Our final project is to build a general-purpose reasoning agent
ASU Research Compute (SOL), Solve arbitrary problem-solving tasks, Use inference-time algorithms to implement an agent, Core constraints to use only the provided LLM API, Efficiency requirement of under 20 LLM calls per question
Algorithm Tester API LLM Wrapper Inference Time Algorithms (9)
- Chain of Thought
- Self-Consistency
- Tree of Thought
- ReACT
- Decomposition
- Tool-Augmented Reasoning
- Self-Refine
- Analogical Reasoning
- Self-Debug