SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2404.09445 · 2024

Exploring Text-to-Motion Generation with Human Preference

arXiv · PDF · Open in the Atlas

Code that ran

4 of 6 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; 4 failures of our instrument; 2 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/2404.09445.json · badge: /api/ran/2404.09445.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
thu-lyj-lab/instructmotion canonical (classifier label)
classifier basis: census
4 ran · 2 no recorded run
FunctionStatusWhere it lives
load_example_input Ran
ran; our contract draft was wrong, not the code
thu-lyj-lab/instructmotion/MotionGPT/generate_npy.py
code served (permissive licence) · get_code("c0eb0f4a912ff31f")
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. 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 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 unittest.mock import Mock
from torch import nn

torch.manual_seed(0)

# Create a minimal VAE mock
class MinimalVAE(nn.Module):
    def __init__(self):
        super().__init__()
        self.linear = nn.Linear(4, 263)
    
    def encode(self, feats):
        # feats shape: (batch, seq_len, 4)
        batch_size = feats.shape[0]
        # Return token tensor of shape (batch, token_len) and dummy loss
        token = torch.randint(0, 512, (batch_size, 10))
        return token, torch.tensor(0.0)

# Create a minimal DataModule mock
class MinimalDataModule:
    def normalize(self, feats):
        return feats / (feats.abs().max() + 1e-8)

# Create a minimal Model mock
class MinimalModel(nn.Module):
    def __init__(self):
        super().__init__()
        self.vae = MinimalVAE()
        self.datamodule = MinimalDataModule()
        self.device = torch.device('cpu')
        self.codebook_size = 512
    
    def feats2joints(self, feats):
        # feats shape: (seq_len, 4)
        # Return joints shape: (seq_len, 22, 3)
        seq_len = feats.shape[0]
        return torch.randn(seq_len, 22, 3)

# Create minimal model instance
model = MinimalModel()

# Create a temporary text file with test data
import tempfile
import os

temp_dir = tempfile.mkdtemp()
txt_path = os.path.join(temp_dir, "test_input.txt")

# Write test content to file
with open(txt_path, 'w') as f:
    f.write("file1.txt#This is a test text without motion\n")
    f.write("file2.txt#Another test text\n")

ARGS = {
    'txt_path': txt_path,
    'task': 'motion_generation',
    'model': model
}
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 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
load_pretrained_reward_rm Ran
ran; our contract draft was wrong, not the code
thu-lyj-lab/instructmotion/src/utils/motiongpt_utils.py
code served (permissive licence) · get_code("0797e250e681c777")
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 MinimalRewardModel. 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 collections import OrderedDict
import tempfile
import os

torch.manual_seed(0)

# Create a minimal reward model that matches the expected structure
class MinimalRewardModel(nn.Module):
    def __init__(self):
        super().__init__()
        self.transformer = nn.Module()
        self.transformer.shared = nn.Embedding(10, 4)
        self.transformer.encoder = nn.Linear(4, 4)
        self.transformer.decoder = nn.Linear(4, 4)
    
    def forward(self, x):
        return x

# Create a temporary checkpoint file with the expected state_dict structure
model = MinimalRewardModel()

# Build a state_dict that mimics what the function expects to load
state_dict_to_save = OrderedDict()
state_dict_to_save['lm.language_model.shared.weight'] = torch.randn(10, 4)
state_dict_to_save['lm.language_model.encoder.weight'] = torch.randn(4, 4)
state_dict_to_save['lm.language_model.encoder.bias'] = torch.randn(4)
state_dict_to_save['lm.language_model.decoder.weight'] = torch.randn(4, 4)
state_dict_to_save['lm.language_model.decoder.bias'] = torch.randn(4)

# Create temporary checkpoint
with tempfile.NamedTemporaryFile(mode='wb', delete=False, suffix='.pt') as f:
    ckpt_path = f.name
    torch.save({'state_dict': state_dict_to_save}, ckpt_path)

try:
    ARGS = {
        'model': MinimalRewardModel(),
        'ckpt_path': ckpt_path,
    }
finally:
    # Clean up will happen after function execution
    pass
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch, transformers — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type MinimalRewardModel — a MinimalRewardModel output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a MinimalRewardModel 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
motion_token_to_string Ran
ran; our contract draft was wrong, not the code
thu-lyj-lab/instructmotion/MotionGPT/generate_npy.py
code served (permissive licence) · get_code("65af45c09615b0ba")
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 torch

torch.manual_seed(0)

