Inverse Scaling: When Bigger Isn't Better Article Swipe
Ian R. McKenzie
,
Alexander Lyzhov
,
Michael Pieler
,
Alicia Parrish
,
Aaron Mueller
,
Ameya Prabhu
,
Euan McLean
,
Aaron Kirtland
,
Alexis Ross
,
Alisa Liu
,
Andrew Gritsevskiy
,
Daniel Wurgaft
,
D.S. Kauffman
,
Gabriel Recchia
,
Jiacheng Liu
,
J. Patrick Cavanagh
,
M. Weiss
,
Sicong Huang
,
The Floating Droid
,
Tom Tseng
,
Tomasz Korbak
,
Xudong Shen
,
Yuhui Zhang
,
Zhengping Zhou
,
Najoung Kim
,
Samuel R. Bowman
,
Ethan Perez
·
YOU?
·
· 2023
· Open Access
·
· DOI: https://doi.org/10.48550/arxiv.2306.09479
YOU?
·
· 2023
· Open Access
·
· DOI: https://doi.org/10.48550/arxiv.2306.09479
Work on scaling laws has found that large language models (LMs) show predictable improvements to overall loss with increased scale (model size, training data, and compute). Here, we present evidence for the claim that LMs may show inverse scaling, or worse task performance with increased scale, e.g., due to flaws in the training objective and data. We present empirical evidence of inverse scaling on 11 datasets collected by running a public contest, the Inverse Scaling Prize, with a substantial prize pool. Through analysis of the datasets, along with other examples found in the literature, we identify four potential causes of inverse scaling: (i) preference to repeat memorized sequences over following in-context instructions, (ii) imitation of undesirable patterns in the training data, (iii) tasks containing an easy distractor task which LMs could focus on, rather than the harder real task, and (iv) correct but misleading few-shot demonstrations of the task. We release the winning datasets at https://inversescaling.com/data to allow for further investigation of inverse scaling. Our tasks have helped drive the discovery of U-shaped and inverted-U scaling trends, where an initial trend reverses, suggesting that scaling trends are less reliable at predicting the behavior of larger-scale models than previously understood. Overall, our results suggest that there are tasks for which increased model scale alone may not lead to progress, and that more careful thought needs to go into the data and objectives for training language models.
Related Topics
Concepts
Metadata
- Type
- preprint
- Language
- en
- Landing Page
- http://arxiv.org/abs/2306.09479
- https://arxiv.org/pdf/2306.09479
- OA Status
- green
- Cited By
- 22
- Related Works
- 10
- OpenAlex ID
- https://openalex.org/W4381245716
All OpenAlex metadata
Raw OpenAlex JSON
- OpenAlex ID
-
https://openalex.org/W4381245716Canonical identifier for this work in OpenAlex
- DOI
-
https://doi.org/10.48550/arxiv.2306.09479Digital Object Identifier
- Title
-
Inverse Scaling: When Bigger Isn't BetterWork title
- Type
-
preprintOpenAlex work type
- Language
-
enPrimary language
- Publication year
-
2023Year of publication
- Publication date
-
2023-06-15Full publication date if available
- Authors
-
Ian R. McKenzie, Alexander Lyzhov, Michael Pieler, Alicia Parrish, Aaron Mueller, Ameya Prabhu, Euan McLean, Aaron Kirtland, Alexis Ross, Alisa Liu, Andrew Gritsevskiy, Daniel Wurgaft, D.S. Kauffman, Gabriel Recchia, Jiacheng Liu, J. Patrick Cavanagh, M. Weiss, Sicong Huang, The Floating Droid, Tom Tseng, Tomasz Korbak, Xudong Shen, Yuhui Zhang, Zhengping Zhou, Najoung Kim, Samuel R. Bowman, Ethan PerezList of authors in order
- Landing page
-
https://arxiv.org/abs/2306.09479Publisher landing page
- PDF URL
-
https://arxiv.org/pdf/2306.09479Direct link to full text PDF
- Open access
-
YesWhether a free full text is available
- OA status
-
greenOpen access status per OpenAlex
- OA URL
-
https://arxiv.org/pdf/2306.09479Direct OA link when available
- Concepts
-
