Browse › Natural Language Processing › Question Answering › HotpotQA
HotpotQA Benchmark (Question Answering)
Question answering can be segmented into domain-specific tasks like community question answering and knowledge-base question answering. Popular benchmark datasets for evaluation question answering systems include SQuAD, HotPotQA, bAbI, TriviaQA, WikiQA, and many others. Models for question answering are typically evaluated on metrics like EM and F1. Some recent top performing models are T5 and XLNet.
The archive carries no text for this table; the description above is the archive's text for the task Question Answering. archive 2025-07-28
Over time archive 2025-07-28
The chart needs JavaScript; the table below carries every value.
Direction inferred from the metric name, not from the archive: JOINT-F1 (higher is better), ANS-F1 (higher is better), SUP-F1 (higher is better). Not inferred (points only, no best-so-far line): ANS-EM, SUP-EM, JOINT-EM. Points are placed at the row's paper date; 23 of 72 rows carry one.
Results archive 2025-07-28
Archive rows end at the archive snapshot, 2025-07-28: no result published after that date is in this table. Rank is the archive's row order at that snapshot; not re-ranked here. Metric values are the archive's strings. Column headers sort the table in your browser; each row keeps its archive rank.
| Paper | Code | Ran Syntology | Report | ||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| 1 | Beam Retrieval | 0.775 | 0.727 | 0.850 | 0.663 | 0.901 | 0.505 | – | Paper | Code | 2023 | 10 of 13 ran · 3 unverified | report |
| 2 | BigBird-etc | 0.736 | 0.755 | 0.891 | – | Paper | Code | 2020 | 10 of 15 ran · 5 unverified | report | |||
| 3 | AISO | 0.720 | 0.675 | 0.805 | 0.612 | 0.860 | 0.449 | – | Paper | Code | 2021 | linked, not harvested | report |
| 4 | Chain-of-Skills | 0.717 | 0.674 | 0.801 | 0.613 | 0.853 | 0.457 | – | Paper | Code | 2023 | linked, not harvested | report |
| 5 | TPRR | 0.708 | 0.670 | 0.795 | 0.594 | 0.843 | 0.444 | – | – | – | not matched | report | |
| 6 | HopRetriever + Sp-search | 0.706 | 0.671 | 0.799 | 0.574 | 0.835 | 0.432 | – | Paper | – | 2020 | no code linked | report |
| 7 | EBS-Large | 0.700 | 0.662 | 0.793 | 0.573 | 0.840 | 0.420 | – | – | – | not matched | report | |
| 8 | HopRetriever | 0.698 | 0.671 | 0.799 | 0.572 | 0.826 | 0.431 | – | – | – | not matched | report | |
| 9 | IRRR+ | 0.696 | 0.663 | 0.791 | 0.569 | 0.832 | 0.428 | – | Paper | Code | 2020 | 3 of 3 ran · 0 unverified | report |
| 10 | EBS-SH | 0.689 | 0.655 | 0.786 | 0.559 | 0.831 | 0.409 | – | – | – | not matched | report | |
| 11 | IRRR | 0.686 | 0.657 | 0.782 | 0.559 | 0.821 | 0.421 | – | Paper | Code | 2020 | 3 of 3 ran · 0 unverified | report |
| 12 | HopRetriever-V2 | 0.678 | 0.648 | 0.778 | 0.561 | 0.818 | 0.410 | – | – | – | not matched | report | |
| 13 | AFSGraph-retriever | 0.670 | 0.646 | 0.778 | 0.557 | 0.812 | 0.411 | – | – | – | not matched | report | |
| 14 | Recursive Dense Retriever | 0.666 | 0.623 | 0.753 | 0.575 | 0.809 | 0.418 | – | Paper | Code | 2020 | linked, not harvested | report |
| 15 | Step-by-Step Retriever | 0.662 | 0.630 | 0.754 | 0.546 | 0.800 | 0.404 | – | – | – | not matched | report | |
| 16 | DDRQA | 0.639 | 0.625 | 0.759 | 0.510 | 0.789 | 0.360 | – | Paper | – | 2020 | no code linked | report |
| 17 | HopRetriever-V1 | 0.639 | 0.608 | 0.739 | 0.531 | 0.793 | 0.380 | – | – | – | not matched | report | |
| 18 | DR model large | 0.630 | 0.620 | 0.753 | 0.499 | 0.778 | 0.354 | – | – | – | not matched | report | |
| 19 | Model name | 0.629 | 0.617 | 0.746 | 0.500 | 0.772 | 0.368 | – | – | – | not matched | report | |
| 20 | HopAns | 0.629 | 0.617 | 0.746 | 0.500 | 0.772 | 0.368 | – | – | – | not matched | report | |
