AI Paper Atlas

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

Temporal-Spatial Causal Interpretations for Vision-Based Reinforcement Learning

Wenjie Shi, Gao Ming Huang, Shiji Song, Cheng Wu

TPAMI 2021 · Multi-agent systems and games › Human-AI and goal-conditioned · approach: Language-grounded reinforcement agents · 21 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 TSCI-discovered features instead of SARFA features as the only pixels the policy seespolicy trained on masked pixels · Enduroaverage return: 2369 -> 2903Improved
use TSCI features instead of perturbation-based featurespolicy trained on masked pixels · Enduroaverage return: 1741 -> 2903Improved
use TSCI features instead of gradient-based featurespolicy trained on masked pixels · Enduroaverage return: 819 -> 2903Improved
use TSCI features instead of SARFA featurespolicy trained on masked pixels · Seaquestaverage return: 2085 -> 2517Improved
use TSCI features instead of perturbation-based featurespolicy trained on masked pixels · Seaquestaverage return: 1830 -> 2517Improved
use TSCI features instead of gradient-based featurespolicy trained on masked pixels · Seaquestaverage return: 836 -> 2517Improved
use temporal causal objective instead of CXPlain or imitation objectives (ablation)TSCI causal discovery network · Atari behavior-matching reliabilitybehavior matching errors and temporal causality errorImproved
use temporal causal objective instead of CXPlain or imitation objectives (ablation)TSCI causal discovery network · Atarimean normalized returnNo clear effect
mask semantic information in more of the stacked frames (4 -> 34 -> 234 -> 1234) (ablation)PPO recurrent agent · Atari counterfactual interventionaverage returnHurt
mask semantic information in a single previous frame (schemes 1, 2, 3) (ablation)PPO recurrent agent · Atari counterfactual interventionaverage returnNo clear effect
stack only the current frame (no frame stacking) in a recurrent agent (ablation)recurrent PPO agent · Atariaverage returnHurt
stack more consecutive frames in recurrent agent input (ablation)recurrent PPO agent · Atariaverage returnImproved

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