Scaling, Computer science, Context (archaeology), Task (project management), Inverse, Scale (ratio), CONTEST, Imitation, Artificial intelligence, Mathematics, Psychology, Engineering, Political science, Systems engineering, Geometry, Physics, Quantum mechanics, Law, Social psychology, Biology, PaleontologyTop concepts (fields/topics) attached by OpenAlex
- Cited by
-
22Total citation count in OpenAlex
- Citations by year (recent)
-
2025: 2, 2024: 10, 2023: 10Per-year citation counts (last 5 years)
- Related works (count)
-
10Other works algorithmically related by OpenAlex
Full payload
| id | https://openalex.org/W4381245716 |
|---|---|
| doi | https://doi.org/10.48550/arxiv.2306.09479 |
| ids.doi | https://doi.org/10.48550/arxiv.2306.09479 |
| ids.openalex | https://openalex.org/W4381245716 |
| fwci | 5.61974353 |
| type | preprint |
| title | Inverse Scaling: When Bigger Isn't Better |
| biblio.issue | |
| biblio.volume | |
| biblio.last_page | |
| biblio.first_page | |
| topics[0].id | https://openalex.org/T10028 |
| topics[0].field.id | https://openalex.org/fields/17 |
| topics[0].field.display_name | Computer Science |
| topics[0].score | 0.9994000196456909 |
| topics[0].domain.id | https://openalex.org/domains/3 |
| topics[0].domain.display_name | Physical Sciences |
| topics[0].subfield.id | https://openalex.org/subfields/1702 |
| topics[0].subfield.display_name | Artificial Intelligence |
| topics[0].display_name | Topic Modeling |
| topics[1].id | https://openalex.org/T10181 |
| topics[1].field.id | https://openalex.org/fields/17 |
| topics[1].field.display_name | Computer Science |
| topics[1].score | 0.9921000003814697 |
| topics[1].domain.id | https://openalex.org/domains/3 |
| topics[1].domain.display_name | Physical Sciences |
| topics[1].subfield.id | https://openalex.org/subfields/1702 |
| topics[1].subfield.display_name | Artificial Intelligence |
| topics[1].display_name | Natural Language Processing Techniques |
| topics[2].id | https://openalex.org/T13629 |
| topics[2].field.id | https://openalex.org/fields/17 |
| topics[2].field.display_name | Computer Science |
| topics[2].score | 0.9401000142097473 |
| topics[2].domain.id | https://openalex.org/domains/3 |
| topics[2].domain.display_name | Physical Sciences |
| topics[2].subfield.id | https://openalex.org/subfields/1702 |
| topics[2].subfield.display_name | Artificial Intelligence |
| topics[2].display_name | Text Readability and Simplification |
| is_xpac | False |
| apc_list | |
| apc_paid | |
| concepts[0].id | https://openalex.org/C99844830 |
| concepts[0].level | 2 |
| concepts[0].score | 0.8214654922485352 |
| concepts[0].wikidata | https://www.wikidata.org/wiki/Q102441924 |
| concepts[0].display_name | Scaling |
| concepts[1].id | https://openalex.org/C41008148 |
| concepts[1].level | 0 |
| concepts[1].score | 0.6869125962257385 |
| concepts[1].wikidata | https://www.wikidata.org/wiki/Q21198 |
| concepts[1].display_name | Computer science |
| concepts[2].id | https://openalex.org/C2779343474 |
| concepts[2].level | 2 |
| concepts[2].score | 0.6791302561759949 |
| concepts[2].wikidata | https://www.wikidata.org/wiki/Q3109175 |
| concepts[2].display_name | Context (archaeology) |
| concepts[3].id | https://openalex.org/C2780451532 |
| concepts[3].level | 2 |
| concepts[3].score | 0.6455990076065063 |
| concepts[3].wikidata | https://www.wikidata.org/wiki/Q759676 |
| concepts[3].display_name | Task (project management) |
| concepts[4].id | https://openalex.org/C207467116 |
| concepts[4].level | 2 |
| concepts[4].score | 0.5967094898223877 |
| concepts[4].wikidata | https://www.wikidata.org/wiki/Q4385666 |
| concepts[4].display_name | Inverse |
| concepts[5].id | https://openalex.org/C2778755073 |
