AI Paper Atlas

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

Learning Dynamic Belief Graphs to Generalize on Text-Based Games

Ashutosh Adhikari, Xingdi Yuan, Marc-Alexandre Côté, Mikuláš Zelinka, Marc-Antoine Rondeau, Romain Laroche, Pascal Poupart, Jian Tang, Adam Trischler, Will Hamilton

NeurIPS 2020 · Multi-agent systems and games › Human-AI and goal-conditioned · approach: Language-grounded reinforcement agents · 52 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 text observations as input alongside the OG belief graph (ablation)GATA (OG) · TextWorld, avg across levelsrelative improvement (%): +8.0 -> +20.4Improved
add text observations as input alongside the COC belief graph (ablation)GATA (COC) · TextWorld, avg across levelsrelative improvement (%): +20.2 -> +24.2Improved
add text observations to ground-truth-pretrained discrete graph agent (ablation)GATA-GTP · TextWorld, avg across levelsrelative improvement (%): -1.2 -> +14.9Improved
add learned belief graph (COC pre-trained) to text-only Tr-DQNTr-DQN · TextWorld, avg across levelsrelative improvement (%): +20.2 (COC, graph only) vs Tr-DQN baselineImproved
use COC instead of OG self-supervised pre-training for graph updater (graph only) (ablation)GATA · TextWorld, avg across levelsrelative improvement (%): +8.0 -> +20.2Improved
use COC instead of OG pre-training (with text obs) (ablation)GATA · TextWorld, avg across levelsrelative improvement (%): +20.4 -> +24.2Improved
add recurrent memory (DRQN) to transformer DQNTr-DQN · TextWorld, avg across levelsrelative improvement (%): +3.9 avg; 20 games +10.3, 100 games -2.6Mixed
add episodic counting bonus to Tr-DRQNTr-DRQN · TextWorld, avg across levelsrelative improvement (%): +3.9 -> +3.6No clear effect
replace learned continuous belief graph with ground-truth-pretrained discrete graph updaterGATA · TextWorld, avg across levelsrelative improvement (%): +24.2 -> +14.9 (both with text obs)Hurt
give agent the ground-truth full graph instead of an inferred belief graphGATA · TextWorld, avg across levelsrelative improvement (%): +24.2 -> +81.6Improved
train on 100 instead of 20 gamesGATA (OG, graph only) · TextWorld, relative improvementrelative improvement (%): -0.2 -> +16.1Improved

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