Context Pruning for Cost Reduction

Background: LLM-based agents performing iterative optimization searches are subject to high API costs due to frequent, token-intensive LLM calls.

Question / Future Work: Future work should investigate context pruning techniques, such as SWE-Pruner, to reduce the input token consumption for the LLM in each search iteration, thereby lowering API costs.