| concepts[5].level | 2 |
| concepts[5].score | 0.5439737439155579 |
| concepts[5].wikidata | https://www.wikidata.org/wiki/Q10858537 |
| concepts[5].display_name | Scale (ratio) |
| concepts[6].id | https://openalex.org/C2777582232 |
| concepts[6].level | 2 |
| concepts[6].score | 0.450141966342926 |
| concepts[6].wikidata | https://www.wikidata.org/wiki/Q5013414 |
| concepts[6].display_name | CONTEST |
| concepts[7].id | https://openalex.org/C126388530 |
| concepts[7].level | 2 |
| concepts[7].score | 0.4382326900959015 |
| concepts[7].wikidata | https://www.wikidata.org/wiki/Q1131737 |
| concepts[7].display_name | Imitation |
| concepts[8].id | https://openalex.org/C154945302 |
| concepts[8].level | 1 |
| concepts[8].score | 0.35103729367256165 |
| concepts[8].wikidata | https://www.wikidata.org/wiki/Q11660 |
| concepts[8].display_name | Artificial intelligence |
| concepts[9].id | https://openalex.org/C33923547 |
| concepts[9].level | 0 |
| concepts[9].score | 0.1756339967250824 |
| concepts[9].wikidata | https://www.wikidata.org/wiki/Q395 |
| concepts[9].display_name | Mathematics |
| concepts[10].id | https://openalex.org/C15744967 |
| concepts[10].level | 0 |
| concepts[10].score | 0.14313235878944397 |
| concepts[10].wikidata | https://www.wikidata.org/wiki/Q9418 |
| concepts[10].display_name | Psychology |
| concepts[11].id | https://openalex.org/C127413603 |
| concepts[11].level | 0 |
| concepts[11].score | 0.07435375452041626 |
| concepts[11].wikidata | https://www.wikidata.org/wiki/Q11023 |
| concepts[11].display_name | Engineering |
| concepts[12].id | https://openalex.org/C17744445 |
| concepts[12].level | 0 |
| concepts[12].score | 0.0 |
| concepts[12].wikidata | https://www.wikidata.org/wiki/Q36442 |
| concepts[12].display_name | Political science |
| concepts[13].id | https://openalex.org/C201995342 |
| concepts[13].level | 1 |
| concepts[13].score | 0.0 |
| concepts[13].wikidata | https://www.wikidata.org/wiki/Q682496 |
| concepts[13].display_name | Systems engineering |
| concepts[14].id | https://openalex.org/C2524010 |
| concepts[14].level | 1 |
| concepts[14].score | 0.0 |
| concepts[14].wikidata | https://www.wikidata.org/wiki/Q8087 |
| concepts[14].display_name | Geometry |
| concepts[15].id | https://openalex.org/C121332964 |
| concepts[15].level | 0 |
| concepts[15].score | 0.0 |
| concepts[15].wikidata | https://www.wikidata.org/wiki/Q413 |
| concepts[15].display_name | Physics |
| concepts[16].id | https://openalex.org/C62520636 |
| concepts[16].level | 1 |
| concepts[16].score | 0.0 |
| concepts[16].wikidata | https://www.wikidata.org/wiki/Q944 |
| concepts[16].display_name | Quantum mechanics |
| concepts[17].id | https://openalex.org/C199539241 |
| concepts[17].level | 1 |
| concepts[17].score | 0.0 |
| concepts[17].wikidata | https://www.wikidata.org/wiki/Q7748 |
| concepts[17].display_name | Law |
| concepts[18].id | https://openalex.org/C77805123 |
| concepts[18].level | 1 |
| concepts[18].score | 0.0 |
| concepts[18].wikidata | https://www.wikidata.org/wiki/Q161272 |
| concepts[18].display_name | Social psychology |
| concepts[19].id | https://openalex.org/C86803240 |
| concepts[19].level | 0 |
| concepts[19].score | 0.0 |
| concepts[19].wikidata | https://www.wikidata.org/wiki/Q420 |
| concepts[19].display_name | Biology |
| concepts[20].id | https://openalex.org/C151730666 |
| concepts[20].level | 1 |
| concepts[20].score | 0.0 |
| concepts[20].wikidata | https://www.wikidata.org/wiki/Q7205 |
| concepts[20].display_name | Paleontology |
| keywords[0].id | https://openalex.org/keywords/scaling |
| keywords[0].score | 0.8214654922485352 |
| keywords[0].display_name | Scaling |
