Not all uncertainty is alike: volatility, stochasticity, and exploration

In three linesTheoretical work on adaptive exploration under uncertainty. Distinguishes volatility (reward drift) from stochasticity (observation noise): volatility enhances optimal exploration, stochasticity suppresses it. Introduces CAUSE, a closed-form exploration bonus via control-as-inference, validated on Gaussian state-space bandits with latent dynamics.

Summary generated by Claude — human-verified