LARP: Language-Agent Role Play for Open-World Games

Language agents have shown impressive problem-solving skills within defined settings and brief timelines. Yet, with the ever-evolving complexities of open-world simulations, there’s a pressing need for agents that can flexibly adapt to complex environments and consistently maintain a long-term memory to ensure coherent actions.

To bridge the gap between language agents and open-world games, we introduce Language Agent for Role-Playing (LARP), which includes a cognitive architecture that encompasses memory processing and a decision-making assistant, an environment interaction module with a feedback-driven learnable action space, and a postprocessing method that promotes the alignment of various personalities.

The LARP framework refines interactions between users and agents, predefined with unique backgrounds and personalities, ultimately enhancing the gaming experience in open-world contexts. Furthermore, it highlights the diverse uses of language models in a range of areas such as entertainment, education, and various simulation scenarios. The project page is released at this https URL.


Awesome stuff. I was just thinking the past days whether people were expanding on the earlier work on LLM-powered agents. (I had in mind the paper which did the rounds several months ago, in which the authors had built a small town and a The Sims-like world where the agents organized a party.)

Please keep posting material like this, it is very much appreciated.

For the record - I just discovered the “gaming” tag on the forum and realized I missed a lot of interesting posts, so I am going through them… Expect other replies.

1 Like

The latency in generating the response will be quite the problem to try to overcome for the game developers. Game players are not patient people on average.

Besides the gaming applications, “LARPing” has all the main ingredients for creating good bots in general.

Lot’s to learn here IMO.


Rapid iteration on llm based npcs in games as a safe and rapid path towards enterprise grade agents.

I pitched the idea to several companies a while back but there wasn’t much interest at the time.

1 Like

The pitch would have to not include the “gaming” context for them to take you seriously, I’m afraid.

But I take it seriously. Creating good AI bot game characters and good enterprise level agents are almost the same thing. Lot’s of overlap.

1 Like

Not all the time. Some companies are getting a lot of dough…

1 Like

True … twas only thinking of non-gaming enterprises :rofl:

Im thinking the general enterprise dollars >> gaming enterprises. But could be wrong, as gaming is huge.

1 Like

Those were gaming companies. But I’m a complete outsider to games.

You obviously can’t talk to enterprise or agency people like that. They’d rather wait until someone else makes the technology mature.

I’m questioning a lot of these numbers and valuations, but I could just be bitter :sweat_smile:


Totally shocked a gaming company wouldn’t consider this. :scream_cat:

Is it the latency of LLM’s they are concerned with? Or just that the NPC’s are rendered on small compute devices like consoles, phones, personal computers, and there is no suitable hardware?

It could work if the NPC’s were cloud based, but maybe not as profitable, at least until LLM scale compute gets cheaper.