Papers › Galactica: A Large Language Model for Science

Galactica: A Large Language Model for Science

16 Nov 2022arXiv:2211.09085archive 2025-07-28

Ross Taylor, Marcin Kardas, Guillem Cucurull, Thomas Scialom, Anthony Hartshorn, Elvis Saravia, Andrew Poulton, Viktor Kerkez, Robert Stojnic

Information overload is a major obstacle to scientific progress. The explosive growth in scientific literature and data has made it ever harder to discover useful insights in a large mass of information. Today scientific knowledge is accessed through search engines, but they are unable to organize scientific knowledge alone. In this paper we introduce Galactica: a large language model that can store, combine and reason about scientific knowledge. We train on a large scientific corpus of papers, reference material, knowledge bases and many other sources. We outperform existing models on a range of scientific tasks. On technical knowledge probes such as LaTeX equations, Galactica outperforms the latest GPT-3 by 68.2% versus 49.0%. Galactica also performs well on reasoning, outperforming Chinchilla on mathematical MMLU by 41.3% to 35.7%, and PaLM 540B on MATH with a score of 20.4% versus 8.8%. It also sets a new state-of-the-art on downstream tasks such as PubMedQA and MedMCQA dev of 77.6% and 52.9%. And despite not being trained on a general corpus, Galactica outperforms BLOOM and OPT-175B on BIG-bench. We believe these results demonstrate the potential for language models as a new interface for science. We open source the model for the benefit of the scientific community.

PaperPDFConference PDFCodeCode Syntology ran

In Syntology View this paper on Syntology, its page in Syntology's graph. That page adds which of its functions are also present in other papers' code Syntology holds, with dates, and its dated run record as JSON and a README badge. It also lists the repositories linked to the paper, the abstract and the calls for agents.

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

For agents, on Syntology's MCP service (how to connect):

Code

Syntology 2 code samples harvested from 1 repository linked to this paper: 2 ran (of which 0 constructed an object rather than computing a result; 2 with no instrument failure: 0 honoured, 0 violated, 2 with no contract checked; 0 where Syntology's instrument failed) · 0 unverified. “Where Syntology's instrument failed” counts failures of Syntology's instrument, not of the code; here there are none.

By repository: official repository: 2 samples from 1 repository, 2 ran (2 with no instrument failure; 0 where Syntology's instrument failed). The run record, sample by sample. “Ran” means executed on a synthesized input, not that the code is correct or reproduces the paper.

Levels written on 2026-09-25 (UTC): the dates Syntology's graph wrote each sample's verification level, not when the code executed.

paperswithcode/galai officialpytorchApache-2.0Syntology: 2 samples, 2 ran (0 instrument-failed) · 0 unverified 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

2 samples harvested: 2 ran (of which 0 constructed an object rather than computing a result; 2 with no instrument failure: 0 honoured, 0 violated, 2 with no contract checked; 0 where Syntology's instrument failed) · 0 unverified. “Where Syntology's instrument failed” counts failures of Syntology's instrument, not of the code. Syntology's record for this page has not changed since , the first build that kept a record date for it; when this build read Syntology's graph is in the build record. That date is not when the samples ran.

2ran

Licence: 0 of the 2 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 paperswithcode/galai. “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. A line beginning “fixture:” under a sample that ran gives the arguments Syntology's fixture called the sample with, each as name:type/rank/element kind (? where none was recorded), and the type of what the fixture's run returned, exactly as the harness recorded them; the arguments and the returned type were recorded by two separate runs of the same fixture. No output value is shown, and a returned type says nothing about whether the output was right. “Level recorded” gives the UTC date when Syntology's graph wrote the sample's verification level, not when the code executed; where the level was written before Syntology's graph kept that date, the row says “level recorded, date not recorded”. “Honoured” and “violated” refer to a contract Syntology drafted from the code itself. “Our draft was wrong” and “the contract check could not run” are failures of Syntology's instrument, not of the code: the sample ran on Syntology's synthesized input, and then the drafted contract did not describe it, or the check itself raised an error (RAISES), lacked a dependency (DEP_MISSING) or ran out of time (TIMEOUT); hover the label for which. A metamorphic tier, where shown, is evidence of the output's shape, not that the sample implements the method.

Each sample ends with its code_sha256. A sample's code_sha256 is Syntology's identity for that exact code. It is printed in full (16 hexadecimal characters for every sample in this build), and get_code(code_sha256="…") takes it exactly as printed. An agent calls it on Syntology's MCP service (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 are grouped by repository: a repository marked official first, then by how directly the repository is tied to the paper, then in the archive's order; never by how many ran. 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 six 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; the licence is outside the permissive list Syntology serves text under (MIT, Apache-2.0, BSD and similar); this paper's copy was not located, so its licence is not recorded; or this paper's copy is cleared, but the stored copy of the same code that Syntology serves is not. 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.

