Skip to content

PokéLLMon

An LLM agent for tactical battle games

Visit official site

What PokéLLMon is

The agent is described as an LLM embodied system for tactical battle games, built from three strategies that shape how it acts during a fight.

What you can do with it

  • Use text feedback from battles to refine generation
  • Draw on external knowledge when choosing actions
  • Reduce abrupt action switches through self-consistency

Application scenarios

  • Battle against human players online
  • Review how the agent performs alongside friends
  • Learn tactics from its battle decisions

Pros & cons

✓ What we like

  • Online battles against human players are described
  • Win rates of 49% and 56% are stated in the material

! What to watch out for

  • The material notes vulnerability to consuming strategies and deception
  • Battle feedback is described as text extracted from fights

FAQ

What strategies are combined?

The material names in-context reinforcement learning, knowledge-augmented generation and self-consistent action generation.

What win rates are stated?

The material states 49% in ladder matches and 56% in invitational matches.

Last reviewed: 2026-11-15

More AI office assistant tools

View all →

How we review