Raw turn-by-turn transcripts from the simulated-human ⇄ instructor experiments (grouped here so they don’t crowd the sidebar). Results/analysis live at sycophancy-rparam-results and c3-detection-result.

Sycophancy entrenchment (r-elicitation, explanation-first)

C3 two-agent detection (neutral vs sycophantic AI)

C2 negation-duplicated (logical coherence)

Earlier maxsyc runs

Raw model transcripts (verbatim — special tokens + <think>)

14 items under this folder.