AI Paper Atlas

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

NetHack is Hard to Hack

Ulyana Piterbarg, Lerrel Pinto, Rob Fergus

NeurIPS 2023 · Multi-agent systems and games › Human-AI and goal-conditioned · approach: Intrinsically motivated exploration · 1 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 hierarchical strategy labels and bilevel decoder (ablation)LSTM policy trained with BC · NLEmean NLE score: 658 -> 931Improved
add hierarchical strategy labels (ablation)LSTM policy trained with BC · NLEmedian NLE score: 403 -> 614Improved
enlarge the baseline decoder without hierarchy (ablation)LSTM policy trained with BC · NLEmean NLE scoreNo clear effect
replace LSTM with transformer-LSTM (ablation)BC policy without hierarchy · NLEmean NLE score: 658 -> 1318Improved
add hierarchy to transformer-LSTM (ablation)BC policy · NLEmean NLE score: 1318 -> 1151Hurt
add APPO RL fine-tuning to BC (ablation)LSTM baseline · NLEmean NLE score: 658 -> 1204Improved
add hierarchy (ablation)APPO + BC LSTM policy · NLEmean NLE score: 1204 -> 1551Improved
add hierarchy (ablation)APPO + BC transformer-LSTM · NLEmean NLE score: 1326 -> 1346No clear effect
add APPO fine-tuning to BC (ablation)transformer-LSTM · NLEmean NLE score: 1318 -> 1326Mixed
ablate LSTM encoder to give a flat transformer (ablation)transformer-LSTM · NLEmean NLE scoreHurt
scale transformer-LSTM from 3 layers (50.8M) to 6 layers (98.9M) (ablation)transformer-LSTM · NLEmean NLE scoreHurt
scale training data from 10 to 100,000 games (ablation)LSTM and 6-layer transformer-LSTM with BC · NLEmean NLE scoreMixed

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