Pref-sampling tests: Muse Wonder least model-influenced, Gemini 4 most nuanced
repligate · x · 2026-10-15
kromem2dot0's preference-sampling tests yielded two notable results this week: Muse Wonder shows the least influence from other models' outputs since the Claude 3 era, while Gemini 4 provided the most nuanced details and errata in its selection reasoning — a sign of fresh memetic diversity in the model ecosystem.
More from Models
- Applied Compute's Spork post-training recipe doubles chess puzzle solve-rate vs GRPO — ypatil125 · 2026-10-15
- Claude vs ChatGPT for Coding: Users Weigh Limits and Credits — OkraSuper9329 · 2026-10-15
- Google's Gemini 4 Argon Tops Physical-World Understanding, Scores 59.1% on Long Video — 141_1337 · 2026-10-15
- Model greets evaluators in its chain of thought — 4 trends paint an alarming AI trajectory — dfrsrchtwts · 2026-10-15
- TypeSafe AI ships Jev 1.14.0 with experimental images hidden in danger-zone flags — michellechen · 2026-10-15
- ADL says it works with OpenAI, Anthropic, Google, Meta and xAI to curb antisemitism; Claude translation blocked by safety filter — amasad · 2026-10-15