MSR separates generation into three stages.
Corpus-guided sampling makes generative outputs more diverse and representative
Mine-Sample-Render extracts attributes from reference data, samples their joint distribution, then conditions existing models on the resulting configurations.
Big Tech
Kostas Triaridis · Morris E. Sharp · Nebojsa Jojic
Stony Brook University · Microsoft
Research Digest··2 min read
Triaridis, Sharp and Jojic present Mine-Sample-Render (MSR), a training-free method for aligning the diversity of generated outputs with a reference corpus.
Why this paper
From Microsoft and Stony Brook University
In one line
Generative models over-concentrate samples; Mine-Sample-Render samples interpretable attribute configurations from a real-data corpus and uses the model as a conditional renderer.
What we could check
- ·No code link found
- ·No weights link found
- ·No dataset link found
- ·No compute details found
- ✓Limitations stated by the authors (2 noted)
- ·No benchmark numbers found
Observed from the paper text and links we have. Absence here means we did not find it, not that it does not exist.
§