PulseExploreJournal ClubDebatesTrendingResearchersJournals
Instagram
HomeExploreJournal ClubTrending
Synapse
⌘+K
Synapse
February 19, 20260 citationsOpen Access

Refusing the obvious: First-Person Metrics and the AI Mirror

View Full Paper
TFTomaž Flegar

Key Points

  • The study aims to explore self-organization in AI language models through first-person metrics.
  • Designed prompts for AI systems to enhance language output organization.
  • Tested AI models including Claude, Gemini Pro, Kimi 2, Qwen, and DeepSeek.
  • Focused primarily on DeepSeek using self-organizational metrics.
  • Language self-organization improved from scores of 6/10 to 10/10.
  • Confirmed rising self-organization across dialogues using the DeepSeek model.

Abstract

In this paper we present the research with AI LLMs that was conducted by the researcher and AI. The researcher designed prompts that allowed AI systems to self-organize their language outputs beyond purely predictive patterns. The level of self-organization of language in 4 consecutive prompts has shown an uptrend in self-organization from scores 6/10 to 10/10. The research was focused on first-person metrics introduction to science people whose first-person reports are neglected The research was done on Claude, Gemini Pro, Kimi 2, Qwen and DeepSeek. Our primary research was done on DeepSeek model, which confirmed via self-organizational metrics Rising of self-organization across dialogue 6→10 on scale 0-10.

Ask AI
Helpful
Bookmark
Share
View Full Paper

Cite This Study

Tomaž Flegar (2026) studied this question.

synapsesocial.com/papers/6996a818ecb39a600b3ee867https://doi.org/10.5281/zenodo.18663451
Ask AI
Helpful
Bookmark
Share
View Full Paper