| keywords[1].id | https://openalex.org/keywords/computer-science |
| keywords[1].score | 0.6869125962257385 |
| keywords[1].display_name | Computer science |
| keywords[2].id | https://openalex.org/keywords/context |
| keywords[2].score | 0.6791302561759949 |
| keywords[2].display_name | Context (archaeology) |
| keywords[3].id | https://openalex.org/keywords/task |
| keywords[3].score | 0.6455990076065063 |
| keywords[3].display_name | Task (project management) |
| keywords[4].id | https://openalex.org/keywords/inverse |
| keywords[4].score | 0.5967094898223877 |
| keywords[4].display_name | Inverse |
| keywords[5].id | https://openalex.org/keywords/scale |
| keywords[5].score | 0.5439737439155579 |
| keywords[5].display_name | Scale (ratio) |
| keywords[6].id | https://openalex.org/keywords/contest |
| keywords[6].score | 0.450141966342926 |
| keywords[6].display_name | CONTEST |
| keywords[7].id | https://openalex.org/keywords/imitation |
| keywords[7].score | 0.4382326900959015 |
| keywords[7].display_name | Imitation |
| keywords[8].id | https://openalex.org/keywords/artificial-intelligence |
| keywords[8].score | 0.35103729367256165 |
| keywords[8].display_name | Artificial intelligence |
| keywords[9].id | https://openalex.org/keywords/mathematics |
| keywords[9].score | 0.1756339967250824 |
| keywords[9].display_name | Mathematics |
| keywords[10].id | https://openalex.org/keywords/psychology |
| keywords[10].score | 0.14313235878944397 |
| keywords[10].display_name | Psychology |
| keywords[11].id | https://openalex.org/keywords/engineering |
| keywords[11].score | 0.07435375452041626 |
| keywords[11].display_name | Engineering |
| language | en |
| locations[0].id | pmh:oai:arXiv.org:2306.09479 |
| locations[0].is_oa | True |
| locations[0].source.id | https://openalex.org/S4306400194 |
| locations[0].source.issn | |
| locations[0].source.type | repository |
| locations[0].source.is_oa | True |
| locations[0].source.issn_l | |
| locations[0].source.is_core | False |
| locations[0].source.is_in_doaj | False |
| locations[0].source.display_name | arXiv (Cornell University) |
| locations[0].source.host_organization | https://openalex.org/I205783295 |
| locations[0].source.host_organization_name | Cornell University |
| locations[0].source.host_organization_lineage | https://openalex.org/I205783295 |
| locations[0].license | |
| locations[0].pdf_url | https://arxiv.org/pdf/2306.09479 |
| locations[0].version | submittedVersion |
| locations[0].raw_type | text |
| locations[0].license_id | |
| locations[0].is_accepted | False |
| locations[0].is_published | False |
| locations[0].raw_source_name | |
| locations[0].landing_page_url | http://arxiv.org/abs/2306.09479 |
| locations[1].id | doi:10.48550/arxiv.2306.09479 |
| locations[1].is_oa | True |
| locations[1].source.id | https://openalex.org/S4306400194 |
| locations[1].source.issn | |
| locations[1].source.type | repository |
| locations[1].source.is_oa | True |
| locations[1].source.issn_l | |
| locations[1].source.is_core | False |
| locations[1].source.is_in_doaj | False |
| locations[1].source.display_name | arXiv (Cornell University) |
| locations[1].source.host_organization | https://openalex.org/I205783295 |
| locations[1].source.host_organization_name | Cornell University |
| locations[1].source.host_organization_lineage | https://openalex.org/I205783295 |
| locations[1].license | cc-by |
| locations[1].pdf_url | |
| locations[1].version | |
| locations[1].raw_type | article-journal |
| locations[1].license_id | https://openalex.org/licenses/cc-by |
| locations[1].is_accepted | False |
| locations[1].is_published | |
| locations[1].raw_source_name | |
| locations[1].landing_page_url | https://doi.org/10.48550/arxiv.2306.09479 |
