Keftek

Spend Less, Reason Better: Budget-Aware Value Tree Search for LLM Agents

Lets agents hit the same accuracy with far fewer LLM calls by scoring which reasoning paths are worth exploring — cuts agentic inference costs.

AI AgentsLLM Infra
Read original on arXiv