Pith. sign in

REVIEW

Domain-matched Pre-training Tasks for Dense Retrieval

Not yet reviewed by Pith; the record is open.

This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.

SPECIMEN: schema-true, not a live event

T0 review · schema-true

One-sentence machine reading of the paper's core claim.

pith:XXXXXXXX · record.json · timestamp

arxiv 2107.13602 v1 pith:77I535AK submitted 2021-07-28 cs.CL cs.IR

classification cs.CLcs.IR
keywords pre-trainingretrievalinformationmilliontasksacrossadditionalalmost
verification ladder T0 review T1 audit T2 compute T3 formal
0 comments
read the original abstract

Pre-training on larger datasets with ever increasing model size is now a proven recipe for increased performance across almost all NLP tasks. A notable exception is information retrieval, where additional pre-training has so far failed to produce convincing results. We show that, with the right pre-training setup, this barrier can be overcome. We demonstrate this by pre-training large bi-encoder models on 1) a recently released set of 65 million synthetically generated questions, and 2) 200 million post-comment pairs from a preexisting dataset of Reddit conversations made available by pushshift.io. We evaluate on a set of information retrieval and dialogue retrieval benchmarks, showing substantial improvements over supervised baselines.

Discussion (0). Continue with ORCID to comment.

Pith tools