The QSG Model: Persona Plasticity Controls Agent Belief Adaptation
Hidenori8Tanaka · x · 2026-09-09
Part 4 of Tanaka's thread: the Quantized Simplex Gossip (QSG) model represents AI output as sampling from internal belief distributions — xL′ = (1−α)xL + α·Samplem(xS), where Samplem returns relative frequencies of m independent symbol samples, and persona plasticity α controls the listener's adaptation.
See the main thread post for full context.
More from Research
- Fixing compute convention and fitted floor cuts scaling forecast error ~79% — dlwh · 2026-10-27
- Retrospective: Hero scaling forecast fixed without using Hero losses, code open-sourced — dlwh · 2026-10-27
- Scaling-fit discrepancy: output-head FLOPs are ~65% of compute at the smallest rung, ~3% at 535B — dlwh · 2026-10-27
- Marin's dlwh Explains the Scaling Law Fit: Estimating the Loss Floor Early — dlwh · 2026-10-27
- Marin's Open Scaling Law Extrapolates 300× and Predicts Loss Within 0.2% — dlwh · 2026-10-27
- Delegating scaling-law asymptote fitting to Codex while planning a birthday party — dlwh · 2026-10-27