Deep Research, Shallow Evaluation: A Case Study in Meta-Evaluation for Long-Form QA Benchmarks

Fuente: arXiv
Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Hwang, Jena D., Kishore, Varsha, Singh, Amanpreet, Haddad, Dany, Naik, Aakanksha, Hamada, Malachi, Bragg, Jonathan, D'Arcy, Mike, Weld, Daniel S., Wang, Lucy Lu, Downey, Doug, Feldman, Sergey
Format: Preprint
Veröffentlicht: 2026
Schlagworte:
Online-Zugang:
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866911495327055872
author Hwang, Jena D.
Kishore, Varsha
Singh, Amanpreet
Haddad, Dany
Naik, Aakanksha
Hamada, Malachi
Bragg, Jonathan
D'Arcy, Mike
Weld, Daniel S.
Wang, Lucy Lu
Downey, Doug
Feldman, Sergey
author_facet Hwang, Jena D.
Kishore, Varsha
Singh, Amanpreet
Haddad, Dany
Naik, Aakanksha
Hamada, Malachi
Bragg, Jonathan
D'Arcy, Mike
Weld, Daniel S.
Wang, Lucy Lu
Downey, Doug
Feldman, Sergey
contents Recent advances have made long-form report-generating systems widely available. This has prompted evaluation frameworks that use LLM-as-judge protocols and claim verification, along with meta-evaluation frameworks that seek to validate these methods. Many of the meta-evaluations estimate an evaluation quality's by comparing its assessments against human pairwise preferences. Prior work, however, suggests that human pairwise preference may be overly simplistic and can fail to capture nuances of expert expectations. We conduct a case study in meta-evaluation for long-form QA benchmarks using ScholarQA-CS2, a benchmark designed for assessing retrieval-augmented deep-research QA in the scientific domain. We comprehensively validate the benchmark through human pairwise preference judgments, then critically examine the strengths, weaknesses, and confounders of this approach. We show that pairwise preference rankings are best suited for system-level evaluation, while explicit metric-wise annotations and expert annotators are critical for reliable metric-level assessment, with subjectivity remaining a key challenge. Based on our findings, we offer practical guidelines for designing future meta-evaluations that better align evaluation methods, annotator expertise, and reporting practices. By surfacing these methodological challenges, we aim to advance evaluation standards for deep-research systems.
format Preprint
id arxiv_https___arxiv_org_abs_2603_06942
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Deep Research, Shallow Evaluation: A Case Study in Meta-Evaluation for Long-Form QA Benchmarks
Hwang, Jena D.
Kishore, Varsha
Singh, Amanpreet
Haddad, Dany
Naik, Aakanksha
Hamada, Malachi
Bragg, Jonathan
D'Arcy, Mike
Weld, Daniel S.
Wang, Lucy Lu
Downey, Doug
Feldman, Sergey
Computation and Language
Recent advances have made long-form report-generating systems widely available. This has prompted evaluation frameworks that use LLM-as-judge protocols and claim verification, along with meta-evaluation frameworks that seek to validate these methods. Many of the meta-evaluations estimate an evaluation quality's by comparing its assessments against human pairwise preferences. Prior work, however, suggests that human pairwise preference may be overly simplistic and can fail to capture nuances of expert expectations. We conduct a case study in meta-evaluation for long-form QA benchmarks using ScholarQA-CS2, a benchmark designed for assessing retrieval-augmented deep-research QA in the scientific domain. We comprehensively validate the benchmark through human pairwise preference judgments, then critically examine the strengths, weaknesses, and confounders of this approach. We show that pairwise preference rankings are best suited for system-level evaluation, while explicit metric-wise annotations and expert annotators are critical for reliable metric-level assessment, with subjectivity remaining a key challenge. Based on our findings, we offer practical guidelines for designing future meta-evaluations that better align evaluation methods, annotator expertise, and reporting practices. By surfacing these methodological challenges, we aim to advance evaluation standards for deep-research systems.
title Deep Research, Shallow Evaluation: A Case Study in Meta-Evaluation for Long-Form QA Benchmarks
topic Computation and Language
url https://arxiv.org/abs/2603.06942