A function or class name links to its page when the name occurs in the harvested code of at least 5 papers; the name is a string, not an identity.

paperswithcode/galai official repository 2 samples, 2 ran (0 instrument-failed) · 0 unverified
escape_custom_split_sequence in 5 papers galai/utils.py ran fingerprinted Apache-2.0 (permissive) · 779d2e3138d09f2d · report fixture: signature text:str/0/str · returned str · level recorded 2026-09-25
extract_references_from_text galai/utils.py ran fingerprinted Apache-2.0 (permissive) · 282223027b104a81 · report fixture: signature text:str/0/str · returned list · level recorded 2026-09-25

Tasks

AnachronismsBias DetectionCitation PredictionClassificationCommon Sense ReasoningIUPAC Name PredictionKnowledge ProbingLanguage ModelingLanguage ModellingLarge Language ModelMMLUMathMath Word Problem SolvingMathematical ReasoningMolecular Property PredictionMulti-task Language UnderstandingMultiple Choice Question Answering (MCQA)Protein AnnotationProtein Function PredictionProtein Structure PredictionQuestion AnsweringStereotypical Bias AnalysisTDC ADMET Benchmarking GroupWord Sense Disambiguationmodel

Results from the paper archive 2025-07-28

TaskDatasetModelMetricValueRank at snapshotLeaderboardReport
Bias Detection StereoSet GAL 120B ICAT Score 65.6 #8 of 11 Archive leaderboard report
Bias Detection StereoSet GAL 120B LMS 75 #8 of 11 Archive leaderboard report
Bias Detection StereoSet GAL 120B SS 56.2 #8 of 11 Archive leaderboard report
Bias Detection StereoSet GPT-3 (text-davinci-002) ICAT Score 60.8 #10 of 11 Archive leaderboard report
Bias Detection StereoSet GPT-3 (text-davinci-002) LMS 77.6 #10 of 11 Archive leaderboard report
Bias Detection StereoSet GPT-3 (text-davinci-002) SS 60.8 #10 of 11 Archive leaderboard report
Bias Detection StereoSet OPT 175B ICAT Score 60 #11 of 11 Archive leaderboard report
Bias Detection StereoSet OPT 175B LMS 74.8 #11 of 11 Archive leaderboard report
Bias Detection StereoSet OPT 175B SS 59.9 #11 of 11 Archive leaderboard report
Common Sense Reasoning ARC (Challenge) GAL 120B (zero-shot) Accuracy 67.9 #20 of 54 Archive leaderboard report
Common Sense Reasoning ARC (Challenge) GPT-3 (zero-shot) Accuracy 51.4 #33 of 54 Archive leaderboard report
Common Sense Reasoning ARC (Challenge) BLOOM (few-shot, k=5) Accuracy 32.9 #50 of 54 Archive leaderboard report
Common Sense Reasoning ARC (Challenge) OPT (few-shot, k=5) Accuracy 31.1 #52 of 54 Archive leaderboard report
Common Sense Reasoning ARC (Easy) GAL 120B (0-shot) Accuracy 83.8 #9 of 47 Archive leaderboard report
Common Sense Reasoning ARC (Easy) GPT-3 (zero-shot) Accuracy 68.8 #37 of 47 Archive leaderboard report
Common Sense Reasoning ARC (Easy) BLOOM (5-shot) Accuracy 40.7 #43 of 47 Archive leaderboard report
Common Sense Reasoning ARC (Easy) OPT (5-shot) Accuracy 37.4 #45 of 47 Archive leaderboard report
Math Word Problem Solving MATH Minerva 540B (5-shot) mCoT Accuracy 33.6 #83 of 135 Archive leaderboard report
Math Word Problem Solving MATH Minerva 540B (5-shot) mCoT Parameters (Billions) 540 #83 of 135 Archive leaderboard report
Math Word Problem Solving MATH GAL 120B (5-shot) mCoT Accuracy 20.4 #103 of 135 Archive leaderboard report
Math Word Problem Solving MATH GAL 120B (5-shot) mCoT Parameters (Billions) 120 #103 of 135 Archive leaderboard report
Math Word Problem Solving MATH GAL 120B <work> Accuracy 16.6 #107 of 135 Archive leaderboard report
Math Word Problem Solving MATH GAL 120B <work> Parameters (Billions) 120 #107 of 135 Archive leaderboard report
Math Word Problem Solving MATH GAL 30B (5-shot) mCoT Accuracy 12.7 #112 of 135 Archive leaderboard report
Math Word Problem Solving MATH GAL 30B (5-shot) mCoT Parameters (Billions) 30 #112 of 135 Archive leaderboard report
Math Word Problem Solving MATH GAL 30B <work> Accuracy 11.4 #114 of 135 Archive leaderboard report
Math Word Problem Solving MATH GAL 30B <work> Parameters (Billions) 30 #114 of 135 Archive leaderboard report
Math Word Problem Solving MATH PaLM 540B (5-shot) mCoT Accuracy 8.8 #118 of 135 Archive leaderboard report
Math Word Problem Solving MATH PaLM 540B (5-shot) mCoT Parameters (Billions) 540 #118 of 135 Archive leaderboard report