| indexed_in | arxiv, datacite |
| authorships[0].author.id | https://openalex.org/A5015269815 |
| authorships[0].author.orcid | |
| authorships[0].author.display_name | Ian R. McKenzie |
| authorships[0].author_position | first |
| authorships[0].raw_author_name | McKenzie, Ian R. |
| authorships[0].is_corresponding | False |
| authorships[1].author.id | https://openalex.org/A5077626187 |
| authorships[1].author.orcid | |
| authorships[1].author.display_name | Alexander Lyzhov |
| authorships[1].author_position | middle |
| authorships[1].raw_author_name | Lyzhov, Alexander |
| authorships[1].is_corresponding | False |
| authorships[2].author.id | https://openalex.org/A5029968236 |
| authorships[2].author.orcid | https://orcid.org/0000-0001-9186-7045 |
| authorships[2].author.display_name | Michael Pieler |
| authorships[2].author_position | middle |
| authorships[2].raw_author_name | Pieler, Michael |
| authorships[2].is_corresponding | False |
| authorships[3].author.id | https://openalex.org/A5082569485 |
| authorships[3].author.orcid | https://orcid.org/0000-0002-1054-0516 |
| authorships[3].author.display_name | Alicia Parrish |
| authorships[3].author_position | middle |
| authorships[3].raw_author_name | Parrish, Alicia |
| authorships[3].is_corresponding | False |
| authorships[4].author.id | https://openalex.org/A5020998070 |
| authorships[4].author.orcid | https://orcid.org/0009-0005-1148-5001 |
| authorships[4].author.display_name | Aaron Mueller |
| authorships[4].author_position | middle |
| authorships[4].raw_author_name | Mueller, Aaron |
| authorships[4].is_corresponding | False |
| authorships[5].author.id | https://openalex.org/A5113274335 |
| authorships[5].author.orcid | |
| authorships[5].author.display_name | Ameya Prabhu |
| authorships[5].author_position | middle |
| authorships[5].raw_author_name | Prabhu, Ameya |
| authorships[5].is_corresponding | False |
| authorships[6].author.id | https://openalex.org/A5057429364 |
| authorships[6].author.orcid | |
| authorships[6].author.display_name | Euan McLean |
| authorships[6].author_position | middle |
| authorships[6].raw_author_name | McLean, Euan |
| authorships[6].is_corresponding | False |
| authorships[7].author.id | https://openalex.org/A5012554092 |
| authorships[7].author.orcid | |
| authorships[7].author.display_name | Aaron Kirtland |
| authorships[7].author_position | middle |
| authorships[7].raw_author_name | Kirtland, Aaron |
| authorships[7].is_corresponding | False |
| authorships[8].author.id | https://openalex.org/A5073199348 |
| authorships[8].author.orcid | |
| authorships[8].author.display_name | Alexis Ross |
| authorships[8].author_position | middle |
| authorships[8].raw_author_name | Ross, Alexis |
| authorships[8].is_corresponding | False |
| authorships[9].author.id | https://openalex.org/A5060080064 |
| authorships[9].author.orcid | |
| authorships[9].author.display_name | Alisa Liu |
| authorships[9].author_position | middle |
| authorships[9].raw_author_name | Liu, Alisa |
| authorships[9].is_corresponding | False |
| authorships[10].author.id | https://openalex.org/A5021580744 |
| authorships[10].author.orcid | https://orcid.org/0000-0001-8138-8796 |
| authorships[10].author.display_name | Andrew Gritsevskiy |
| authorships[10].author_position | middle |
| authorships[10].raw_author_name | Gritsevskiy, Andrew |
| authorships[10].is_corresponding | False |
| authorships[11].author.id | https://openalex.org/A5092206144 |
| authorships[11].author.orcid | |
| authorships[11].author.display_name | Daniel Wurgaft |
| authorships[11].author_position | middle |
| authorships[11].raw_author_name | Wurgaft, Daniel |
