
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 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

torch.manual_seed(0)

# Construct pred_mesh: (batch, v_num, 3)
# batch=2, v_num=4, coordinates=3
batch_size = 2
v_num = 4
pred_mesh = torch.randn(batch_size, v_num, 3)

# Construct floor_height: (batch,)
# Should have shape (batch_size,) to match the logic in the function
floor_height = torch.randn(batch_size)

ARGS = {
    "pred_mesh": pred_mesh,
    "floor_height": floor_height,
}
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 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
footContactLoss Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
zxz267/AvatarJLM/models/loss.py
code served (permissive licence) · get_code("0d68cdd2c818c959")
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 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 as nn

torch.manual_seed(0)

# Create a minimal loss function that accepts two tensors
loss_func = nn.MSELoss()

# Create pred and gt tensors
# The function expects shape (batch, seq_len, features)
# It extracts the first 22*3=66 elements and reshapes to (batch, seq_len, 22, 3)
# So we need at least 66 features in the last dimension
batch = 2
seq_len = 5
num_joints = 22
coord_dim = 3
total_features = num_joints * coord_dim  # 66

# pred and gt should have shape (batch, seq_len, total_features)
pred = torch.randn(batch, seq_len, total_features)
gt = torch.randn(batch, seq_len, total_features)

ARGS = {
    "loss_func": loss_func,
    "pred": pred,
    "gt": gt,
}
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 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
local2global_pose Ran
ran; our contract draft was wrong, not the code
zxz267/AvatarJLM/models/module.py
code served (permissive licence) · get_code("42e1b193cf19125f")
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 4, dtype family float, shape (2, 22, 3, 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)

# local_pose: shape (batch_size, 22*3*3) -> reshaped to (batch_size, 22, 3, 3)
# This represents local rotation matrices for 22 joints
local_pose = torch.randn(2, 22, 3, 3)

# kintree: array of parent joint indices for each of 22 joints
# -1 means root (no parent), otherwise 0-21 index of parent joint
# Standard SMPL kinematic tree structure
kintree = np.array([
    -1,  # 0: root has no parent
    0,   # 1: parent is 0
    0,   # 2: parent is 0
    0,   # 3: parent is 0
    1,   # 4: parent is 1
    2,   # 5: parent is 2
    3,   # 6: parent is 3
    4,   # 7: parent is 4
    5,   # 8: parent is 5
    6,   # 9: parent is 6
    7,   # 10: parent is 7
    8,   # 11: parent is 8
    9,   # 12: parent is 9
    9,   # 13: parent is 9
    9,   # 14: parent is 9
    12,  # 15: parent is 12
    13,  # 16: parent is 13
    14,  # 17: parent is 14
    16,  # 18: parent is 16
    17,  # 19: parent is 17
    18,  # 20: parent is 18
    19,  # 21: parent is 19
])

ARGS = {
    "local_pose": local_pose,
    "kintree": kintree
}
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 4, dtype family float, shape (2, 22, 3, 3) — shape (2, 22, 3, 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
matrot2sixd Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
zxz267/AvatarJLM/utils/utils_transform.py
code served (permissive licence) · get_code("0cc7759dd018622e")
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, 6) 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 Nx3x3 tensor for pose_matrot
# N=2 (batch size), 3x3 (rotation matrix)
pose_matrot = torch.randn(2, 3, 3)

ARGS = {
    "pose_matrot": pose_matrot
}
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, 6) — shape (2, 6) 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 Ran
ran; our contract draft was wrong, not the code
zxz267/AvatarJLM/utils/utils_option.py
code served (permissive licence) · get_code("630044843d91b1bb")
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 OrderedDict. 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 os
import json
import tempfile
from collections import OrderedDict

# Create a temporary JSON config file
config_data = {
    "task": "test_task",
    "scale": 1,
    "gpu_ids": [0],
    "merge_bn": False,
    "merge_bn_startpoint": -1,
    "find_unused_parameters": True,
    "dist": False,
    "path": {
        "root": "./experiments",
        "pretrained_netG": "./models/pretrained.pth",
        "pretrained": "./models/pretrained.pth"
    },
    "datasets": {
        "train": {
            "dataset_type": "amass_p1",
            "dataroot": "./data/protocol_1",
            "dataroot_H": None,
            "dataroot_L": None
        },
        "test": {
            "dataset_type": "amass_p1",
            "dataroot": "./data/protocol_1",
            "dataroot_H": None,
            "dataroot_L": None
        }
    },
    "netG": {
        "net_type": "generator",
        "in_nc": 3,
        "out_nc": 3,
        "nf": 64
    },
    "netD": {
        "net_type": "discriminator_patchgan",
        "in_nc": 3,
        "base_nc": 64,
        "n_layers": 3,
        "norm_type": "spectral"
    },
    "train": {
        "F_feature_layer": 34,
        "F_weights": 1.0,
        "F_lossfn_type": "l1",
        "F_use_input_norm": True,
        "F_use_range_norm": False,
        "G_optimizer_reuse": False,
        "D_optimizer_reuse": False,
        "G_param_strict": True,
        "D_param_strict": True,
        "E_param_strict": True,
        "E_decay": 0
    }
}

