Research map / Multi-agent systems and games / Human-AI and goal-conditioned
Dungeons and Data: A Large-Scale NetHack Dataset
Eric Hambro, Roberta Raileanu, Danielle Rothermel, Vegard Mella, Tim Rocktäschel, Heinrich Kuttler, Naila Murray
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 changed | Where | Measured | Effect |
|---|---|---|---|
| Combine offline demonstrations with online APPO training using BC | APPO on NLD-AA · NLD-AA, Random character | average episode return: 868 ± 200 -> 1282 ± 275 | Improved |
| Combine offline demonstrations with online APPO training using BC | APPO on NLD-AA-Monk · NLD-AA-Monk, Human Monk | average episode return: 2046 ± 360 -> 2809 ± 324 | Improved |
| Combine offline observations with online APPO training using BCO | APPO on NLD-NAO · NLD-NAO, Random character | average episode return | Improved |
| Train with behavioral cloning rather than offline RL | Offline learning on NLD-AA · NLD-AA, Random character | average episode return: 33 ± 70; 160 ± 70; 554 ± 142 | Improved |
| Learn an inverse dynamics model from observation-only trajectories and use BCO | NLD-NAO · NLD-NAO, Random character | average episode return | Improved |
| Use APPO + BCO instead of APPO alone | Online learning with observation-only demonstrations · NLD-NAO, Random character | training stability | Improved |
| Use current offline RL and demonstration-learning methods to learn from NLD-AA | Offline RL and imitation learning baselines · NLD-AA | average episode return | Hurt |
Closest papers by meaning
- Amortized Planning with Large-Scale Transformers: A Case Study on Chess (NeurIPS 2024 · similarity 0.86)
- Large Language Models Play StarCraft II:Benchmarks and A Chain of Summarization Approach (NeurIPS 2024 · similarity 0.86)
- Chess-World-Model: A 10M-Game Benchmark for Exact State Tracking from Chess Move Sequences (NeurIPS 2026 · similarity 0.85)
- LLM-PySC2: Starcraft II learning environment for Large Language Models (NeurIPS 2025 · similarity 0.85)
- BALROG: Benchmarking Agentic LLM and VLM Reasoning On Games (ICLR 2025 · similarity 0.85)
- Efficient Active Imitation Learning with Random Network Distillation (ICLR 2025 · similarity 0.84)
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.