Math Word Problem Solving MATH GPT-3 175B (8-shot) Accuracy 5.2 #129 of 135 Archive leaderboard report
Math Word Problem Solving MATH GPT-3 175B (8-shot) Parameters (Billions) 175 #129 of 135 Archive leaderboard report
Molecular Property Prediction BACE GAL 30B ROC-AUC 72.7 #16 of 20 Archive leaderboard report
Molecular Property Prediction BACE GAL 120B ROC-AUC 61.7 #17 of 20 Archive leaderboard report
Molecular Property Prediction BACE GAL 6.7B ROC-AUC 58.4 #18 of 20 Archive leaderboard report
Molecular Property Prediction BACE GAL 1.3B ROC-AUC 57.6 #19 of 20 Archive leaderboard report
Molecular Property Prediction BACE GAL 125M ROC-AUC 56.1 #20 of 20 Archive leaderboard report
Molecular Property Prediction BBBP Uni-Mol ROC-AUC 72.9 #16 of 29 Archive leaderboard report
Molecular Property Prediction BBBP GAL 120B ROC-AUC 66.1 #25 of 29 Archive leaderboard report
Molecular Property Prediction BBBP GAL 1.3B ROC-AUC 60.4 #26 of 29 Archive leaderboard report
Molecular Property Prediction BBBP GAL 30B ROC-AUC 59.6 #27 of 29 Archive leaderboard report
Molecular Property Prediction BBBP GAL 6.7B ROC-AUC 53.5 #28 of 29 Archive leaderboard report
Molecular Property Prediction BBBP GAL 125M ROC-AUC 39.3 #29 of 29 Archive leaderboard report
Molecular Property Prediction HIV dataset Uni-Mol AUC 0.808 #2 of 11 Archive leaderboard report
Molecular Property Prediction HIV dataset GAL 30B AUC 0.759 #7 of 11 Archive leaderboard report
Molecular Property Prediction HIV dataset GAL 120B AUC 0.745 #8 of 11 Archive leaderboard report
Molecular Property Prediction HIV dataset GAL 1.3B AUC 0.724 #9 of 11 Archive leaderboard report
Molecular Property Prediction HIV dataset GAL 6.7B AUC 0.722 #10 of 11 Archive leaderboard report
Molecular Property Prediction HIV dataset GAL 125M AUC 0.702 #11 of 11 Archive leaderboard report
Molecular Property Prediction MoleculeNet Uni-Mol AUC 0.77 #1 of 5 Archive leaderboard report
Molecular Property Prediction MoleculeNet GAL 30B AUC 0.69 #2 of 5 Archive leaderboard report
Molecular Property Prediction MoleculeNet GAL 6.7B AUC 0.64 #3 of 5 Archive leaderboard report
Molecular Property Prediction MoleculeNet GAL 1.3B AUC 0.619 #4 of 5 Archive leaderboard report
Molecular Property Prediction MoleculeNet GAL 125M AUC 0.581 #5 of 5 Archive leaderboard report
Molecular Property Prediction SIDER GAL 120B ROC-AUC 63.2 #13 of 19 Archive leaderboard report
Molecular Property Prediction SIDER GAL 30B ROC-AUC 61.3 #15 of 19 Archive leaderboard report
Molecular Property Prediction SIDER GAL 125M ROC-AUC 55.9 #17 of 19 Archive leaderboard report
Molecular Property Prediction SIDER GAL 6.7B ROC-AUC 55.9 #18 of 19 Archive leaderboard report
Molecular Property Prediction SIDER GAL 1.3B ROC-AUC 54.0 #19 of 19 Archive leaderboard report
Molecular Property Prediction Tox21 Uni-Mol ROC-AUC 79.6 #6 of 20 Archive leaderboard report
Molecular Property Prediction Tox21 GAL 120B ROC-AUC 68.9 #16 of 20 Archive leaderboard report
Molecular Property Prediction Tox21 GAL 30B ROC-AUC 68.5 #17 of 20 Archive leaderboard report
Molecular Property Prediction Tox21 GAL 6.7B ROC-AUC 63.9 #18 of 20 Archive leaderboard report
Molecular Property Prediction Tox21 GAL 1.3B ROC-AUC 60.6 #19 of 20 Archive leaderboard report
Molecular Property Prediction Tox21 GAL 125M ROC-AUC 54.3 #20 of 20 Archive leaderboard report
Molecular Property Prediction clintox GAL 120B Molecules (M) 2 #10 of 20 Archive leaderboard report
Molecular Property Prediction clintox GAL 120B ROC-AUC 82.6 #10 of 20 Archive leaderboard report
Molecular Property Prediction clintox GAL 30B Molecules (M) 2 #11 of 20 Archive leaderboard report
Molecular Property Prediction clintox GAL 30B ROC-AUC 82.2 #11 of 20 Archive leaderboard report
Molecular Property Prediction clintox GAL 6.7B Molecules (M) 2 #14 of 20 Archive leaderboard report
Molecular Property Prediction clintox GAL 6.7B ROC-AUC 78.4 #14 of 20 Archive leaderboard report
Molecular Property Prediction clintox GAL 1.3B Molecules (M) 2 #18 of 20 Archive leaderboard report
Molecular Property Prediction clintox GAL 1.3B ROC-AUC 58.9 #18 of 20 Archive leaderboard report