# Write config to temporary file
temp_fd, temp_path = tempfile.mkstemp(suffix='.json', text=True)
os.write(temp_fd, json.dumps(config_data).encode('utf-8'))
os.close(temp_fd)

# Create a simple args class
class SimpleArgs:
    def __init__(self):
        self.task = "test_task"
        self.protocol = None
        self.checkpoint = None
    
    @property
    def __dict__(self):
        return {
            'task': self.task,
            'protocol': self.protocol,
            'checkpoint': self.checkpoint
        }

args_obj = SimpleArgs()

ARGS = {
    'opt_path': temp_path,
    'args': args_obj,
    'is_train': True
}
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 OrderedDict — a OrderedDict output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a OrderedDict 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
penetrationLoss Ran
ran; our contract draft was wrong, not the code
zxz267/AvatarJLM/models/loss.py
code served (permissive licence) · get_code("11384aef90bf1e42")
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 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

torch.manual_seed(0)

# Analyze the function:
# - pred_mesh: (batch, v_num, 3) tensor - batch of meshes with vertices
# - floor_height: (batch_floors,) tensor - floor heights
# - seq_len = pred_mesh.shape[0] // floor_height.shape[0]
# - The function computes loss based on mesh penetration below floor

# For valid execution:
# - pred_mesh.shape[0] must be divisible by floor_height.shape[0]
# - pred_mesh needs shape (batch, v_num, 3)
# - floor_height needs shape (batch_floors,)

# Let's set up:
# - batch = 4 (pred_mesh batch dimension)
# - batch_floors = 2 (floor_height batch dimension)
# - seq_len = 4 // 2 = 2
# - v_num = 5 (number of vertices)

batch = 4
batch_floors = 2
v_num = 5

# Create pred_mesh with shape (4, 5, 3)
# Contains z-coordinates that may be below floor
pred_mesh = torch.randn(batch, v_num, 3)

# Create floor_height with shape (2,)
# These will be repeated to match pred_mesh batch size
floor_height = torch.tensor([0.5, 1.0])

ARGS = {
    "pred_mesh": pred_mesh,
    "floor_height": floor_height,
}
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 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
penetration_error Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
zxz267/AvatarJLM/utils/utils_metric.py
code served (permissive licence) · get_code("f60dcc669fe7bec8")
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 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

torch.manual_seed(0)

# pred_mesh: (batch, v_num, 3)
# The function calls .min(1) on pred_mesh, expecting shape (batch, v_num, 3)
# It then indexes with [:, 2] on the result, expecting (batch, 3)
# So we need batch samples, each with v_num vertices in 3D space
batch_size = 2
v_num = 4
pred_mesh = torch.randn(batch_size, v_num, 3)

# floor_height: used with .float() and .shape[0]
# It's compared with lowest_z which has shape (batch,)
# Then floor_height[:, None].repeat(1, seq_len).view(-1) is called
# This means floor_height should have shape (num_sequences,) where num_sequences divides batch_size
# seq_len = batch_size // floor_height.shape[0]
# For batch_size=2, we can have floor_height.shape[0]=1 or 2
# Let's use shape (1,) so seq_len = 2
floor_height = torch.randn(1)

ARGS = {
    "pred_mesh": pred_mesh,
    "floor_height": floor_height,
}
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 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
skating_error Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
zxz267/AvatarJLM/utils/utils_metric.py
code served (permissive licence) · get_code("e2cfd0597b21ea18")
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 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

torch.manual_seed(0)

# Construct valid arguments for skating_error
# The function expects:
# - pred: (batch, v_num, 3) where it will reshape to (batch, seq_len, -1)
# - gt: same shape as pred

# Looking at the function logic:
# - seq_len = pred.shape[0]
# - batch = pred.shape[0] // seq_len
# This implies pred.shape[0] must be divisible by itself, so batch = 1
# OR the logic is: total_frames = pred.shape[0], and batch is derived from that

# Actually, re-reading: seq_len = pred.shape[0], then batch = pred.shape[0] // seq_len = 1
# So we have shape (total_size, v_num, 3) where total_size gets divided into (batch=1, seq_len=total_size)

