AI Paper Atlas

Research map / Reinforcement learning / Visual and representation RL

ELLA: Exploration through Learned Language Abstraction

Suvir Mirchandani, Siddharth Karamcheti, Dorsa Sadigh

NeurIPS 2021 · Reinforcement learning › Visual and representation RL · approach: Novelty and impact-driven exploration · 8 citations (OpenAlex)

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
add relevant low-level GoTo rewardsPPO on PutNext-Room · PutNext-Roomsample efficiencyImproved
add relevant low-level GoTo rewardsPPO on PutNext-Maze · PutNext-Mazesample efficiencyImproved
add low-level Pick rewards for collecting relevant keysPPO on Unlock-Maze · Unlock-Mazesample efficiencyImproved
add relevant low-level rewardsPPO on Combo-Maze · Combo-Mazesample efficiencyNo clear effect
add low-level Open and GoTo rewardsPPO on Open&Pick-Maze · Open&Pick-Mazesample efficiencyImproved
add relevant low-level rewardsPPO on Sequence-Maze · Sequence-Mazesample efficiencyImproved
reduce termination-classifier training data from 15K to 2K (ablation)ELLA on Open&Pick-Maze · Open&Pick-Mazeaverage reward: 15K; 2KNo clear effect
reduce termination-classifier training data to 500 (ablation)ELLA on Open&Pick-Maze · Open&Pick-Mazeaverage reward: 500Hurt
reduce termination-classifier training data to 100 (ablation)ELLA on Open&Pick-Maze · Open&Pick-Mazetermination-classifier accuracy: 66%Hurt
reduce termination-classifier training data to 100 (ablation)ELLA on Open&Pick-Maze · Open&Pick-Mazelearning-curve quality: 100Hurt
replace online relevance classifier with an oracle after initially rewarding all low-level tasks (ablation)ELLA · Four evaluated tasksaverage reward: 2.5 million framesImproved
replace online relevance classifier with an oracle after initially rewarding all low-level tasks (ablation)ELLA · Sequence-Mazeaverage reward: 2.5 million framesNo 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