AI Paper Atlas

Research map / Reinforcement learning / Offline RL

Katakomba: Tools and Benchmarks for Data-Driven NetHack

Vladislav Kurenkov, Alexander Nikulin, Denis Tarasov, Sergey Kolesnikov

NeurIPS 2023 · Reinforcement learning › Offline RL · approach: Simple practical offline RL · 0 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 HDF5 memory-mapped storage instead of TTYRecNetHack dataset loading · NetHack data loading, batch_size=256, seq_len=64loading time: 255ms -> 18msImproved
Use HDF5 in-memory storage instead of TTYRecNetHack dataset loading · NetHack data loading, batch_size=256, seq_len=64loading time: 255ms -> 14msImproved
Use compressed HDF5 with on-read decompression instead of TTYRecNetHack dataset loading · NetHack data loading, batch_size=256, seq_len=64loading time: 255ms -> 2.42sHurt
Set CQL alpha to 0.0001 rather than 0.0005 (ablation)CQL · NetHack, final unnormalized game scoreReturn: 396.50 $\pm$ 39.35 -> 526.72 $\pm$ 71.37Improved
Set CQL alpha to 0.0001 rather than 0.001 (ablation)CQL · NetHack, final unnormalized game scoreReturn: 395.56 $\pm$ 139.92 -> 526.72 $\pm$ 71.37Improved
Set CQL alpha to 0.0001 rather than 0.05 (ablation)CQL · NetHack, final unnormalized game scoreReturn: 226.55 $\pm$ 196.45 -> 526.72 $\pm$ 71.37Improved
Set CQL alpha to 0.0001 rather than 0.01 (ablation)CQL · NetHack, final unnormalized game scoreReturn: 32.42 $\pm$ 43.69 -> 526.72 $\pm$ 71.37Improved
Set CQL alpha to 0.0001 rather than 0.1 (ablation)CQL · NetHack, final unnormalized game scoreReturn: 14.04 $\pm$ 10.29 -> 526.72 $\pm$ 71.37Improved
Set CQL alpha to 0.0001 rather than 0.5 (ablation)CQL · NetHack, final unnormalized game scoreReturn: 0.00 $\pm$ 0.00 -> 526.72 $\pm$ 71.37Improved
Set CQL alpha to 0.0001 rather than 1.0 (ablation)CQL · NetHack, final unnormalized game scoreReturn: 0.59 $\pm$ 0.45 -> 526.72 $\pm$ 71.37Improved
Use a policy-constraint-free method rather than methods with policy constraintsRecurrent offline RL on NetHack · Katakomba benchmark, normalized scorenormalized scoreHurt
Use offline RL algorithms instead of Behavioral CloningKatakomba benchmark · NetHack, normalized score and death levelnormalized scoreNo clear effect

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