Ananto Nayan Bala

Recall Is Not Enough: A Reader-Context Diagnostic for Budget-Constrained Retrieval-Augmented Generation

Casting reader-context construction as budgeted submodular maximization gives a packer that beats both deployed top-k truncation and LLMLingua-2 compression - across three reader families, four scales, and four budgets, at equal-or-lower token cost.