Browse › Methodology › Multi-task Language Understanding › MML
MML Benchmark (Multi-task Language Understanding)
The test covers 57 tasks including elementary mathematics, US history, computer science, law, and more. https://arxiv.org/pdf/2009.03300.pdf
The archive carries no text for this table; the description above is the archive's text for the task Multi-task Language Understanding. archive 2025-07-28
Over time archive 2025-07-28
The chart needs JavaScript; the table below carries every value.
Not inferred: Average (%). Points are placed at the row's paper date; 40 of 44 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 | GPT-4 o1(300b) | 87 | ✓ | Paper | – | 2024 | no code linked | report |
| 2 | Llama 3.1 (405B) | 86.6 | ✓ | Paper | Code | 2024 | linked, not harvested | report |
| 3 | Llama 3.1 (70B) | 86.0 | ✓ | Paper | Code | 2024 | linked, not harvested | report |
| 4 | Gemini Ultra (5-shot) | 83.7 | – | – | – | not matched | report | |
| 5 | Claude 3 Sonnet (5-shot) | 79 | – | Paper | – | 2024 | no code linked | report |
| 6 | Qwen1.5 72B (5-shot) | 77.5 | – | – | – | not matched | report | |
| 7 | Claude 3 Haiku (5-shot) | 75.2 | – | Paper | – | 2024 | no code linked | report |
| 8 | DBRX Instruct 132B (5-shot) | 73.7 | – | Paper | Code | 2024 | 2 of 9 ran · 7 unverified | report |
| 9 | llama 2(65b) | 73.5 | – | Paper | Code | 2022 | 8 of 17 ran · 9 unverified | report |
| 10 | Llama 3.1 8B (CoT) | 73.0 | ✓ | Paper | Code | 2024 | 2 of 9 ran · 7 unverified | report |
| 11 | Mixtral 8x7B (5-shot) | 70.6 | – | Paper | Code | 2024 | 5 of 5 ran · 0 unverified | report |
| 12 | GPT-3.5 Turbo | 70.0 | ✓ | Paper | Code | 2023 | 2 of 5 ran · 3 unverified | report |
| 13 | LLaMA 65B (fine-tuned) | 68.9 | – | Paper | Code | 2023 | 26 of 58 ran · 32 unverified | report |
| 14 | chatgpt/gpt3.5(20B) | 67.5 | – | Paper | Code | 2022 | 8 of 11 ran · 3 unverified | report |
| 15 | LLaMA 65B (5-shot) | 63.4 | – | Paper | Code | 2023 | 26 of 58 ran · 32 unverified | report |
| 16 | LLaMA 2 34B (5-shot) | 62.6 | – | Paper | Code | 2023 | 31 of 52 ran · 21 unverified | report |
| 17 | Mistral 7B (5-shot) | 62.5 | ✓ | Paper | Code | 2024 | 5 of 5 ran · 0 unverified | report |
| 18 | Mistral 7B (5-shot) | 60.1 | – | Paper | Code | 2023 | 9 of 11 ran · 2 unverified | report |
| 19 | GPT-3 Davinci 175B (CoT) | 59.5 | – | Paper | Code | 2022 | 8 of 17 ran · 9 unverified | report |
| 20 | LLaMA 33B (5-shot) | 57.8 | – | Paper | Code | 2023 | 26 of 58 ran · 32 unverified | report |
| 21 | Falcon 40B | 57.0 | – | Paper | – | 2023 | no code linked | report |
| 22 | Qwen 7B (5-shot) | 56.7 | – | – | – | not matched | report | |
| 23 | LLaMA 2 13B (5-shot) | 54.8 | – | Paper | Code | 2023 | 31 of 52 ran · 21 unverified | report |
| 24 | Branch-Train-MiX 4x7B (sampling top-1 experts) | 53.2 | – | Paper | Code | 2024 | linked, not harvested | report |
| 25 | GAL 120B (zero-shot) | 52.6 | – | Paper | Code | 2022 | 0 of 2 ran · 2 unverified | report |
| 26 | Atlas (5-shot) | 47.9 | – | Paper | Code | 2022 | 2 of 2 ran · 0 unverified | report |
| 27 | Flan-T5-XL 3B (CoT) | 45.5 | – | Paper | Code | 2022 | 8 of 17 ran · 9 unverified | report |
| 28 | LLaMA 2 7B (5-shot) | 45.3 | – | Paper | Code | 2023 | 31 of 52 ran · 21 unverified | report |
| 29 | Flan-T5-Large 780M | 45.1 | – | Paper | Code | 2022 | 8 of 17 ran · 9 unverified | report |
| 30 | GLM-130B | 44.8 | – | Paper | Code | 2022 | 5 of 21 ran · 16 unverified | report |
| 31 | Flan-T5-Large 780M (CoT) | 40.5 | – | Paper | Code | 2022 | 8 of 17 ran · 9 unverified | report |
| 32 | GPT-3 Davinci 175B (5-shot) | 39.7 | – | Paper | Code | 2022 | 8 of 17 ran · 9 unverified | report |
| 33 | Bloomberg GPT 50B (5-shot) | 39.2 | – | Paper | Code | 2023 | linked, not harvested | report |
| 34 | UL2 20B (5-shot) | 39.2 | – | Paper | Code | 2022 | 0 of 16 ran · 16 unverified | report |
| 35 | BLOOM 176B (5-shot) | 39.1 | – | Paper | Code | 2023 | linked, not harvested | report |
| 36 | phi-1.5-web 1.3B | 37.9 | – | Paper | Code | 2023 | linked, not harvested | report |
| 37 | OPT 66B (5-shot) | 36 | – | Paper | Code | 2023 | linked, not harvested | report |
| 38 | Flan-T5-Base 250M | 35.9 | – | Paper | Code | 2022 | 8 of 17 ran · 9 unverified | report |
| 39 | Flan-T5-Base 250M (CoT) | 33.7 | – | Paper | Code | 2022 | 8 of 17 ran · 9 unverified | report |
| 40 | GPT-NeoX 20B (5-shot) | 33.6 | – | Paper | Code | 2022 | 0 of 3 ran · 3 unverified | report |
| 41 | RWKV v5 Eagle 7B | 31 | – | – | – | not matched | report | |
| 42 | LLaMA7B-MiLe-Loss(5-shot) | 29.68 | – | Paper | Code | 2023 | linked, not harvested | report |
| 43 | Flan-T5-Small 80M | 28.7 | – | Paper | Code | 2022 | 8 of 17 ran · 9 unverified | report |
| 44 | Falcon 7B (5-shot) | 28.0 | – | Paper | – | 2023 | no code linked | report |
All 44 rows shown. 40 link to a paper page on this site; 6 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). 27 rows have a graph line, from 13 distinct papers; 24 rows (10 papers) have at least one sample that ran. Counting each paper once: Syntology ran 98 of 212 samples; 114 unverified. Separately, 30 of those 212 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