# Construct motion_token: a tensor of shape (batch_size, sequence_length)
# with integer values representing motion token IDs
motion_token = torch.randint(0, 512, (2, 4), dtype=torch.long)

# Construct lengths: a list indicating the actual length of each sequence
# (to handle variable-length sequences within the batch)
lengths = [3, 4]

# Construct codebook_size: the size of the codebook
codebook_size = 512

ARGS = {
    "motion_token": motion_token,
    "lengths": lengths,
    "codebook_size": codebook_size,
}
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
parse_args Ran
ran; our contract draft was wrong, not the code
thu-lyj-lab/instructmotion/MotionGPT/render.py
code served (permissive licence) · get_code("e80c033b31c865ac")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned Namespace. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import sys
from argparse import ArgumentParser, Namespace

class MockParser:
    """Stand-in ArgumentParser with parse_args_bak method"""
    def __init__(self):
        self.parser = ArgumentParser()
        self.parser.add_argument('--test', default='value')
    
    def parse_args_bak(self, args=None, namespace=None):
        """Mock implementation of parse_args_bak"""
        return self.parser.parse_args(args=args, namespace=namespace)

# Create instance
parser_instance = MockParser()

# Define arguments for parse_args
ARGS = {
    'self': parser_instance,
    'args': None,
    'namespace': None
}
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 Namespace — a Namespace output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a Namespace output the recorded ndim 0 and dtype family float describe that coercion, not the output, and are not stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
get_motion_tokenizer Not yet run
unverified (harvested, no recorded run)
thu-lyj-lab/instructmotion/src/utils/motiongpt_utils.py
code served (permissive licence) · get_code("59126a0d078805a0")
How it ran

no invocation recorded.

load_pretrained_reward_ppo Not yet run
unverified (harvested, no recorded run)
thu-lyj-lab/instructmotion/src/utils/motiongpt_utils.py
code served (permissive licence) · get_code("77ac2997bd266444")
How it ran

no invocation recorded.

Also present in other papers' code

Of 6 samples lifted from this paper's linked repositories, 6 carry a function identity and 0 do not. Of the 6 distinct functions with an identity, 1 are also present in other papers whose code we hold (1 with at least one paper dated earlier, 0 the same month, 0 later) and 5 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
parse_args 2 other papers
2 dated earlier
2306.14795 (2023-06)
MotionGPT: Human Motion as a Foreign Language
2311.01015 (2023-11)
Act As You Wish: Fine-Grained Control of Motion Diffusion Model with Hierarchical Semantic Graphs

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

This paper presents an exploration of preference learning in text-to-motion generation. We find that current improvements in text-to-motion generation still rely on datasets requiring expert labelers with motion capture systems. Instead, learning from human preference data does not require motion capture systems; a labeler with no expertise simply compares two generated motions. This is particularly efficient because evaluating the model's output is easier than gathering the motion that performs a desired task (e.g. backflip). To pioneer the exploration of this paradigm, we annotate 3,528 preference pairs generated by MotionGPT, marking the first effort to investigate various algorithms for learning from preference data. In particular, our exploration highlights important design choices when using preference data. Additionally, our experimental results show that preference learning has the potential to greatly improve current text-to-motion generative models. Our code and dataset are publicly available at https://github.com/THU-LYJ-Lab/InstructMotion}{https://github.com/THU-LYJ-Lab/InstructMotion to further facilitate research in this area.

For agents

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

get_harvested_code_for_paper("2404.09445")
get_code_for_paper("2404.09445")
have("2404.09445")

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

curl https://syntology.ai/api/ran/2404.09445.json

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

[![Syntology run record](https://syntology.ai/api/ran/2404.09445.svg)](https://syntology.ai/paper/2404.09445)

Connect an agent — have() is free.