# Then it reshapes to (batch, seq_len, -1) and expects to extract 22*3 values
# So v_num must be at least 22

# Let's construct:
# - Total frames: 4 (seq_len)
# - Batch: 1 (derived as 4 // 4)
# - Vertices: 22 (minimum needed)
# Shape: (4, 22, 3)

batch_size = 1
seq_len = 4
num_vertices = 22

pred = torch.randn(seq_len, num_vertices, 3)
gt = torch.randn(seq_len, num_vertices, 3)

ARGS = {
    "pred": pred,
    "gt": gt,
}
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 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
trunc_normal_ Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
zxz267/AvatarJLM/models/module.py
code served (permissive licence) · get_code("02566da69866c48c")
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 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 math

torch.manual_seed(0)

# Create a small tensor to be filled with truncated normal values
tensor = torch.empty(2, 4)

# Define arguments for trunc_normal_
ARGS = {
    'tensor': tensor,
    'mean': 0.0,
    'std': 1.0,
    'a': -2.0,
    'b': 2.0
}
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 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
velocityLoss Ran
ran; our contract draft was wrong, not the code
zxz267/AvatarJLM/models/loss.py
code served (permissive licence) · get_code("b07eed3dde8d7b16")
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 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 as nn

torch.manual_seed(0)

# Create a simple loss function (MSELoss)
loss_func = nn.MSELoss()

# Create pred tensor: (batch=2, seq_len=5, joint_num=22, 3)
# joint_num=22 because the function uses :22*3 slicing
pred = torch.randn(2, 5, 22, 3)

# Create gt tensor: (batch=2, seq_len=5, joint_num=22, 3)
gt = torch.randn(2, 5, 22, 3)

# interval parameter
interval = 1

ARGS = {
    "loss_func": loss_func,
    "pred": pred,
    "gt": gt,
    "interval": interval
}
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 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 14 samples lifted from this paper's linked repositories, 14 carry a function identity and 0 do not. Of the 14 distinct functions with an identity, 3 are also present in other papers whose code we hold (3 with at least one paper dated earlier, 3 the same month, 3 later) and 11 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
trunc_normal_ 160 other papers
66 dated earlier, 6 the same month, 86 later, 2 same year, order not determinable · 36 only as an assembly brick
1911.10194 (2019-11)
Panoptic-DeepLab: A Simple, Strong, and Fast Baseline for Bottom-Up Panoptic Segmentation
2005.10242 (2020-05)
Understanding Contrastive Representation Learning through Alignment and Uniformity on the Hypersphere
2007.12911 (2020-07)
Tighter risk certificates for neural networks
and 157 more not named here
dict2str 23 other papers
11 dated earlier, 1 the same month, 11 later
1905.02716 (2019-05)
EDVR: Video Restoration with Enhanced Deformable Convolutional Networks
1907.03222 (2019-07)
IRNet: A General Purpose Deep Residual Regression Framework for Materials Discovery
2005.05650 (2020-05)
Invertible Image Rescaling
and 20 more not named here
find_last_checkpoint 5 other papers
3 dated earlier, 1 the same month, 1 later
2103.14006 (2021-03)
Designing a Practical Degradation Model for Deep Blind Image Super-Resolution
2106.15368 (2021-06)
Text Prior Guided Scene Text Image Super-resolution
2207.13784 (2022-07)
AvatarPoser: Articulated Full-Body Pose Tracking from Sparse Motion Sensing
and 2 more not named here

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

To bridge the physical and virtual worlds for rapidly developed VR/AR applications, the ability to realistically drive 3D full-body avatars is of great significance. Although real-time body tracking with only the head-mounted displays (HMDs) and hand controllers is heavily underconstrained, a carefully designed end-to-end neural network is of great potential to solve the problem by learning from large-scale motion data. To this end, we propose a two-stage framework that can obtain accurate and smooth full-body motions with the three tracking signals of head and hands only. Our framework explicitly models the joint-level features in the first stage and utilizes them as spatiotemporal tokens for alternating spatial and temporal transformer blocks to capture joint-level correlations in the second stage. Furthermore, we design a set of loss terms to constrain the task of a high degree of freedom, such that we can exploit the potential of our joint-level modeling. With extensive experiments on the AMASS motion dataset and real-captured data, we validate the effectiveness of our designs and show our proposed method can achieve more accurate and smooth motion compared to existing approaches.

For agents

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

get_harvested_code_for_paper("2308.08855")
get_code_for_paper("2308.08855")
have("2308.08855")

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

curl https://syntology.ai/api/ran/2308.08855.json

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

[![Syntology run record](https://syntology.ai/api/ran/2308.08855.svg)](https://syntology.ai/paper/2308.08855)

Connect an agent — have() is free.