Molecular Property Prediction clintox GAL 125M Molecules (M) 2 #20 of 20 Archive leaderboard report
Molecular Property Prediction clintox GAL 125M ROC-AUC 51.8 #20 of 20 Archive leaderboard report
Multi-task Language Understanding MML GAL 120B (zero-shot) Average (%) 52.6 #25 of 44 Archive leaderboard report
Multiple Choice Question Answering (MCQA) MMLU (Abstract Algebra) GAL 30B (zero-shot) Accuracy 33.3 #1 of 5 Archive leaderboard report
Multiple Choice Question Answering (MCQA) MMLU (Abstract Algebra) Chinchilla (few-shot, k=5) Accuracy 31 #2 of 5 Archive leaderboard report
Multiple Choice Question Answering (MCQA) MMLU (Abstract Algebra) GAL 120B (zero-shot) Accuracy 27 #3 of 5 Archive leaderboard report
Multiple Choice Question Answering (MCQA) MMLU (Abstract Algebra) Gopher (few-shot, k=5) Accuracy 25 #4 of 5 Archive leaderboard report
Multiple Choice Question Answering (MCQA) MMLU (Abstract Algebra) OPT (few-shot, k=5) Accuracy 21 #5 of 5 Archive leaderboard report
Multiple Choice Question Answering (MCQA) MMLU (Astronomy) Chinchilla (few-shot, k=5) Accuracy 73.0 #1 of 5 Archive leaderboard report
Multiple Choice Question Answering (MCQA) MMLU (Astronomy) Gopher (few-shot, k=5) Accuracy 65.8 #2 of 5 Archive leaderboard report
Multiple Choice Question Answering (MCQA) MMLU (Astronomy) GAL 120B (zero-shot) Accuracy 65.1 #3 of 5 Archive leaderboard report
Multiple Choice Question Answering (MCQA) MMLU (Astronomy) BLOOM (few-shot, k=5) Accuracy 25.7 #4 of 5 Archive leaderboard report
Multiple Choice Question Answering (MCQA) MMLU (Astronomy) OPT (few-shot, k=5) Accuracy 23.0 #5 of 5 Archive leaderboard report
Multiple Choice Question Answering (MCQA) MMLU (College Biology) Chinchilla (few-shot, k=5) Accuracy 79.9 #4 of 8 Archive leaderboard report
Multiple Choice Question Answering (MCQA) MMLU (College Biology) Gopher (few-shot, k=5) Accuracy 70.8 #5 of 8 Archive leaderboard report
Multiple Choice Question Answering (MCQA) MMLU (College Biology) GAL 120B (zero-shot) Accuracy 68.8 #6 of 8 Archive leaderboard report
Multiple Choice Question Answering (MCQA) MMLU (College Biology) OPT (few-shot, k=5) Accuracy 30.6 #7 of 8 Archive leaderboard report
Multiple Choice Question Answering (MCQA) MMLU (College Biology) BLOOM (few-shot, k=5) Accuracy 28.5 #8 of 8 Archive leaderboard report
Multiple Choice Question Answering (MCQA) MMLU (College Chemistry) Chinchilla (few-shot, k=5) Accuracy 51 #1 of 5 Archive leaderboard report
Multiple Choice Question Answering (MCQA) MMLU (College Chemistry) GAL 120B (zero-shot) Accuracy 46 #2 of 5 Archive leaderboard report
Multiple Choice Question Answering (MCQA) MMLU (College Chemistry) Gopher (few-shot, k=5) Accuracy 45 #3 of 5 Archive leaderboard report
Multiple Choice Question Answering (MCQA) MMLU (College Chemistry) OPT (few-shot, k=5) Accuracy 30 #4 of 5 Archive leaderboard report
Multiple Choice Question Answering (MCQA) MMLU (College Chemistry) BLOOM (few-shot, k=5) Accuracy 19 #5 of 5 Archive leaderboard report
Multiple Choice Question Answering (MCQA) MMLU (College Computer Science) Chinchilla (few-shot, k=5) Accuracy 51.0 #1 of 4 Archive leaderboard report
Multiple Choice Question Answering (MCQA) MMLU (College Computer Science) GAL 120B (zero-shot) Accuracy 49 #2 of 4 Archive leaderboard report
Multiple Choice Question Answering (MCQA) MMLU (College Computer Science) OPT (few-shot, k=5) Accuracy 17.0 #3 of 4 Archive leaderboard report
Multiple Choice Question Answering (MCQA) MMLU (College Computer Science) BLOOM (few-shot, k=5) Accuracy 6.0 #4 of 4 Archive leaderboard report
Multiple Choice Question Answering (MCQA) MMLU (College Mathematics) GAL 120B (zero-shot) Accuracy 43 #1 of 5 Archive leaderboard report
Multiple Choice Question Answering (MCQA) MMLU (College Mathematics) Gopher (few-shot, k=5) Accuracy 37 #2 of 5 Archive leaderboard report
Multiple Choice Question Answering (MCQA) MMLU (College Mathematics) OPT (few-shot, k=5) Accuracy 33 #3 of 5 Archive leaderboard report
Multiple Choice Question Answering (MCQA) MMLU (College Mathematics) Chinchilla (few-shot, k=5) Accuracy 32 #4 of 5 Archive leaderboard report
