AI Paper Atlas

Research map / Reinforcement learning / Visual and representation RL

Honor of Kings Arena: an Environment for Generalization in Competitive Reinforcement Learning

Hua Wei, Jingxiao Chen, Xiyang Ji, Hongyang Qin, Minwen Deng, Siqin Li, Liang Wang, Weinan Zhang, Yong Yu, Liu Linc, Lanxiao Huang, Deheng Ye and 2 more

NeurIPS 2022 · Reinforcement learning › Visual and representation RL · approach: Curriculum and environment design · 12 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
Increase the number of CPU cores for PPO self-play data collectionPPO self-play · Diaochan vs. BTtraining hours to beat BT: 6.16$\pm$0.009 -> 0.89$\pm$0.06Improved
Transfer a policy trained on Diaochan vs. Diaochan to other opponent heroesPPO · Diaochan vs. different BT opponent heroeswin rateHurt
Train on five opponent-hero tasks instead of only Diaochan vs. DiaochanPPO · Diaochan vs. 20 BT opponent-hero taskswin rateImproved
Distill a policy from models trained on five opponent-hero tasksPPO · Diaochan vs. 20 BT opponent-hero taskswin rateImproved
Transfer a policy trained on Diaochan vs. Diaochan to control other target heroesPPO · Different target heroes vs. Diaochan BTwin rateHurt
Train on five target-hero tasks instead of only Diaochan vs. DiaochanPPO · 20 target-hero tasks vs. Diaochan BTwin rateImproved
Distill a policy from models trained on five target-hero tasksPPO · 20 target-hero tasks vs. Diaochan BTwin rateImproved
Add dual clipping to PPO (ablation)PPO · Diaochan vs. DiaochanperformanceImproved
Remove legal action masking (ablation)PPO · Diaochan vs. DiaochanperformanceHurt
Add an LSTM to the policy network (ablation)PPO · Diaochan vs. DiaochanperformanceImproved
Train Model B against a frozen Model A rather than BTModel B · Model B vs. BTwin rate: 0.91$\pm$0.03 vs. 0.81$\pm$0.25Hurt
Use Elo score instead of a single-opponent win rate to evaluate agentsHonor of Kings Arena agent evaluation · Model A, Model B, and BTElo score: Model A 2372; Model B 1186; BT 0Improved

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