Browse › Natural Language Processing › Common Sense Reasoning › ReCoRD
ReCoRD Benchmark (Common Sense Reasoning)
Common sense reasoning tasks are intended to require the model to go beyond pattern recognition. Instead, the model should use "common sense" or world knowledge to make inferences.
The archive carries no text for this table; the description above is the archive's text for the task Common Sense Reasoning. 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: F1 (higher is better). Not inferred (points only, no best-so-far line): EM. Points are placed at the row's paper date; 34 of 45 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 | Turing NLR v5 XXL 5.4B (fine-tuned) | 95.9 | 96.4 | – | Paper | – | 2022 | no code linked | report |
| 2 | ST-MoE-32B 269B (fine-tuned) | 95.1 | – | Paper | Code | 2022 | 5 of 5 ran · 0 unverified | report | |
| 3 | DeBERTa-1.5B | 94.1 | 94.5 | – | Paper | Code | 2020 | 4 of 13 ran · 9 unverified | report |
| 4 | PaLM 540B (finetuned) | 94.0 | 94.6 | – | Paper | Code | 2022 | 30 of 37 ran · 7 unverified | report |
| 5 | Vega v2 6B (fine-tuned) | 93.9 | 94.4 | – | Paper | – | 2022 | no code linked | report |
| 6 | T5-XXL 11B (fine-tuned) | 93.4 | – | Paper | Code | 2019 | 2 of 31 ran · 29 unverified | report | |
| 7 | GESA 500M | 91.7 | 92.2 | – | Paper | – | 2023 | no code linked | report |
| 8 | LUKE-Graph | 91.2 | 91.5 | – | Paper | – | 2023 | no code linked | report |
| 9 | LUKE (single model) | 90.640 | 91.209 | – | – | – | not matched | report | |
| 10 | LUKE 483M | 90.6 | 91.2 | – | Paper | Code | 2020 | 3 of 10 ran · 7 unverified | report |
| 11 | KELM (finetuning RoBERTa-large based single model) | 89.1 | 89.6 | – | Paper | Code | 2021 | 1 of 6 ran · 5 unverified | report |
| 12 | ST-MoE-L 4.1B (fine-tuned) | 88.9 | – | Paper | Code | 2022 | 5 of 5 ran · 0 unverified | report | |
| 13 | FLAN 137B (prompt-tuned) | 85.1 | – | Paper | Code | 2021 | 0 of 1 ran · 1 unverified | report | |
| 14 | XLNet + MTL + Verifier (ensemble) | 83.090 | 83.737 | – | – | – | not matched | report | |
| 15 | GPT-3 Large 760M (0-shot) | 82.1 | – | Paper | Code | 2020 | 15 of 65 ran · 50 unverified | report | |
| 16 | CSRLM (single model) | 81.780 | 82.584 | – | – | – | not matched | report | |
| 17 | XLNet + Verifier | 81.5 | 82.7 | – | Paper | – | 2019 | no code linked | report |
| 18 | XLNet + MTL + Verifier (single model) | 81.460 | 82.664 | – | – | – | not matched | report | |
| 19 | Switch Transformer 9B | 79.9 | – | Paper | – | 2022 | no code linked | report | |
| 20 | {SKG-NET} (single model) | 79.480 | 80.038 | – | – | – | not matched | report | |
| 21 | KELM (finetuning BERT-large based single model) | 76.2 | 76.7 | – | Paper | Code | 2021 | 1 of 6 ran · 5 unverified | report |
| 22 | sMLP – deterministic 9.4B (0-shot) | 73.4 | – | Paper | – | 2022 | no code linked | report | |
| 23 | FLAN 137B (zero-shot) | 72.5 | – | Paper | Code | 2021 | 0 of 1 ran · 1 unverified | report | |
| 24 | Gshard 9B | 72.4 | – | Paper | – | 2022 | no code linked | report | |
| 25 | SKG-BERT (single model) | 72.240 | 72.778 | – | – | – | not matched | report | |
| 26 | KT-NET (single model) | 71.600 | 73.620 | – | – | – | not matched | report | |
| 27 | DCReader+BERT (single model) | 69.490 | 71.138 | – | – | – | not matched | report | |
| 28 | HASH Layers 10B (0-shot) | 67.2 | – | Paper | – | 2022 | no code linked | report | |
| 29 | GraphBert (single) | 60.800 | 62.986 | – | – | – | not matched | report | |
| 30 | Base Layers 10B (0-shot) | 60.7 | – | Paper | – | 2022 | no code linked | report | |
| 31 | GraphBert-WordNet (single) | 59.860 | 61.885 | – | – | – | not matched | report | |
| 32 | GraphBert-NELL (single) | 59.410 | 61.515 | – | – | – | not matched | report | |
| 33 | BERT-Base (single model) | 54.040 | 56.065 | – | Paper | Code | 2018 | 204 of 659 ran · 455 unverified | report |
| 34 | DocQA + ELMo | 45.4 | 46.7 | – | Paper | – | 2018 | no code linked | report |
| 35 | N-Grammer 343M | 28.9 | 29.9 | – | Paper | Code | 2022 | 0 of 6 ran · 6 unverified | report |
| 36 | T5-11B | 94.1 | – | Paper | Code | 2019 | 2 of 31 ran · 29 unverified | report | |
| 37 | PaLM 2-L (one-shot) | 93.8 | – | Paper | Code | 2023 | linked, not harvested | report | |
| 38 | PaLM 2-M (one-shot) | 92.4 | – | Paper | Code | 2023 | linked, not harvested | report | |
| 39 | PaLM 2-S (one-shot) | 92.1 | – | Paper | Code | 2023 | linked, not harvested | report | |
| 40 | GPT-3 175B (one-shot) | 90.2 | – | Paper | Code | 2022 | 0 of 4 ran · 4 unverified | report | |
| 41 | AlexaTM 20B | 88.4 | – | Paper | Code | 2022 | 1 of 1 ran · 0 unverified | report | |
| 42 | Bloomberg GPT 50B (1-shot) | 82.8 | – | Paper | Code | 2023 | linked, not harvested | report | |
| 43 | OPT 66B (1-shot) | 82.5 | – | Paper | Code | 2023 | linked, not harvested | report | |
| 44 | BLOOM 176B (1-shot) | 78 | – | Paper | Code | 2023 | linked, not harvested | report | |
| 45 | GPT-NeoX 20B (1-shot) | 67.9 | – | Paper | Code | 2023 | linked, not harvested | report |
All 45 rows shown. 34 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). 16 rows have a graph line, from 12 distinct papers; 12 rows (9 papers) have at least one sample that ran. Counting each paper once: Syntology ran 265 of 838 samples; 573 unverified. Separately, 162 of those 838 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