⌁ Awesome World Model Hub Papers · Datasets · Projects
中文 GitHub
← Back to papers

From Observations to Events: Event-Aware World Model for Reinforcement Learning

ICLR 26 2026 76 method

TLDR

Proposes Event-Aware World Model (EAWM) that learns event representations from raw observations to improve MBRL generalization and robustness.

Reasoning

Strengths: Novel cognitive-inspired event segmentation approach, strong empirical gains (10-45%) across multiple benchmarks, and unified formulation of world model architectures. Weaknesses: Abstract lacks details on the event generator and GES mechanisms; generalization claims may depend on specific benchmarks.

Read-first score

Read-first score 76, weighted from topical fit, citation, graph, method, reproducibility, and recency signals. Original total remains 46.

Recency 8%
100

Uses a gentle age decay so recent papers surface without erasing older foundations. 2026

Reproducibility 25%
81

Screens links and visible text for paper, code, dataset, artifact, and repository signals. pdf=True; code=True; dataset=False; markers=code,github

Methodology quality 25%
80

Screens visible abstract and analysis fields for experiment, dataset, baseline, metric, and limitation evidence. markers=baseline,benchmark,experiment,result

Topical relevance 42%
65.7

Uses existing LLM keyword relevance scores normalized to 0-100. world model,world simulator,generative world model,interactive world model,video world model,world dynamics prediction,model-based reinforcement learning world model

Field roles

FrontierMethodology anchorReproducibility anchor

Rank sensitivity

Stability: volatile; rank range: 90.

Keyword Scores

world model
10
model-based reinforcement learning world model
10
world dynamics prediction
8
generative world model
6
interactive world model
5
video world model
4
world simulator
3

Deep Analysis

Innovations

  • Event-Aware World Model (EAWM) framework that learns event-aware representations without handcrafted labels
  • Automated event generator to derive events from raw observations
  • Generic Event Segmentor (GES) to identify event boundaries marking start and end of event segments
  • Unified formulation of seemingly distinct world model architectures demonstrating broad applicability

Methodology

EAWM employs an automated event generator to derive events from raw observations and a Generic Event Segmentor (GES) to identify event boundaries. Through event prediction, the representation space is shaped to capture meaningful spatio-temporal transitions, streamlining policy learning in a model-based reinforcement learning setting.

Key Results

EAWM consistently boosts the performance of strong MBRL baselines by 10%-45% on Atari 100K, Craftax 1M, DeepMind Control 500K, and DMC-GB2 500K, setting new state-of-the-art results across these benchmarks.

Resources

PaperCode

Tags