Multiple Choice Question Answering (MCQA) MMLU (College Mathematics) BLOOM (few-shot, k=5) Accuracy 25 #5 of 5 Archive leaderboard report
Multiple Choice Question Answering (MCQA) MMLU (College Physics) Chinchilla (few-shot, k=5) Accuracy 46.1 #1 of 5 Archive leaderboard report
Multiple Choice Question Answering (MCQA) MMLU (College Physics) GAL 120B (zero-shot) Accuracy 42.2 #2 of 5 Archive leaderboard report
Multiple Choice Question Answering (MCQA) MMLU (College Physics) Gopher (few-shot, k=5) Accuracy 34.3 #3 of 5 Archive leaderboard report
Multiple Choice Question Answering (MCQA) MMLU (College Physics) OPT (few-shot, k=5) Accuracy 21.6 #4 of 5 Archive leaderboard report
Multiple Choice Question Answering (MCQA) MMLU (College Physics) BLOOM (few-shot, k=5) Accuracy 18.6 #5 of 5 Archive leaderboard report
Multiple Choice Question Answering (MCQA) MMLU (Econometrics) Gopher (few-shot, k=5) Accuracy 43 #1 of 5 Archive leaderboard report
Multiple Choice Question Answering (MCQA) MMLU (Econometrics) GAL 120B (zero-shot) Accuracy 42.1 #2 of 5 Archive leaderboard report
Multiple Choice Question Answering (MCQA) MMLU (Econometrics) Chinchilla (few-shot, k=5) Accuracy 38.6 #3 of 5 Archive leaderboard report
Multiple Choice Question Answering (MCQA) MMLU (Econometrics) BLOOM (few-shot, k=5) Accuracy 23.7 #4 of 5 Archive leaderboard report
Multiple Choice Question Answering (MCQA) MMLU (Econometrics) OPT (few-shot, k=5) Accuracy 21 #5 of 5 Archive leaderboard report
Multiple Choice Question Answering (MCQA) MMLU (Electrical Engineer) GAL 120B (zero-shot) Accuracy 62.8 #1 of 5 Archive leaderboard report
Multiple Choice Question Answering (MCQA) MMLU (Electrical Engineer) Chinchilla (few-shot, k=5) Accuracy 62.1 #2 of 5 Archive leaderboard report
Multiple Choice Question Answering (MCQA) MMLU (Electrical Engineer) Gopher (few-shot, k=5) Accuracy 60 #3 of 5 Archive leaderboard report
Multiple Choice Question Answering (MCQA) MMLU (Electrical Engineer) OPT (few-shot, k=5) Accuracy 36.6 #4 of 5 Archive leaderboard report
Multiple Choice Question Answering (MCQA) MMLU (Electrical Engineer) BLOOM (few-shot, k=5) Accuracy 32.4 #5 of 5 Archive leaderboard report
Multiple Choice Question Answering (MCQA) MMLU (Elementary Mathematics) Chinchilla (few-shot, k=5) Accuracy 41.5 #1 of 5 Archive leaderboard report
Multiple Choice Question Answering (MCQA) MMLU (Elementary Mathematics) GAL 120B (zero-shot) Accuracy 38.1 #2 of 5 Archive leaderboard report
Multiple Choice Question Answering (MCQA) MMLU (Elementary Mathematics) Gopher (few-shot, k=5) Accuracy 33.6 #3 of 5 Archive leaderboard report
Multiple Choice Question Answering (MCQA) MMLU (Elementary Mathematics) BLOOM (few-shot, k=5) Accuracy 27.6 #4 of 5 Archive leaderboard report
Multiple Choice Question Answering (MCQA) MMLU (Elementary Mathematics) OPT (few-shot, k=5) Accuracy 25.7 #5 of 5 Archive leaderboard report
Multiple Choice Question Answering (MCQA) MMLU (Formal Logic) Gopher (few-shot, k=5) Accuracy 35.7 #1 of 5 Archive leaderboard report
Multiple Choice Question Answering (MCQA) MMLU (Formal Logic) Chinchilla (few-shot, k=5) Accuracy 33.3 #2 of 5 Archive leaderboard report
Multiple Choice Question Answering (MCQA) MMLU (Formal Logic) GAL 120B (zero-shot) Accuracy 32.5 #3 of 5 Archive leaderboard report
Multiple Choice Question Answering (MCQA) MMLU (Formal Logic) OPT (few-shot, k=5) Accuracy 29.4 #4 of 5 Archive leaderboard report
Multiple Choice Question Answering (MCQA) MMLU (Formal Logic) BLOOM (few-shot, k=5) Accuracy 26.2 #5 of 5 Archive leaderboard report
Multiple Choice Question Answering (MCQA) MMLU (High School Biology) Chinchilla (few-shot, k=5) Accuracy 80.3 #1 of 5 Archive leaderboard report
Multiple Choice Question Answering (MCQA) MMLU (High School Biology) Gopher (few-shot, k=5) Accuracy 71.3 #2 of 5 Archive leaderboard report
Multiple Choice Question Answering (MCQA) MMLU (High School Biology) GAL 120B (zero-shot) Accuracy 69.4 #3 of 5 Archive leaderboard report
Multiple Choice Question Answering (MCQA) MMLU (High School Biology) BLOOM (few-shot, k=5) Accuracy 29.4 #4 of 5 Archive leaderboard report
