Subculture Science Research Group's Blog
  • Blog
  • Portfolio
  • About us
  • Tags
Subculture Science Research Group's Blog

Reinforcement Learning


The Small LLM That Answers Before It Thinks: Four Roles and a Comparison Design

 Posted on Sun Nov 15 2026  |  6 min  |  1120 words  |    • Other languages: 日本語

Small LLMs commit to a hunch and stop tracking candidates. This post breaks down the failure modes, describes a loop that grows a creator and a solver in turn, and reports how far a per-turn reward actually moved the numbers. [Read More]
LLM  LoRA  reinforcement learning  evaluation  problem solving 

Co-evolving Four Roles, GAN-Style: The Testbed Behind umigame_GAN

 Posted on Sun Nov 15 2026  |  7 min  |  1322 words  |    • Other languages: 日本語

A testbed for strengthening trial-and-error reasoning in small LLMs: four competing roles, a domain-independent loop, R-Zero and EDA style rewards, and an MLX LoRA comparison design. Code excerpts and measured numbers from the repository. [Read More]
LLM  LoRA  MLX  reinforcement learning  GAN  engineering 

 • © 2026  •  Subculture Science Research Group's Blog

Hugo v0.142.0 powered  •  Theme Beautiful Hugo adapted from Beautiful Jekyll

  • 日本語
  • English