by xiwan · Agent Tool · ★ 73
Last updated: · Indexed by AgentSkillsHub · Auto-synced every 8h
LLM-Game-Agents This is a repo for studying the application of LLM Agents on Games DISCLAIM You need to prepare an access token for your LLM model. Description There are usually 4 types of intervention methods for LLM models: Prompt Engineering: Using prompt templates to guide the LLM's output. RAG: Typically interfaced with a vector database. Fine-Tuning: Not training the full model, can be analogous to LoRA. Pre-Training: Specifically pre-training the large model. Among these, Prompt Engineering has the best cost-performance ratio. Here we will mainly use langchain to complete LLM's contextual awareness and logical reasoning abilities. Examples LLM-werewolf-cn LLM-werewolf-en This social game with LLM(ClaudeV2) demostrates the following capabilities: Cooperation Werewolf Player 1, Player 6 agree to vote at night Suspicion Villager Player 2's dying words: Suspect P4 Argument Villager Player 4 argues that he is not a werewolf Disguise Werewolf Player 6 disguises himself as a villager Su
| Stars | 73 |
| Forks | 8 |
| Language | Jupyter Notebook |
| Category | Agent Tool |
| License | MIT |
| Quality Score | 49.3070672189722/100 |
| Open Issues | 1 |
| Last Updated | 2025-06-15 |
| Created | 2024-02-15 |
| Est. Tokens | ~1125k |
Explore other popular agent tool tools:
LLM-Game-Agents is This is a repo for studying the application of LLM Agents on Games. It is categorized as a Agent Tool with 73 GitHub stars.
LLM-Game-Agents is primarily written in Jupyter Notebook.
You can find installation instructions and usage details in the LLM-Game-Agents GitHub repository at github.com/xiwan/LLM-Game-Agents. The project has 73 stars and 8 forks, indicating an active community.
LLM-Game-Agents is released under the MIT license, making it free to use and modify according to the license terms.