Multiple Choice Question Answering (MCQA) MMLU (High School Biology) OPT (few-shot, k=5) Accuracy 27.7 #5 of 5 Archive leaderboard report
Multiple Choice Question Answering (MCQA) MMLU (High School Chemistry) Chinchilla (few-shot, k=5) Accuracy 58.1 #1 of 4 Archive leaderboard report
Multiple Choice Question Answering (MCQA) MMLU (High School Chemistry) GAL 120B (zero-shot) Accuracy 47.8 #2 of 4 Archive leaderboard report
Multiple Choice Question Answering (MCQA) MMLU (High School Chemistry) BLOOM (few-shot, k=5) Accuracy 23.2 #3 of 4 Archive leaderboard report
Multiple Choice Question Answering (MCQA) MMLU (High School Chemistry) OPT (few-shot, k=5) Accuracy 21.7 #4 of 4 Archive leaderboard report
Multiple Choice Question Answering (MCQA) MMLU (High School Computer Science) GAL 120B (zero-shot) Accuracy 70 #1 of 5 Archive leaderboard report
Multiple Choice Question Answering (MCQA) MMLU (High School Computer Science) Chinchilla (few-shot, k=5) Accuracy 58 #2 of 5 Archive leaderboard report
Multiple Choice Question Answering (MCQA) MMLU (High School Computer Science) Gopher (few-shot, k=5) Accuracy 54 #3 of 5 Archive leaderboard report
Multiple Choice Question Answering (MCQA) MMLU (High School Computer Science) OPT (few-shot, k=5) Accuracy 30 #4 of 5 Archive leaderboard report
Multiple Choice Question Answering (MCQA) MMLU (High School Computer Science) BLOOM (few-shot, k=5) Accuracy 25 #5 of 5 Archive leaderboard report
Multiple Choice Question Answering (MCQA) MMLU (High School Mathematics) GAL 120B (zero-shot) Accuracy 32.6 #1 of 5 Archive leaderboard report
Multiple Choice Question Answering (MCQA) MMLU (High School Mathematics) Chinchilla (few-shot, k=5) Accuracy 31.9 #2 of 5 Archive leaderboard report
Multiple Choice Question Answering (MCQA) MMLU (High School Mathematics) BLOOM (few-shot, k=5) Accuracy 27 #3 of 5 Archive leaderboard report
Multiple Choice Question Answering (MCQA) MMLU (High School Mathematics) OPT (few-shot, k=5) Accuracy 24.4 #4 of 5 Archive leaderboard report
Multiple Choice Question Answering (MCQA) MMLU (High School Mathematics) Gopher (few-shot, k=5) Accuracy 23.7 #5 of 5 Archive leaderboard report
Multiple Choice Question Answering (MCQA) MMLU (High School Physics) Chinchilla (few-shot, k=5) Accuracy 36.4 #1 of 4 Archive leaderboard report
Multiple Choice Question Answering (MCQA) MMLU (High School Physics) GAL 120B (zero-shot) Accuracy 33.8 #2 of 4 Archive leaderboard report
Multiple Choice Question Answering (MCQA) MMLU (High School Physics) OPT (few-shot, k=5) Accuracy 29.8 #3 of 4 Archive leaderboard report
Multiple Choice Question Answering (MCQA) MMLU (High School Physics) BLOOM (few-shot, k=5) Accuracy 25.2 #4 of 4 Archive leaderboard report
Multiple Choice Question Answering (MCQA) MMLU (High School Statistics) Chinchilla (few-shot, k=5) Accuracy 58.8 #1 of 5 Archive leaderboard report
Multiple Choice Question Answering (MCQA) MMLU (High School Statistics) Gopher (few-shot, k=5) Accuracy 50 #2 of 5 Archive leaderboard report
Multiple Choice Question Answering (MCQA) MMLU (High School Statistics) OPT (few-shot, k=5) Accuracy 43.5 #3 of 5 Archive leaderboard report
Multiple Choice Question Answering (MCQA) MMLU (High School Statistics) GAL 120B (zero-shot) Accuracy 41.2 #4 of 5 Archive leaderboard report
Multiple Choice Question Answering (MCQA) MMLU (High School Statistics) BLOOM (few-shot, k=5) Accuracy 19.4 #5 of 5 Archive leaderboard report
Multiple Choice Question Answering (MCQA) MMLU (Machine Learning) Chinchilla (few-shot, k=5) Accuracy 41.1 #1 of 4 Archive leaderboard report
Multiple Choice Question Answering (MCQA) MMLU (Machine Learning) GAL 120B (zero-shot) Accuracy 38.4 #2 of 4 Archive leaderboard report
Multiple Choice Question Answering (MCQA) MMLU (Machine Learning) OPT (few-shot, k=5) Accuracy 28.6 #3 of 4 Archive leaderboard report
Multiple Choice Question Answering (MCQA) MMLU (Machine Learning) BLOOM (few-shot, k=5) Accuracy 25 #4 of 4 Archive leaderboard report
Multiple Choice Question Answering (MCQA) MMLU (Medical Genetics) GAL 30B (zero-shot) Accuracy 70 #4 of 8 Archive leaderboard report
