Papers › Re2G: Retrieve, Rerank, Generate

Re2G: Retrieve, Rerank, Generate

13 Jul 2022NAACL 2022 7arXiv:2207.06300archive 2025-07-28

Michael Glass, Gaetano Rossiello, Md Faisal Mahbub Chowdhury, Ankita Rajaram Naik, Pengshan Cai, Alfio Gliozzo

As demonstrated by GPT-3 and T5, transformers grow in capability as parameter spaces become larger and larger. However, for tasks that require a large amount of knowledge, non-parametric memory allows models to grow dramatically with a sub-linear increase in computational cost and GPU memory requirements. Recent models such as RAG and REALM have introduced retrieval into conditional generation. These models incorporate neural initial retrieval from a corpus of passages. We build on this line of research, proposing Re2G, which combines both neural initial retrieval and reranking into a BART-based sequence-to-sequence generation. Our reranking approach also permits merging retrieval results from sources with incomparable scores, enabling an ensemble of BM25 and neural initial retrieval. To train our system end-to-end, we introduce a novel variation of knowledge distillation to train the initial retrieval, reranker, and generation using only ground truth on the target sequence output. We find large gains in four diverse tasks: zero-shot slot filling, question answering, fact-checking, and dialog, with relative gains of 9% to 34% over the previous state-of-the-art on the KILT leaderboard. We make our code available as open source at https://github.com/IBM/kgi-slot-filling/tree/re2g.

PaperPDFConference PDFCodeCode Syntology ran

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

For agents, Syntology's MCP tool lists every function and class Syntology harvested from this paper and whether it ran (how to connect): get_harvested_code_for_paper(arxiv_id="2207.06300")

Code

Syntology Ran 1 of 8 code samples harvested from 1 repository linked to this paper; 7 have no recorded run. Of those that ran: 1 ran · our draft was wrong.

By repository: official repository: 8 samples from 1 repository, 1 ran. The run record, sample by sample. “Ran” means executed on a synthesized input, not that the code is correct or reproduces the paper.

ibm/kgi-slot-filling officialmentioned in paperpytorchApache-2.0 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

8 samples harvested; 1 ran; 0 honoured the contract we drafted; 7 have no recorded run. Read from Syntology's graph 2026-09-24; that is when this build read the record, not when the samples ran.

1ran · our draft was wrong
7unverified

Licence: 0 of the 8 samples are pointer only, meaning Syntology does not serve that copy's text. This page shows no code text for any sample; each one links to its file in the repository.

Harvested from ibm/kgi-slot-filling. “Ran” means the sample executed on a synthesized input. It does not mean the output is correct, and nothing here reproduces the paper's results. “Honoured” and “violated” refer to a contract Syntology drafted from the code itself; “our draft was wrong” and “fixture could not drive it” are failures of Syntology's instrument, not of the code.

Each sample ends with its code_sha256, Syntology's identity for that exact code. An agent fetches the stored sample with Syntology's MCP tool get_code(code_sha256="…") (how to connect); click an identity to copy that call.

Repository labels, per sample. official repository: The archive marks this repository official for the paper. named in the paper: The archive records that the paper mentions this repository; it is not marked official. community (archive-listed): In the archive's code links for this paper, not marked official and not recorded as mentioned in the paper. found in paper text by Syntology: Syntology found this repository in the paper's own text; whether it is the authors' implementation is not asserted. community: Not in the archive's code links for this paper; a community repository Syntology harvested. Samples from a repository marked official are listed first. Licence labels name the repository's licence as recorded at harvest. “Pointer only” means Syntology does not serve that copy's text, for one of four reasons: no licence file was found; the licence was not identified; the licence is recorded as permissive but that copy's record is not marked cleared; or the licence is outside the permissive list Syntology serves text under (MIT, Apache-2.0, BSD and similar). Some licences outside that list permit redistribution, such as WTFPL, and GPL-3.0 under its conditions; they are simply not on the list. Hover a licence label for the reason. File links open the file on GitHub at the default branch, which may have changed since the harvest.

normalize_answer ibm/kgi-slot-filling/slot_filling/rag_util.py official repository ran · our draft was wrong fingerprinted Apache-2.0 (permissive) · e7e75981cb464788 · report
clean_text ibm/kgi-slot-filling/slot_filling/kilt_passage_corpus.py official repository unverified Apache-2.0 (permissive) · d890771bd3e4b3f7 · report
find_para_range ibm/kgi-slot-filling/slot_filling/kilt_passage_corpus.py official repository unverified Apache-2.0 (permissive) · 50ec835b3753c6ec · report
get_relation_from_inst ibm/kgi-slot-filling/slot_filling/dataset_stats.py official repository unverified Apache-2.0 (permissive) · e57ce1467b8c1988 · report
matching_passage_ids ibm/kgi-slot-filling/dpr/dpr_kilt_slot_filling_dataset.py official repository unverified Apache-2.0 (permissive) · 0fa853d294de8b23 · report
postprocess_docs ibm/kgi-slot-filling/slot_filling/rag_util.py official repository unverified Apache-2.0 (permissive) · 89afe8d9e930bb10 · report
prefix_allowed_tokens_fn ibm/kgi-slot-filling/slot_filling/rag_client_server_apply.py official repository unverified Apache-2.0 (permissive) · 1d4a7d4d6e3eb7e0 · report
prepare_seq2seq_batch ibm/kgi-slot-filling/slot_filling/rag_util.py official repository unverified Apache-2.0 (permissive) · f92394bf8b4685cc · report

