Papers › Hurdles to Progress in Long-form Question Answering

Hurdles to Progress in Long-form Question Answering

10 Mar 2021NAACL 2021 4arXiv:2103.06332archive 2025-07-28

Kalpesh Krishna, Aurko Roy, Mohit Iyyer

The task of long-form question answering (LFQA) involves retrieving documents relevant to a given question and using them to generate a paragraph-length answer. While many models have recently been proposed for LFQA, we show in this paper that the task formulation raises fundamental challenges regarding evaluation and dataset creation that currently preclude meaningful modeling progress. To demonstrate these challenges, we first design a new system that relies on sparse attention and contrastive retriever learning to achieve state-of-the-art performance on the ELI5 LFQA dataset. While our system tops the public leaderboard, a detailed analysis reveals several troubling trends: (1) our system's generated answers are not actually grounded in the documents that it retrieves; (2) ELI5 contains significant train / validation overlap, as at least 81% of ELI5 validation questions occur in paraphrased form in the training set; (3) ROUGE-L is not an informative metric of generated answer quality and can be easily gamed; and (4) human evaluations used for other text generation tasks are unreliable for LFQA. We offer suggestions to mitigate each of these issues, which we hope will lead to more rigorous LFQA research and meaningful progress in the future.

PaperPDFConference PDFCode

In Syntology Open this paper in Syntology's Atlas, the map of the papers in Syntology's graph and their citations.

Code

martiansideofthemoon/hurdles-longform-qa officialmentioned in papermentioned on GitHubpytorch report
carriex/lfqa_eval mentioned on GitHub report

Repository list and official/mentioned flags are the archive's, frozen 2025-07-28. Reachability, where shown, is from one Syntology probe window (2026-09-16 to 2026-09-18); repositories not probed show nothing. GitHub stars are not tracked.

Code Syntology ran Syntology

Not run by Syntology. Nothing on this page verifies that the listed code works.

Tasks

FormLong Form Question AnsweringOpen-Domain DialogOpen-Domain Question AnsweringQuestion AnsweringText Generation

Results from the paper archive 2025-07-28

TaskDatasetModelMetricValueRank at snapshotLeaderboardReport
Open-Domain Question Answering KILT: ELI5 arxiv.org/abs/2103.06332 F1 22.88 #3 of 16 Archive leaderboard report
Open-Domain Question Answering KILT: ELI5 arxiv.org/abs/2103.06332 KILT-F1 2.34 #3 of 16 Archive leaderboard report
Open-Domain Question Answering KILT: ELI5 arxiv.org/abs/2103.06332 KILT-RL 2.36 #3 of 16 Archive leaderboard report
Open-Domain Question Answering KILT: ELI5 arxiv.org/abs/2103.06332 R-Prec 10.67 #3 of 16 Archive leaderboard report
Open-Domain Question Answering KILT: ELI5 arxiv.org/abs/2103.06332 ROUGE-L 23.19 #3 of 16 Archive leaderboard report
Open-Domain Question Answering KILT: ELI5 arxiv.org/abs/2103.06332 Recall@5 24.56 #3 of 16 Archive leaderboard report
Question Answering KILT: ELI5 c-REALM F1 23.1 #3 of 7 Archive leaderboard report
Question Answering KILT: ELI5 c-REALM Rouge-L 23.4 #3 of 7 Archive leaderboard report

Ranks are positions in the archive's leaderboards as they stood at the 2025-07-28 snapshot. Results published since then are not among these rows, so a rank here is not a current standing.

Methods

AdamAttentionDense ConnectionsDropoutLayer NormalizationLinear LayerMulti-Head AttentionReLUResidual ConnectionRouting AttentionRouting TransformerSoftmax

Report a problem or propose a change · a person checks every report against the paper or source before anything changes; decisions are listed on /corrections