SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2402.07087 · ICML · 2024

Self-Correcting Self-Consuming Loops for Generative Model Training

Yonglong Tian, Calvin Luo, Chen Sun, Nate Gillman, Michael Freeman, Daksh Aggarwal, Chia-Hong Hsu

arXiv · PDF · Open in the Atlas

Code that ran

3 of 7 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; 1 honoured the contract we drafted, 0 violated it and 0 have no contract verdict; 2 failures of our instrument; 4 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/2402.07087.json · badge: /api/ran/2402.07087.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
nate-gillman/self-correcting-self-consuming canonical (classifier label)
classifier basis: census
3 ran · 4 no recorded run
FunctionStatusWhere it lives
approx_standard_normal_cdf Ran
ran; our contract draft was wrong, not the code
nate-gillman/self-correcting-self-consuming/diffusion/losses.py
pointer only (licence: MIT); the stored copy get_code serves, from quandao10/slct, is pointer only (licence: NONE) · get_code("cfd76fd0d89574a4")
How it ran

The body of this sample is withheld: this paper's link allows the body inline (MIT), but get_code serves the stored copy, first harvested from quandao10/slct, and GitHub detected no licence in that repository when it was checked. The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. A run is recorded.

discretized_gaussian_log_likelihood Ran
ran; our contract draft was wrong, not the code
nate-gillman/self-correcting-self-consuming/diffusion/losses.py
pointer only (licence: MIT); the stored copy get_code serves, from quandao10/slct, is pointer only (licence: NONE) · get_code("cd33283d615fb3d7")
How it ran

The body of this sample is withheld: this paper's link allows the body inline (MIT), but get_code serves the stored copy, first harvested from quandao10/slct, and GitHub detected no licence in that repository when it was checked. The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. A run is recorded.

normal_kl Ran
ran, honoured the contract we drafted
nate-gillman/self-correcting-self-consuming/diffusion/losses.py
pointer only (licence: MIT); the stored copy get_code serves, from facebookresearch/agrol, is pointer only (licence: NOASSERTION) · get_code("cf2798b666b231ca")
How it ran

The body of this sample is withheld: this paper's link allows the body inline (MIT), but get_code serves the stored copy, first harvested from facebookresearch/agrol, and GitHub could not classify the licence of that repository (NOASSERTION). The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. A run is recorded.

fix_height_smpl_vanilla Not yet run
unverified (harvested, no recorded run)
nate-gillman/self-correcting-self-consuming/UniversalHumanoidControl/amass_to_imitation.py
code served (permissive licence) · get_code("8c9f1d7fbffacb11")
How it ran

no invocation recorded.

get_imitation Not yet run
unverified (harvested, no recorded run)
nate-gillman/self-correcting-self-consuming/UniversalHumanoidControl/amass_to_imitation.py
code served (permissive licence) · get_code("e38ff66c5ff1a166")
How it ran

no invocation recorded.

process_qpos_list Not yet run
unverified (harvested, no recorded run)
nate-gillman/self-correcting-self-consuming/UniversalHumanoidControl/amass_to_imitation.py
code served (permissive licence) · get_code("34ce6e41f1d1c586")
How it ran

no invocation recorded.

visualize_samples_and_save_to_disk Not yet run
unverified (harvested, no recorded run)
nate-gillman/self-correcting-self-consuming/UniversalHumanoidControl/visualize_chains.py
code served (permissive licence) · get_code("1b96783243e1247f")
How it ran

no invocation recorded.

Also present in other papers' code

Of 7 samples lifted from this paper's linked repositories, 7 carry a function identity and 0 do not. Of the 7 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 4 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
normal_kl 160 other papers
81 dated earlier, 9 the same month, 68 later, 2 same year, order not determinable · 19 only as an assembly brick
2010.02502 (2020-10)
Denoising Diffusion Implicit Models
2105.05233 (2021-05)
Diffusion Models Beat GANs on Image Synthesis
2111.11672 (2021-11)
Few-shot Image Generation with Mixup-based Distance Learning
and 157 more not named here
approx_standard_normal_cdf 154 other papers
77 dated earlier, 8 the same month, 67 later, 2 same year, order not determinable · 17 only as an assembly brick
2010.02502 (2020-10)
Denoising Diffusion Implicit Models
2105.05233 (2021-05)
Diffusion Models Beat GANs on Image Synthesis
2111.11672 (2021-11)
Few-shot Image Generation with Mixup-based Distance Learning
and 151 more not named here
discretized_gaussian_log_likelihood 128 other papers
76 dated earlier, 5 the same month, 46 later, 1 same year, order not determinable · 16 only as an assembly brick
2010.02502 (2020-10)
Denoising Diffusion Implicit Models
2105.05233 (2021-05)
Diffusion Models Beat GANs on Image Synthesis
2111.11672 (2021-11)
Few-shot Image Generation with Mixup-based Distance Learning
and 125 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

As synthetic data becomes higher quality and proliferates on the internet, machine learning models are increasingly trained on a mix of humanand machine-generated data. Despite the successful stories of using synthetic data for representation learning, using synthetic data for generative model training creates "self-consuming loops" which may lead to training instability or even collapse, unless certain conditions are met. Our paper aims to stabilize self-consuming generative model training. Our theoretical results demonstrate that by introducing an idealized correction function, which maps a data point to be more likely under the true data distribution, selfconsuming loops can be made exponentially more stable. We then propose self-correction functions, which rely on expert knowledge (e.g. the laws of physics programmed in a simulator), and aim to approximate the idealized corrector automatically and at scale. We empirically validate the effectiveness of self-correcting self-consuming loops on the challenging human motion synthesis task, and observe that it successfully avoids model collapse, even when the ratio of synthetic data to real data is as high as 100%.

For agents

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

get_harvested_code_for_paper("2402.07087")
get_code_for_paper("2402.07087")
have("2402.07087")

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

curl https://syntology.ai/api/ran/2402.07087.json

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

[![Syntology run record](https://syntology.ai/api/ran/2402.07087.svg)](https://syntology.ai/paper/2402.07087)

Connect an agent — have() is free.