{"paper":{"title":"Leveraging Passage Retrieval with Generative Models for Open Domain Question Answering","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"Generative models for open-domain question answering gain from retrieving multiple passages and combining their evidence.","cross_cats":["cs.LG"],"primary_cat":"cs.CL","authors_text":"Edouard Grave, Gautier Izacard","submitted_at":"2020-07-02T17:44:57Z","abstract_excerpt":"Generative models for open domain question answering have proven to be competitive, without resorting to external knowledge. While promising, this approach requires to use models with billions of parameters, which are expensive to train and query. In this paper, we investigate how much these models can benefit from retrieving text passages, potentially containing evidence. We obtain state-of-the-art results on the Natural Questions and TriviaQA open benchmarks. Interestingly, we observe that the performance of this method significantly improves when increasing the number of retrieved passages."},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"We obtain state-of-the-art results on the Natural Questions and TriviaQA open benchmarks. Interestingly, we observe that the performance of this method significantly improves when increasing the number of retrieved passages. This is evidence that generative models are good at aggregating and combining evidence from multiple passages.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"That the observed gains are attributable to the generative model's ability to aggregate evidence across passages rather than to confounding factors such as retrieval quality, prompt formatting, or benchmark-specific artifacts; the abstract provides no controls or ablation details to isolate this mechanism.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"Augmenting generative models with passage retrieval yields state-of-the-art results on Natural Questions and TriviaQA, with performance scaling positively as more passages are retrieved.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"Generative models for open-domain question answering gain from retrieving multiple passages and combining their evidence.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"6a52f2d0347224fae0afe8a6555442b9015082c8aa97caeccc2e00f2b65eb245"},"source":{"id":"2007.01282","kind":"arxiv","version":2},"verdict":{"id":"02da9189-dbba-4d44-b8a9-f08273771386","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-17T12:43:04.017029Z","strongest_claim":"We obtain state-of-the-art results on the Natural Questions and TriviaQA open benchmarks. Interestingly, we observe that the performance of this method significantly improves when increasing the number of retrieved passages. This is evidence that generative models are good at aggregating and combining evidence from multiple passages.","one_line_summary":"Augmenting generative models with passage retrieval yields state-of-the-art results on Natural Questions and TriviaQA, with performance scaling positively as more passages are retrieved.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"That the observed gains are attributable to the generative model's ability to aggregate evidence across passages rather than to confounding factors such as retrieval quality, prompt formatting, or benchmark-specific artifacts; the abstract provides no controls or ablation details to isolate this mechanism.","pith_extraction_headline":"Generative models for open-domain question answering gain from retrieving multiple passages and combining their evidence."},"references":{"count":0,"sample":[],"resolved_work":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","internal_anchors":0},"formal_canon":{"evidence_count":1,"snapshot_sha256":"8ff57abf21d1c56aa30403874a4ea1b258996024ff81d38e4a624819f885c873"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"}