Tasks

Fact CheckingFact VerificationKnowledge DistillationOpen-Domain DialogOpen-Domain Question AnsweringQuestion AnsweringRAGRerankingRetrievalSlot FillingZero-shot Slot Filling

1 archive task tag without a task page not shown.

Results from the paper archive 2025-07-28

TaskDatasetModelMetricValueRank at snapshotLeaderboardReport
Fact Verification KILT: FEVER Re2G Accuracy 89.55 #1 of 33 Archive leaderboard report
Fact Verification KILT: FEVER Re2G KILT-AC 78.53 #1 of 33 Archive leaderboard report
Fact Verification KILT: FEVER Re2G R-Prec 88.92 #1 of 33 Archive leaderboard report
Fact Verification KILT: FEVER Re2G Recall@5 92.52 #1 of 33 Archive leaderboard report
Open-Domain Dialog KILT: Wizard of Wikipedia Re2G F1 18.9 #2 of 21 Archive leaderboard report
Open-Domain Dialog KILT: Wizard of Wikipedia Re2G KILT-F1 12.98 #2 of 21 Archive leaderboard report
Open-Domain Dialog KILT: Wizard of Wikipedia Re2G KILT-RL 11.39 #2 of 21 Archive leaderboard report
Open-Domain Dialog KILT: Wizard of Wikipedia Re2G R-Prec 60.1 #2 of 21 Archive leaderboard report
Open-Domain Dialog KILT: Wizard of Wikipedia Re2G ROUGE-L 16.76 #2 of 21 Archive leaderboard report
Open-Domain Dialog KILT: Wizard of Wikipedia Re2G Recall@5 79.98 #2 of 21 Archive leaderboard report
Open-Domain Question Answering KILT: Natural Questions Re2G EM 51.73 #1 of 16 Archive leaderboard report
Open-Domain Question Answering KILT: Natural Questions Re2G F1 60.97 #1 of 16 Archive leaderboard report
Open-Domain Question Answering KILT: Natural Questions Re2G KILT-EM 43.56 #1 of 16 Archive leaderboard report
Open-Domain Question Answering KILT: Natural Questions Re2G KILT-F1 49.8 #1 of 16 Archive leaderboard report
Open-Domain Question Answering KILT: Natural Questions Re2G R-Prec 70.78 #1 of 16 Archive leaderboard report
Open-Domain Question Answering KILT: Natural Questions Re2G Recall@5 76.63 #1 of 16 Archive leaderboard report
Open-Domain Question Answering KILT: TriviaQA Re2G EM 76.27 #1 of 15 Archive leaderboard report
Open-Domain Question Answering KILT: TriviaQA Re2G F1 81.4 #1 of 15 Archive leaderboard report
Open-Domain Question Answering KILT: TriviaQA Re2G KILT-EM 57.91 #1 of 15 Archive leaderboard report
Open-Domain Question Answering KILT: TriviaQA Re2G KILT-F1 61.78 #1 of 15 Archive leaderboard report
Open-Domain Question Answering KILT: TriviaQA Re2G R-Prec 72.68 #1 of 15 Archive leaderboard report
Open-Domain Question Answering KILT: TriviaQA Re2G Recall@5 74.23 #1 of 15 Archive leaderboard report
Slot Filling KILT: T-REx Re2G Accuracy 87.68 #1 of 20 Archive leaderboard report
Slot Filling KILT: T-REx Re2G F1 89.93 #1 of 20 Archive leaderboard report
Slot Filling KILT: T-REx Re2G KILT-AC 75.84 #1 of 20 Archive leaderboard report
Slot Filling KILT: T-REx Re2G KILT-F1 77.05 #1 of 20 Archive leaderboard report
Slot Filling KILT: T-REx Re2G R-Prec 80.7 #1 of 20 Archive leaderboard report
Slot Filling KILT: T-REx Re2G Recall@5 89.0 #1 of 20 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

AdafactorAdamAttentionAttention DropoutBARTBERTBPECosine AnnealingDense ConnectionsDropoutGPT-3Gated Linear UnitInverse Square Root ScheduleKnowledge DistillationLayer NormalizationLinear LayerLinear Warmup With Cosine AnnealingLinear Warmup With Linear DecayMulti-Head AttentionRAGResidual ConnectionSentencePieceSoftmaxT5Weight DecayWordPiece

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