GitHub Repository Explorer

Topic: match3 509 repos

data as of 25.09.2026

Advances in reinforcement learning have resulted in breakthroughs for algorithmic gameplay without yielding direct impact on the player experience. However. recent work has shown how to use game playing agents to directly power player assistance modes that improve the player experience. One avenue ripe for assist-mode support is 3D player navigation. For this support to be effective. we need both robust agents and also well-designed assistance methods. This need for an interdisciplinary approach makes it difficult for more specialized researchers to make progress in this new area. In this paper. we contribute the open-source \textit{NavAssist} research pipeline to bridge the gap between different disciplinary groups. We contribute several building blocks: (a) for reinforcement learning researchers. five OpenAI Gym compatible navigation environments to train more effective agents; (b) for game designers. a Unity game project and the pre-trained models to implement new AI-driven assistance methods; (c) and for human-computer interaction researchers. a set of example assistance methods powered by pre-trained models to investigate and evaluate the effects of those assistance methods.

★ 0 1669d ago

This project includes a Match-3 demo. It needs to be developed further in terms of its UI. VFX. and booster features. The following things need to be added in the future: - Some kind of animation queue - A more advanced MVP-level UI - Third-party SDKs necessary for LiveOps (analytics. in-app purchases. ads. etc.). - More advanced booster features

★ 0 463d ago

This is the complete decompiled qud code. it is uploaded for research purposes and all rights for the code are held to the original writers

★ 0 408d ago

Reinforcement learning framework for training martial arts AI agents using self-play PPO in Unity 3D simulations.

★ 0 392d ago