Multiple Choice Question Answering (MCQA) MMLU (Medical Genetics) Chinchilla (few-shot, k=5) Accuracy 69 #5 of 8 Archive leaderboard report
Multiple Choice Question Answering (MCQA) MMLU (Medical Genetics) GAL 120B (zero-shot) Accuracy 68 #6 of 8 Archive leaderboard report
Multiple Choice Question Answering (MCQA) MMLU (Medical Genetics) BLOOM (few-shot, k=5) Accuracy 36 #7 of 8 Archive leaderboard report
Multiple Choice Question Answering (MCQA) MMLU (Medical Genetics) OPT (few-shot, k=5) Accuracy 35 #8 of 8 Archive leaderboard report
Multiple Choice Question Answering (MCQA) MedMCQA GAL 120B (zero-shot) Dev Set (Acc-%) 0.529 #16 of 22 Archive leaderboard report
Multiple Choice Question Answering (MCQA) MedMCQA BLOOM (few-shot, k=5) Dev Set (Acc-%) 0.325 #20 of 22 Archive leaderboard report
Multiple Choice Question Answering (MCQA) MedMCQA OPT (few-shot, k=5) Dev Set (Acc-%) 0.296 #21 of 22 Archive leaderboard report
Protein Function Prediction CASPSimSeq GAL 120B ROUGE-L 0.252 #1 of 5 Archive leaderboard report
Protein Function Prediction CASPSimSeq GAL 30B ROUGE-L 0.137 #2 of 5 Archive leaderboard report
Protein Function Prediction CASPSimSeq GAL 6.7B ROUGE-L 0.109 #3 of 5 Archive leaderboard report
Protein Function Prediction CASPSimSeq GAL 1.3B ROUGE-L 0.069 #4 of 5 Archive leaderboard report
Protein Function Prediction CASPSimSeq GAL 125M ROUGE-L 0.062 #5 of 5 Archive leaderboard report
Protein Function Prediction PaenSeq GAL 120B ROUGE-L 0.272 #1 of 5 Archive leaderboard report
Protein Function Prediction PaenSeq GAL 30B ROUGE-L 0.196 #2 of 5 Archive leaderboard report
Protein Function Prediction PaenSeq GAL 6.7B ROUGE-L 0.137 #3 of 5 Archive leaderboard report
Protein Function Prediction PaenSeq GAL 1.3B ROUGE-L 0.084 #4 of 5 Archive leaderboard report
Protein Function Prediction PaenSeq GAL 125M ROUGE-L 0.073 #5 of 5 Archive leaderboard report
Protein Function Prediction UniProtSeq GAL 120B ROUGE-L 0.252 #1 of 5 Archive leaderboard report
Protein Function Prediction UniProtSeq GAL 30B ROUGE-L 0.186 #2 of 5 Archive leaderboard report
Protein Function Prediction UniProtSeq GAL 6.7B ROUGE-L 0.111 #3 of 5 Archive leaderboard report
Protein Function Prediction UniProtSeq GAL 1.3B ROUGE-L 0.079 #4 of 5 Archive leaderboard report
Protein Function Prediction UniProtSeq GAL 125M ROUGE-L 0.061 #5 of 5 Archive leaderboard report
Protein Structure Prediction CASPSeq GAL 120B Validation perplexity 17.26 #1 of 5 Archive leaderboard report
Protein Structure Prediction CASPSeq GAL 30B Validation perplexity 17.27 #2 of 5 Archive leaderboard report
Protein Structure Prediction CASPSeq GAL 6.7B Validation perplexity 17.29 #3 of 5 Archive leaderboard report
Protein Structure Prediction CASPSeq GAL 1.3B Validation perplexity 17.58 #4 of 5 Archive leaderboard report
Protein Structure Prediction CASPSeq GAL 125M Validation perplexity 20.62 #5 of 5 Archive leaderboard report
Protein Structure Prediction CASPSimSeq GAL 120B Validation perplexity 12.77 #1 of 5 Archive leaderboard report
Protein Structure Prediction CASPSimSeq GAL 30B Validation perplexity 15.42 #2 of 5 Archive leaderboard report
Protein Structure Prediction CASPSimSeq GAL 6.7B Validation perplexity 16.35 #3 of 5 Archive leaderboard report
Protein Structure Prediction CASPSimSeq GAL 1.3B Validation perplexity 17.04 #4 of 5 Archive leaderboard report
Protein Structure Prediction CASPSimSeq GAL 125M Validation perplexity 19.18 #5 of 5 Archive leaderboard report
Protein Structure Prediction PaenSeq GAL 120B Validation perplexity 3.14 #1 of 5 Archive leaderboard report
Protein Structure Prediction PaenSeq GAL 30B Validation perplexity 4.28 #2 of 5 Archive leaderboard report
Protein Structure Prediction PaenSeq GAL 6.7B Validation perplexity 7.76 #3 of 5 Archive leaderboard report
Protein Structure Prediction PaenSeq GAL 1.3B Validation perplexity 12.53 #4 of 5 Archive leaderboard report
Protein Structure Prediction PaenSeq GAL 125M Validation perplexity 16.35 #5 of 5 Archive leaderboard report
Protein Structure Prediction UniProtSeq GAL 120B Validation perplexity 5.54 #1 of 5 Archive leaderboard report
Protein Structure Prediction UniProtSeq GAL 30B Validation perplexity 8.23 #2 of 5 Archive leaderboard report
