AI Paper Atlas

Research map / Multi-agent systems and games / Human-AI and goal-conditioned

Dungeons and Data: A Large-Scale NetHack Dataset

Eric Hambro, Roberta Raileanu, Danielle Rothermel, Vegard Mella, Tim Rocktäschel, Heinrich Kuttler, Naila Murray

NeurIPS 2022 · Multi-agent systems and games › Human-AI and goal-conditioned · approach: World-model-based agents · 0 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
Combine offline demonstrations with online APPO training using BCAPPO on NLD-AA · NLD-AA, Random characteraverage episode return: 868 ± 200 -> 1282 ± 275Improved
Combine offline demonstrations with online APPO training using BCAPPO on NLD-AA-Monk · NLD-AA-Monk, Human Monkaverage episode return: 2046 ± 360 -> 2809 ± 324Improved
Combine offline observations with online APPO training using BCOAPPO on NLD-NAO · NLD-NAO, Random characteraverage episode returnImproved
Train with behavioral cloning rather than offline RLOffline learning on NLD-AA · NLD-AA, Random characteraverage episode return: 33 ± 70; 160 ± 70; 554 ± 142Improved
Learn an inverse dynamics model from observation-only trajectories and use BCONLD-NAO · NLD-NAO, Random characteraverage episode returnImproved
Use APPO + BCO instead of APPO aloneOnline learning with observation-only demonstrations · NLD-NAO, Random charactertraining stabilityImproved
Use current offline RL and demonstration-learning methods to learn from NLD-AAOffline RL and imitation learning baselines · NLD-AAaverage episode returnHurt

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