Few-Step-Diffusion-Sampling-Quality-Risk-Model
RandomForest/GradientBoosting models predicting fast-generation quality from a slower preview alone. Forward-looking classifier: 100% accuracy.
About Dataset
Three scikit-learn models, GroupShuffleSplit by config_id. baseline_comparison.py shows the single-signal baseline (raw 20-step quality, no method identity) actually scores worse than majority-class guessing (66.7% vs 68.9%) a good preview genuinely doesn't guarantee 4-step success unless you know the method.
Purpose of Dataset
Lets A Hobbyist Running Diffusion/flow Models On Their Own Hardware Decide Their Interactive-generation Step Count Without Guessing.
Dataset Metadata
Attribution 4.0 International (CC BY- 4.0)
India
Science, Technology and Research
Jai Mali
Structured
NA
NA
N.A.
09/08/26 17:16:04
Open
Hosted
Synthetic
Activity Overview
0
0 
409.89 KB 
4
Tags
- Large Language Model
License Control
Attribution 4.0 International (CC BY- 4.0)
few-step-diffusion-sampling-risk-dataset.json ( 409.89 KB )
To preview this file, you need to be a registered user. Please complete the registration process to gain access and continue viewing the content.
Data Quality Score Beta
Version Control
Version 1(409.89 KB)
- JAI MALI·2 day(s) ago
few-step-diffusion-sampling-risk-dataset.json
© 2026 - Copyright AIKosh. All rights reserved.