AI Paper Atlas

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

Building spatial world models from sparse transitional episodic memories

Zizhan He, Maxime Daigle, Pouya Bashivan

ICLR 2026 · Multi-agent systems and games › Human-AI and goal-conditioned · approach: Intrinsically motivated exploration · 1 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
use a transformer instead of LSTM or Mamba for the world model (ablation)ESWM · Open Arenageneralization accuracyImproved
use ESWM with disjoint memories instead of TEM-T sequence modelESWM-T · Open ArenaaccuracyImproved
increase model sizeESWM-T vs TEM-T · Open ArenaaccuracyMixed
increase number of attention heads beyond 8 (ablation)ESWM-T · Open ArenaaccuracyHurt
increase model depth (ablation)ESWM-T · Open ArenaaccuracyImproved
use denser non-minimal memory banksESWM · Random Wallprediction accuracyImproved
use ESWM world model with planning instead of EPNESWM · Random Wall navigationsuccess rate / optimality: 96.8% success, 99.2% optimality, +18% and +21% over EPNImproved
use ESWM explorer instead of EPNESWM · Random Wall explorationunique states visited: 16.8% more unique statesImproved
update memory bank instead of retraining after structural changesESWM · Random Wall with added wallnavigation success: 93 vs 72 (EPN) and 56 (baseline)Improved
add contradictory observations to memory banks (noise)ESWM · Random Wallprediction accuracy: dropped by 1% and 10%Hurt
train on episodic transitions to build latent map (trained vs untrained)ESWM · sRSAsRSA: 0.5 -> 0.78Improved
hand-engineered memory lookup instead of learnable attention (ablation)hand-engineered lookup · Open ArenaaccuracyHurt

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