AI Paper Atlas

Research map / Reinforcement learning / Visual and representation RL

Beyond Noisy-TVs: Noise-Robust Exploration Via Learning Progress Monitoring

Zhibo Hou, Zhiyu An, Wan Du

ICLR 2026 · Reinforcement learning › Visual and representation RL · approach: Novelty and impact-driven exploration

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
reward learning progress with an error model instead of prediction error or similarityLPM · Noisy MNISTsteps until intrinsic reward converges to zero: about 400 (AMA) -> about 150 (LPM)Improved
use LPM instead of similarity-based EDT bonusLPM · Noisy MNISTintrinsic reward on stochastic vs deterministic transitionsImproved
use LPM intrinsic reward for exploration with noisy visual inputLPM · MiniWorld 3D maze, average over three conditionsvisited states: +95.3 states over second best (7.6%)Improved
add action noise (random CIFAR image on idle action) to MiniWorldbaseline intrinsic-motivation methods · MiniWorld 3D mazestate coverageHurt
LPM in deterministic MountainCar versus EnsembleLPM · MountainCar-Continuous deterministicstate coverage (%): 91.22 -> 76.50Hurt
use LPM under stochastic observation noise in MountainCarLPM · MountainCar-Continuous stochasticstate coverage (%): 76.50 -> 67.04 (12.4% drop vs 33.1% to 85.9% for baselines)Improved
use LPM under action noise on AtariLPM · Atari with idle-action noiseextrinsic reward: 3.9% drop in Space Invader, 4.7% drop in UpNDown, 0.3% improvement in Ms PacManImproved
use LPM on clean and noisy Atari versus strong baselinesLPM · Atari (6 settings)extrinsic reward: best in 4 of 6Mixed
use LPM on Montezuma's RevengeLPM · Montezuma's Revengesteps to non-trivial score: RND 50M steps -> LPM 20M stepsImproved
use pointwise error difference instead of expected previous error from the error model (ablation)LPM · theory (Gaussian likelihood)monotonicity with information gainHurt
LPM on MiniGrid Lava Crossing versus ICMLPM · MiniGrid Lava CrossingPPO success rateNo clear effect

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy