What question did this study set out to answer?

This work explores the maintenance of feedback in chaotic regimes without collapsing dynamics.

May 25, 2026Open Access

Self-referential feedback in chaotic reservoirs with preserved dynamics: a minimal architecture and the role of feedback type

View Full Paper

Ask AI

Bookmark

This work explores the maintenance of feedback in chaotic regimes without collapsing dynamics.
Developed a minimal architecture based on reservoir computing with two specialized reservoirs (chaos and memory).
Implemented a dual-head self-model with asymmetric compression and assessed prediction match metrics over ten random seeds.
Conducted systematic ablation to evaluate the effects of parametric and content feedback loops.
Content feedback loop increased prediction match from 0.33 to 0.66 (Cohen's d about 2.0) across 10 random seeds.
Lyapunov exponent remained stable and positive across modes (0.020 ± 0.002).
Parametric loop alone decreased prediction match, while content loop was both sufficient and crucial for maintaining feedback.

No takes yet. Share an insight, caveat, or question.

Randomized trial investigates feedback types in chaotic systems, suggesting implications for predictive coding.

Alexander Vasil’ev (2026) studied this question.

View Full Paper

Ask AI

Bookmark