| 21 | Anonymous | 0.629 | 0.604 | 0.732 | 0.520 | 0.771 | 0.380 | – | – | – | not matched | report | |
| 22 | Multi-dimensional-AFSGraph | 0.624 | 0.615 | 0.746 | 0.503 | 0.772 | 0.362 | – | – | – | not matched | report | |
| 23 | HGN-albert + SemanticRetrievalMRS IR | 0.623 | 0.597 | 0.714 | 0.510 | 0.774 | 0.379 | – | – | – | not matched | report | |
| 24 | Tree-shaped-cluster | 0.617 | 0.603 | 0.731 | 0.499 | 0.768 | 0.359 | – | – | – | not matched | report | |
| 25 | AFSgraph | 0.617 | 0.601 | 0.730 | 0.500 | 0.769 | 0.359 | – | – | – | not matched | report | |
| 26 | Robustly Fine-tuned Graph-based Recurrent Retriever | 0.612 | 0.600 | 0.730 | 0.491 | 0.764 | 0.354 | – | Paper | Code | 2019 | 3 of 6 ran · 3 unverified | report |
| 27 | AFSgraph model | 0.609 | 0.601 | 0.730 | 0.485 | 0.759 | 0.350 | – | – | – | not matched | report | |
| 28 | HGN-large + SemanticRetrievalMRS IR | 0.607 | 0.579 | 0.699 | 0.510 | 0.768 | 0.372 | – | – | – | not matched | report | |
| 29 | RoBERTa-DenseRetriever-Fast | 0.602 | 0.598 | 0.727 | 0.480 | 0.749 | 0.345 | – | – | – | not matched | report | |
| 30 | DPR-recurrent | 0.602 | 0.598 | 0.727 | 0.480 | 0.749 | 0.345 | – | – | – | not matched | report | |
| 31 | RoBERTa-DenseRetriever | 0.601 | 0.596 | 0.724 | 0.479 | 0.748 | 0.345 | – | – | – | not matched | report | |
| 32 | HGN + SemanticRetrievalMRS IR | 0.599 | 0.567 | 0.692 | 0.500 | 0.764 | 0.356 | – | Paper | Code | 2019 | 3 of 6 ran · 3 unverified | report |
| 33 | DFGN | 0.5982 | – | Paper | Code | 2019 | 1 of 3 ran · 2 unverified | report | |||||
| 34 | SAFSR model | 0.598 | 0.589 | 0.716 | 0.480 | 0.757 | 0.345 | – | Paper | Code | 2018 | 3 of 8 ran · 5 unverified | report |
| 35 | GraphRR-Fast | 0.569 | 0.582 | 0.709 | 0.429 | 0.713 | 0.310 | – | – | – | not matched | report | |
| 36 | DR model | 0.568 | 0.588 | 0.717 | 0.416 | 0.725 | 0.293 | – | – | – | not matched | report | |
| 37 | Quark + SemanticRetrievalMRS IR | 0.562 | 0.555 | 0.675 | 0.456 | 0.730 | 0.329 | – | Paper | – | 2020 | no code linked | report |
| 38 | GAR-BERT | 0.561 | 0.523 | 0.648 | 0.490 | 0.747 | 0.330 | – | – | – | not matched | report | |
| 39 | Graph-based Recurrent Retriever | 0.553 | 0.560 | 0.689 | 0.441 | 0.730 | 0.292 | – | – | – | not matched | report | |
| 40 | MIR+EPS+BERT | 0.548 | 0.529 | 0.648 | 0.428 | 0.720 | 0.312 | – | – | – | not matched | report | |
| 41 | GAR | 0.530 | 0.482 | 0.613 | 0.483 | 0.739 | 0.306 | – | – | – | not matched | report | |
| 42 | Transformer-XH-final | 0.513 | 0.516 | 0.641 | 0.409 | 0.714 | 0.261 | – | Paper | Code | 2020 | linked, not harvested | report |
| 43 | Transformer-XH | 0.496 | 0.490 | 0.608 | 0.417 | 0.700 | 0.271 | – | – | – | not matched | report | |
| 44 | SemanticRetrievalMRS | 0.476 | 0.453 | 0.573 | 0.387 | 0.708 | 0.251 | – | Paper | Code | 2019 | 6 of 17 ran · 11 unverified | report |
| 45 | DrKIT | 0.429 | 0.421 | 0.517 | 0.371 | 0.598 | 0.247 | – | – | – | not matched | report | |
| 46 | Entity-centric BERT Pipeline | 0.392 | 0.418 | 0.531 | 0.263 | 0.573 | 0.170 | – | – | – | not matched | report | |
| 47 | PR-Bert | 0.391 | 0.433 | 0.538 | 0.219 | 0.596 | 0.145 | – | – | – | not matched | report | |
| 48 | GoldEn Retriever | 0.391 | 0.379 | 0.486 | 0.307 | 0.642 | 0.180 | – | Paper | Code | 2019 | 1 of 1 ran · 0 unverified | report |
| 49 | SAFSr-Bert | 0.370 | 0.394 | 0.514 | 0.242 | 0.585 | 0.133 | – | – | – | not matched | report | |
| 50 | Cognitive Graph QA | 0.349 | 0.371 | 0.489 | 0.228 | 0.577 | 0.124 | – | Paper | Code | 2019 | 5 of 9 ran · 4 unverified | report |
| 51 | GAR-NOSF | 0.334 | 0.475 | 0.606 | 0.076 | 0.448 | 0.049 | – | – | – | not matched | report | |
| 52 | IKFGraph | 0.304 | 0.358 | 0.453 | 0.160 | 0.512 | 0.115 | – | – | – | not matched | report | |
