English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Tools annual

Global: Zendy Free Plan

Global: Zendy Tools monthly

Global: Zendy Plus monthly

Screening tasks that aim to identify a small subset of top alternatives froma large pool are common in business decision-making processes. These tasksoften require substantial human effort to evaluate each alternative'sperformance, making them time-consuming and costly. Motivated by recentadvances in large language models (LLMs), particularly their ability togenerate outputs that align well with human evaluations, we consider anLLM-as-human-evaluator approach for conducting screening virtually, therebyreducing the cost burden. To achieve scalability and cost-effectiveness invirtual screening, we identify that the stochastic nature of LLM outputs andtheir cost structure necessitate efficient budget allocation across allalternatives. To address this, we propose using a top-$m$ greedy evaluationmechanism, a simple yet effective approach that keeps evaluating the currenttop-$m$ alternatives, and design the explore-first top-$m$ greedy (EFG-$m$)algorithm. We prove that EFG-$m$ is both sample-optimal and consistent inlarge-scale virtual screening. Surprisingly, we also uncover a bonus rankingeffect, where the algorithm naturally induces an indifference-based rankingwithin the selected subset. To further enhance practicality, we design a suiteof algorithm variants to improve screening performance and computationalefficiency. Numerical experiments validate our results and demonstrate theeffectiveness of our algorithms. Lastly, we conduct a case study on LLM-basedvirtual screening. The study shows that while LLMs alone may not providemeaningful screening and ranking results when directly queried, integratingthem with our sample-optimal algorithms unlocks their potential forcost-effective, large-scale virtual screening.

Efficient Budget Allocation for Large-Scale LLM-Enabled Virtual Screening