BudgetVerify is proposed, a budget-tiered generator-verifier framework that routes each generated answer to one of three verification tiers: no verification, lightweight check-and-revise, or higher-cost solve-first-then-compare verification.
Abstract
Financial question answering often requires precise numerical extraction, unit handling, and arithmetic over tables and text, but applying expensive verification uniformly wastes test-time compute. We propose BudgetVerify, a budget-tiered generator-verifier framework that routes each generated answer to one of three verification tiers: no verification, lightweight check-and-revise, or higher-cost solve-first-then-compare verification. The router is trained from offline correctness and token-cost outcomes and, at test time, selects a verification tier using information available before verification, including the question, context statistics, the generated answer, and associated generator metadata. The selected tier either returns the generated answer directly or invokes the corresponding verifier. Across six commercial and open-weight base models, BudgetVerify consistently produces more efficient accuracy-cost Pareto frontiers than fixed verification policies by selectively allocating stronger verification only when it is useful. Although absolute performance varies across models, these efficiency gains and the resulting qualitative frontier shape are consistent across generator models.
Retrieval-augmented generation systems can route difficult queries to deeper context, but batch deployments must allocate a shared token budget across calls whose costs vary by query. We formulate selective escalation as finite-batch allocation for financial document question answering. Each of 150 FinanceBench questio...
Jun-Ru Zhu, Yi-Xin Yang, Xiao-Qing Ding et al.· 0 citations
Verifying LLM-generated answers under a shared computational budget requires jointly deciding which candidates to inspect and which verification attributes to evaluate. We formulate this problem as multi-attribute good-arm identification under a global budget: each candidate is an arm evaluated along several costly att...
Bo Xue, Ji Cheng, Shenghao Lyu et al.· arXiv.org· 1 citation
FinFIRST is the first financial benchmark to jointly evaluate answers and supporting evidence through atomic rubrics, retaining final-answer correctness as the primary objective while making the supporting research process measurable, verifiable, and diagnosable.
Automatic prompt optimization (APO) has been widely employed to adapt large language models without updating their weights, yielding promising results. However, existing methods such as GEPA and OPRO assume hundreds to thousands of subject-model calls, far more than is practical behind paid, rate-limited APIs. Under ti...
Hao-Yue Liu, Zhi-Chao Wang, Huan-Yu Yan et al.· 0 citations
The VEX-Bench is introduced, a unified benchmark for evaluating the verification complexity of LLM-generated misinformation, as perceived during screening, across models and generation methods, and shows that no single method dominates all dimensions, underscoring the need for multi-dimensional evaluation.
Hanxun Huang, Yu-Tao Wu, Qi-Zhou Wang et al.· 0 citations
LLM evaluation pipelines often have many candidate judges: general LLM-as-a-judge prompts, reward models, safety classifiers, confidence variants, and task-specific verifiers. The deployment question is not only which judge is best, but which judges should be called, on which examples, and when panel construction shoul...
Bin Zhu, Yi Xie, Yanghui Rao· 2 citations
Related blog posts
MIT News · Artificial Intelligence· news.mit.eduOct 13, 2026
Machine-learning networks can learn without detailed instructions from a designer. Those that do may offer insight into how the brain reorganizes itself during learning.
MIT News · Artificial Intelligence· news.mit.eduSep 29, 2026
Professor Sherry Turkle’s new book, “Artificial Intimacy,” offers a withering critique of chatbots and the antisocial dynamics she believes they encourage.