| 53 | AnonymousQ | 0.291 | 0.369 | 0.460 | 0.153 | 0.468 | 0.115 | – | – | – | not matched | report | |
| 54 | HGN Model-reproduce | 0.284 | 0.335 | 0.427 | 0.156 | 0.493 | 0.110 | – | – | – | not matched | report | |
| 55 | MUPPET | 0.270 | 0.306 | 0.403 | 0.167 | 0.473 | 0.109 | – | Paper | Code | 2019 | 0 of 5 ran · 5 unverified | report |
| 56 | GRN + BERT | 0.258 | 0.299 | 0.391 | 0.132 | 0.497 | 0.083 | – | – | – | not matched | report | |
| 57 | Entity-centric IR | 0.255 | 0.354 | 0.463 | 0.001 | 0.432 | 0.000 | – | – | – | not matched | report | |
| 58 | KGNN | 0.247 | 0.277 | 0.372 | 0.127 | 0.472 | 0.070 | – | Paper | – | 2019 | no code linked | report |
| 59 | SAQA | 0.245 | 0.284 | 0.386 | 0.147 | 0.472 | 0.086 | – | – | – | not matched | report | |
| 60 | GRN | 0.236 | 0.273 | 0.365 | 0.122 | 0.488 | 0.074 | – | – | – | not matched | report | |
| 61 | QFE | 0.231 | 0.287 | 0.381 | 0.142 | 0.444 | 0.087 | – | Paper | – | 2019 | no code linked | report |
| 62 | SAFSr_model | 0.209 | 0.289 | 0.391 | 0.080 | 0.406 | 0.041 | – | – | – | not matched | report | |
| 63 | SuppBERT | 0.175 | 0.236 | 0.320 | 0.056 | 0.400 | 0.033 | – | – | – | not matched | report | |
| 64 | Baseline Model | 0.162 | 0.240 | 0.329 | 0.039 | 0.377 | 0.019 | – | Paper | Code | 2018 | 3 of 8 ran · 5 unverified | report |
| 65 | tes | 0.011 | 0.074 | 0.121 | 0.000 | 0.078 | 0.000 | – | – | – | not matched | report | |
| 66 | PromptRank-fewshot-2-demo | 0.000 | 0.581 | 0.711 | 0.000 | 0.000 | 0.000 | – | – | – | not matched | report | |
| 67 | graph-recurrent-retriever+roberta-base w. S/R-pretraining | 0.000 | 0.581 | 0.710 | 0.000 | 0.000 | 0.000 | – | – | – | not matched | report | |
| 68 | TPReasoner w/o BERT | 0.000 | 0.360 | 0.474 | 0.000 | 0.000 | 0.000 | – | – | – | not matched | report | |
| 69 | MultiQA | 0.000 | 0.307 | 0.402 | 0.000 | 0.000 | 0.000 | – | – | – | not matched | report | |
| 70 | DecompRC | 0.000 | 0.300 | 0.407 | 0.000 | 0.000 | 0.000 | – | Paper | Code | 2019 | 1 of 1 ran · 0 unverified | report |
| 71 | 0.000 | 0.300 | 0.407 | 0.000 | 0.000 | 0.000 | – | – | – | not matched | report | ||
| 72 | Mistral multi hop with very large sources | 0.000 | 0.080 | 0.221 | 0.000 | 0.000 | 0.000 | – | – | – | not matched | report |
All 72 rows shown. 23 link to a paper page on this site; 0 are marked as using additional training data in the archive. No GitHub stars are tracked; "Code" is the first repository the archive lists for the row. The archive carries no row tags, review links or community-submitted rows for this table; none are shown. archive 2025-07-28
Syntology Ran reads "N of M ran · U unverified": of the M code samples Syntology harvested from repositories linked to that row's paper (joined by arXiv id), N executed on a synthesized input and the other U = M−N are unverified (harvested, no recorded run). It counts code from repositories linked to that row's paper, not this result: the row's number was not reproduced and nothing here is a correctness claim. The other cell texts mean no graph line for the row: "linked, not harvested" (the archive links code, Syntology has not harvested it), "no code linked" (no code link in the archive), "not matched" (the row's paper URL matched no paper on this site). 14 rows have a graph line, from 12 distinct papers; 13 rows (11 papers) have at least one sample that ran. Counting each paper once: Syntology ran 46 of 87 samples; 41 unverified. Separately, 26 of those 87 are pointer-only (licence): the site points at that code rather than redistributing it, a licence property recorded for ran and unverified samples alike; each cell's tooltip carries the row's own pointer-only count. Read from the graph 2026-09-24. Per-sample status is on the paper page.
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