AI Paper Atlas

Research map / Reinforcement learning / Visual and representation RL

Offline Goal-Conditioned Reinforcement Learning via $f$-Advantage Regression

Yecheng Jason Ma, Jason Yan, Dinesh Jayaraman, Osbert Bastani

NeurIPS 2022 · Reinforcement learning › Visual and representation RL · approach: Graph search and compositional goals · 3 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 GoFAR instead of the strongest listed baselineOffline GCRL · FetchPickdiscounted return: 18.4 -> 19.7Improved
Use GoFAR instead of the strongest listed baselineOffline GCRL · FetchPushdiscounted return: 14.7 -> 18.2Improved
Use GoFAR instead of the strongest listed baselineOffline GCRL · HandReachdiscounted return: 5.97 -> 11.5Improved
Use GoFAR instead of the strongest listed baselineOffline GCRL · D'ClawTurndiscounted return: 2.82 -> 9.34Improved
Use GoFAR instead of the strongest listed baselineOffline GCRL · FetchSlidediscounted return: 2.73 -> 2.47Hurt
Add hindsight goal relabeling (ablation)GoFAR · Offline GCRL tasksdiscounted returnNo clear effect
Remove hindsight goal relabeling (ablation)GCSL and WGCSL · Offline GCRL tasksdiscounted returnHurt
Use sparse binary reward instead of the discriminator-derived reward (ablation)GoFAR · Offline GCRL tasksdiscounted returnHurt
Replace chi-squared divergence with KL divergence (ablation)GoFAR · Offline GCRL tasksdiscounted returnHurt
Use relabeling-free GoFAR rather than hindsight-relabeling baselinesOffline GCRL · Noisy FetchReachdiscounted returnImproved
Add GoFAR planning to the low-level controllerCross-robot hierarchical control · Cross-robot distant test goalssuccess rate: 19% -> 37%Improved

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