AI Paper Atlas

Research map / Reinforcement learning / Visual and representation RL

HIQL: Offline Goal-Conditioned RL with Latent States as Actions

Seohong Park, Dibya Ghosh, Benjamin Eysenbach, Sergey Levine

NeurIPS 2023 · Reinforcement learning › Visual and representation RL · approach: Graph search and compositional goals · 4 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
remove latent representation learning (ablation)HIQL · kitchen-partialnormalized return: 65.0 -> 46.3Hurt
remove latent representation learning (ablation)HIQL · kitchen-mixednormalized return: 67.7 -> 36.8Hurt
remove latent representation learning (ablation)HIQL · calvinnormalized return: 43.8 -> 23.4Hurt
remove latent representation learning (ablation)HIQL · antmaze-ultra-playnormalized return: 39.2 vs 56.0Mixed
remove latent representation learning (ablation)HIQL · antmaze-large-diversenormalized return: 88.2 vs 87.3No clear effect
replace flat policy with hierarchical policy extractionGC-IQL · antmaze-large-diversenormalized return: 50.7 -> 88.2Improved
replace flat policy with hierarchical policy extractionGC-IQL · procgen-maze-1000-testsuccess rate: 60.0 -> 78.2Improved
use temporal abstraction and representation instead of GC-POR hierarchyGC-POR · calvinnormalized return: 12.4 -> 43.8Improved
train with 25% action labels (rest action-free) (ablation)HIQL · antmaze-large-diversenormalized return: 88.2 -> 88.9No clear effect
train with 25% action labels (rest action-free) (ablation)HIQL · antmaze-ultra-diversenormalized return: 52.9 -> 38.2Hurt
train with 25% action labels (rest action-free) (ablation)HIQL · calvinnormalized return: 43.8 -> 35.8Hurt
train with 25% action labels (rest action-free) (ablation)HIQL · procgen-maze-500-testnormalized return: 64.5 -> 65.5No clear effect

Later papers that compared with this method

26 later accepted papers report HIQL as a baseline in their results tables; 20 of them report a better number on most of the results they share with it.

18 more not listed.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

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