AI Paper Atlas

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

Visual Semantic Planning Using Deep Successor Representations

Yuke Zhu, Daniel Gordon, Eric Kolve, Dieter Fox, Li Fei-Fei, Abhinav Gupta, Roozbeh Mottaghi, Ali Farhadi

ICCV 2017 · Multi-agent systems and games › Human-AI and goal-conditioned · approach: Embodied agent learning · 129 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 imitation-learning bootstrapping instead of RL from scratchSR model / A3C · THOR Medium taskssuccess rate: 0.00 -> 0.80Improved
add LSTM memory to the classifier baseline (ablation)CLS-MLP · THOR Medium taskssuccess rate: 0.65 -> 0.80Improved
add LSTM memory to the classifier baseline (ablation)CLS-MLP · THOR Hard tasksmean successful episode length: 475.86 -> 136.94Improved
add RL fine-tuning after imitation learning (ablation)SR IL · THOR Medium tasksmean successful episode length: 32.32 -> 26.56Improved
add RL fine-tuning after imitation learning (ablation)SR IL · THOR Easy tasksmean successful episode length: 2.70 -> 2.57Improved
add RL fine-tuning after imitation learning (ablation)SR IL · THOR Medium taskssuccess rate: 0.80 -> 0.80No clear effect
replace classifier baseline CLS-LSTM with SR ILCLS-LSTM · THOR Hard tasksmean successful episode length: 136.94 -> 34.25Improved
replace classifier baseline CLS-LSTM with SR ILCLS-LSTM · THOR Medium tasksmean successful episode length: 314.05 -> 32.32Improved
replace classifier baseline CLS-LSTM with SR ILCLS-LSTM · THOR Medium taskssuccess rate: 0.80 -> 0.80No clear effect
fine-tune only the reward vector w for new tasks instead of training from scratch (ablation)SR model · THOR hard tasks, 4 new itemsconvergence speedImproved
fine-tune the entire model instead of only w (ablation)SR model · THOR hard tasks, 4 new itemsconvergence speedHurt
train with imitation learning instead of RLA3C-style RL · THORfailed actionsImproved

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