gauraang/MultiAgent

🧠 Multi-Agent LLM — Response Space Explorer

Interactive 6D visualization of 199 agent responses across 5 tasks, projected into embedding space via PCA.

• —15 agents compared side-by-side across tasks
• —4 model runs: TinyLlama, Qwen 3.5 (4b + 27b), Gemma 4
• —6D navigation: three sliders rotate through PC1–PC6
• —Hover any dot to see the full prompt and response
• —Task centroids (○) mark the semantic center of each cluster
• —Dot size = consensus rank within the task cluster

📄 [Methodology & Architecture →](https://github.com/GauraangMalikk/Information-Model-an-LLM-Agent-Architecture)