Debate over LLM training temperature and blurry generated images
Two discussions emerged on LLM details: one argues training and sampling temperatures should match for consistency, while RL needs canonical temperature; another contends that model-generated 'ugly images' show MSE-style blur rather than natural ugliness, revealing a real generative deficiency.
2026-08-21 ~ 2026-08-21 · 2 related posts
- View: LLMs Should Be Trained at Sampling Temperature; RL Targets Canonical Temp — kalomaze · 2026-08-21
- Discussion: Why Model Artifacts Look Like MSE Blur, Not Natural Ugliness — kalomaze · 2026-08-21