Local prediction-error-based curiosity rewards focus on the current transition without considering the world model's cumulative prediction error across all visited transitions. We introduce Curiosity-Critic, which grounds its intrinsic reward in the improvement of this cumulative objective, and show that it reduces to a tractable per-step form: the difference between the current prediction error and the asymptotic error baseline of the current state transition. We estimate this baseline online with a learned critic co-trained alongside the world model; regressing a single scalar, the critic converges well before the world model saturates, redirecting exploration toward learnable transitions without oracle knowledge of the noise floor. The reward is higher for learnable transitions and collapses toward the baseline for stochastic ones, effectively separating epistemic (reducible) from aleatoric (irreducible) prediction error online. Prior prediction-error curiosity formulations, from Schmidhuber (1991) to learned-feature-space variants, emerge as special cases corresponding to specific approximations of this baseline. Experiments on a stochastic grid world show that Curiosity-Critic outperforms prediction-error and visitation-count baselines in convergence speed and final world model accuracy.

Scaling model capacity has been vital in the success of deep learning.

Introspection reveals that our meta learned LAs learn through fast association in a way that is qualitatively different from gradientdescent.

In this paper, we study the problem of unsupervised object segmentation from single images.

Neural Information Processing Systems http://nips.cc/

We introduce a visual challenge, Pathfinder, and describe a novel recurrent neural network architecture called the horizontal gated recurrent unit (hGRU) to learn intrinsic horizontal connections - both within and across feature columns.

The T = 100, itisclearthatT grows.SABstill tocompleteT = 5000, whereasT = 2000bothv self-attention 1/8 = 12.5%).

This sampling process is embodied as the policyπ, which is responsible for outputting an actiona conditioned on past environmental states{s}.

Adaptive parameterization is a means of increasing this flexibility and thereby increasing the model's capacity to learn non-linear patterns. We focus on the feed-forward layer, f(x):= φ(W x+b),for some activation functionφ: R 7 R. Define the pre-activation layer as a = A(x):= Wx+band denote byg(a):= φ(a)/athe activation effect ofφgivena, where divisioniselement-wise.