SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2505.13447 · 2025

Mean Flows for One-step Generative Modeling

arXiv · PDF · Open in the Atlas

Code that ran

We lifted 12 functions out of this paper's own repositories and ran 9 of them in a sandbox. "Ran" means the function executed on a synthesized input and returned a value. It is not a reproduction of the paper's results.

RepositoryRoleRan
Gsunshine/meanflow canonical 0 of 1
andypinxinliu/GestureLSM pwc_unofficial 9 of 11
FunctionStatusWhere it lives
exponential_pdf Ran andypinxinliu/GestureLSM/models/LSM.py
code served (permissive licence) · get_code("f3c44f47824302dc")
find_attention_modules Ran andypinxinliu/GestureLSM/models/MeanFlow.py
code served (permissive licence) · get_code("1b12f7f17b973eee")
get_loss_func Ran andypinxinliu/GestureLSM/optimizers/loss_factory.py
code served (permissive licence) · get_code("4f043bf1bec12652")
get_module_config Ran andypinxinliu/GestureLSM/models/config.py
code served (permissive licence) · get_code("2057a8b48b0d5cf3")
get_obj_from_str Ran andypinxinliu/GestureLSM/models/config.py
code served (permissive licence) · get_code("6f951d437f9f5c5f")
mean_flat Ran andypinxinliu/GestureLSM/models/MeanFlow.py
code served (permissive licence) · get_code("dd96bc6c07c85dae")
sample_beta_distribution Ran andypinxinliu/GestureLSM/models/LSM.py
code served (permissive licence) · get_code("1ba44760ad133e99")
sample_t Ran andypinxinliu/GestureLSM/models/LSM.py
code served (permissive licence) · get_code("f9b3244bb6e382cd")
update_lr_warm_up Ran andypinxinliu/GestureLSM/rvq_beatx_train.py
code served (permissive licence) · get_code("b3786bbc856619f6")
generate Not yet run Gsunshine/meanflow/meanflow.py
code served (permissive licence) · get_code("15cb9cb0de4aadad")
instantiate_from_config Not yet run andypinxinliu/GestureLSM/models/config.py
code served (permissive licence) · get_code("16e41676220876a4")
reshape_coefs Not yet run andypinxinliu/GestureLSM/models/MeanFlow.py
code served (permissive licence) · get_code("24486901843b5f5c")

Repositories linked to this paper

Some links come from the archived Papers with Code dataset (CC BY-SA 4.0): attribution and licence.

Abstract

We propose a principled and effective framework for one-step generative modeling. We introduce the notion of average velocity to characterize flow fields, in contrast to instantaneous velocity modeled by Flow Matching methods. A well-defined identity between average and instantaneous velocities is derived and used to guide neural network training. Our method, termed the MeanFlow model, is self-contained and requires no pre-training, distillation, or curriculum learning. MeanFlow demonstrates strong empirical performance: it achieves an FID of 3.43 with a single function evaluation (1-NFE) on ImageNet 256x256 trained from scratch, significantly outperforming previous state-of-the-art one-step diffusion/flow models. Our study substantially narrows the gap between one-step diffusion/flow models and their multi-step predecessors, and we hope it will motivate future research to revisit the foundations of these powerful models.

For agents

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

get_harvested_code_for_paper("2505.13447")
get_code_for_paper("2505.13447")
have("2505.13447")

Connect an agent — have() is free.