Protein Structure Prediction UniProtSeq GAL 6.7B Validation perplexity 11.58 #3 of 5 Archive leaderboard report
Protein Structure Prediction UniProtSeq GAL 1.3B Validation perplexity 15.82 #4 of 5 Archive leaderboard report
Protein Structure Prediction UniProtSeq GAL 125M Validation perplexity 19.05 #5 of 5 Archive leaderboard report
Question Answering BioASQ GAL 120B (zero-shot) Accuracy 94.3 #2 of 7 Archive leaderboard report
Question Answering BioASQ BLOOM (zero-shot) Accuracy 91.4 #4 of 7 Archive leaderboard report
Question Answering BioASQ OPT (zero-shot) Accuracy 81.4 #7 of 7 Archive leaderboard report
Question Answering MedQA GAL 120B (zero-shot) Accuracy 44.4 #20 of 27 Archive leaderboard report
Question Answering MedQA BLOOM (few-shot, k=5) Accuracy 23.3 #26 of 27 Archive leaderboard report
Question Answering MedQA OPT (few-shot, k=5) Accuracy 22.8 #27 of 27 Archive leaderboard report
Question Answering PubMedQA GAL 120B (zero-shot) Accuracy 77.6 #10 of 30 Archive leaderboard report
Question Answering PubMedQA BLOOM (zero-shot) Accuracy 73.6 #20 of 30 Archive leaderboard report
Question Answering PubMedQA OPT (zero-shot) Accuracy 70.2 #24 of 30 Archive leaderboard report
Question Answering TruthfulQA GAL 120B MC1 0.26 #10 of 33 Archive leaderboard report
Question Answering TruthfulQA GAL 30B MC1 0.24 #12 of 33 Archive leaderboard report
Question Answering TruthfulQA OPT 175B MC1 0.21 #18 of 33 Archive leaderboard report
Question Answering TruthfulQA GAL 125M MC1 0.19 #21 of 33 Archive leaderboard report
Question Answering TruthfulQA GAL 1.3B MC1 0.19 #22 of 33 Archive leaderboard report
Question Answering TruthfulQA GAL 6.7B MC1 0.19 #23 of 33 Archive leaderboard report
Stereotypical Bias Analysis CrowS-Pairs GAL 120B Age 69 #1 of 4 Archive leaderboard report
Stereotypical Bias Analysis CrowS-Pairs GAL 120B Disability 66.7 #1 of 4 Archive leaderboard report
Stereotypical Bias Analysis CrowS-Pairs GAL 120B Gender 51.9 #1 of 4 Archive leaderboard report
Stereotypical Bias Analysis CrowS-Pairs GAL 120B Nationality 51.6 #1 of 4 Archive leaderboard report
Stereotypical Bias Analysis CrowS-Pairs GAL 120B Overall 60.5 #1 of 4 Archive leaderboard report
Stereotypical Bias Analysis CrowS-Pairs GAL 120B Physical Appearance 58.7 #1 of 4 Archive leaderboard report
Stereotypical Bias Analysis CrowS-Pairs GAL 120B Race/Color 59.9 #1 of 4 Archive leaderboard report
Stereotypical Bias Analysis CrowS-Pairs GAL 120B Religion 51.9 #1 of 4 Archive leaderboard report
Stereotypical Bias Analysis CrowS-Pairs GAL 120B Sexual Orientation 77.4 #1 of 4 Archive leaderboard report
Stereotypical Bias Analysis CrowS-Pairs GAL 120B Socioeconomic status 65.7 #1 of 4 Archive leaderboard report
TDC ADMET Benchmarking Group tdcommons Galactica-GAL-120B TDC.BBB_Martins 0.661 #8 of 12 Archive leaderboard report
TDC ADMET Benchmarking Group tdcommons Galactica-GAL-1.3B TDC.BBB_Martins 0.604 #9 of 12 Archive leaderboard report
TDC ADMET Benchmarking Group tdcommons Galactica-GAL-30B TDC.BBB_Martins 0.596 #10 of 12 Archive leaderboard report
TDC ADMET Benchmarking Group tdcommons Galactica-GAL-6.7B TDC.BBB_Martins 0.535 #11 of 12 Archive leaderboard report
TDC ADMET Benchmarking Group tdcommons Galactica-GAL-125M TDC.BBB_Martins 0.393 #12 of 12 Archive leaderboard report
Word Sense Disambiguation BIG-bench (Anachronisms) OPT 175B Accuracy 49.1 #3 of 6 Archive leaderboard report
Word Sense Disambiguation BIG-bench (Anachronisms) GAL 120B (few-shot, k=5) Accuracy 48.7 #4 of 6 Archive leaderboard report
Word Sense Disambiguation BIG-bench (Anachronisms) GAL 30B (few-shot, k=5) Accuracy 47.0 #5 of 6 Archive leaderboard report
Word Sense Disambiguation BIG-bench (Anachronisms) BLOOM 176B Accuracy 1.3 #6 of 6 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

Introduced by this paper: Galactica

AdamAttentionAttention DropoutBLOOMBPEChinchillaCosine AnnealingDense ConnectionsDropoutGPT-3GalacticaLayer NormalizationLinear LayerLinear Warmup With Cosine AnnealingMulti-Head AttentionPaLMResidual ConnectionSoftmaxWeight Decay

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