| authorships[11].is_corresponding | False |
| authorships[12].author.id | https://openalex.org/A5040736333 |
| authorships[12].author.orcid | |
| authorships[12].author.display_name | D.S. Kauffman |
| authorships[12].author_position | middle |
| authorships[12].raw_author_name | Kauffman, Derik |
| authorships[12].is_corresponding | False |
| authorships[13].author.id | https://openalex.org/A5000771881 |
| authorships[13].author.orcid | https://orcid.org/0000-0002-0210-8635 |
| authorships[13].author.display_name | Gabriel Recchia |
| authorships[13].author_position | middle |
| authorships[13].raw_author_name | Recchia, Gabriel |
| authorships[13].is_corresponding | False |
| authorships[14].author.id | https://openalex.org/A5100656305 |
| authorships[14].author.orcid | https://orcid.org/0000-0002-6637-9258 |
| authorships[14].author.display_name | Jiacheng Liu |
| authorships[14].author_position | middle |
| authorships[14].raw_author_name | Liu, Jiacheng |
| authorships[14].is_corresponding | False |
| authorships[15].author.id | https://openalex.org/A5113095180 |
| authorships[15].author.orcid | https://orcid.org/0000-0003-4379-2602 |
| authorships[15].author.display_name | J. Patrick Cavanagh |
| authorships[15].author_position | middle |
| authorships[15].raw_author_name | Cavanagh, Joe |
| authorships[15].is_corresponding | False |
| authorships[16].author.id | https://openalex.org/A5110994341 |
| authorships[16].author.orcid | |
| authorships[16].author.display_name | M. Weiss |
| authorships[16].author_position | middle |
| authorships[16].raw_author_name | Weiss, Max |
| authorships[16].is_corresponding | False |
| authorships[17].author.id | https://openalex.org/A5043900893 |
| authorships[17].author.orcid | https://orcid.org/0000-0003-0854-4734 |
| authorships[17].author.display_name | Sicong Huang |
| authorships[17].author_position | middle |
| authorships[17].raw_author_name | Huang, Sicong |
| authorships[17].is_corresponding | False |
| authorships[18].author.id | https://openalex.org/A5092206145 |
| authorships[18].author.orcid | |
| authorships[18].author.display_name | The Floating Droid |
| authorships[18].author_position | middle |
| authorships[18].raw_author_name | Droid, The Floating |
| authorships[18].is_corresponding | False |
| authorships[19].author.id | https://openalex.org/A5087045282 |
| authorships[19].author.orcid | https://orcid.org/0000-0002-6422-288X |
| authorships[19].author.display_name | Tom Tseng |
| authorships[19].author_position | middle |
| authorships[19].raw_author_name | Tseng, Tom |
| authorships[19].is_corresponding | False |
| authorships[20].author.id | https://openalex.org/A5002211679 |
| authorships[20].author.orcid | https://orcid.org/0000-0002-6258-2013 |
| authorships[20].author.display_name | Tomasz Korbak |
| authorships[20].author_position | middle |
| authorships[20].raw_author_name | Korbak, Tomasz |
| authorships[20].is_corresponding | False |
| authorships[21].author.id | https://openalex.org/A5091771176 |
| authorships[21].author.orcid | https://orcid.org/0000-0001-9115-8572 |
| authorships[21].author.display_name | Xudong Shen |
| authorships[21].author_position | middle |
| authorships[21].raw_author_name | Shen, Xudong |
| authorships[21].is_corresponding | False |
| authorships[22].author.id | https://openalex.org/A5100624916 |
| authorships[22].author.orcid | https://orcid.org/0000-0002-5769-3456 |
| authorships[22].author.display_name | Yuhui Zhang |
| authorships[22].author_position | middle |
| authorships[22].raw_author_name | Zhang, Yuhui |
| authorships[22].is_corresponding | False |
| authorships[23].author.id | https://openalex.org/A5085639468 |
| authorships[23].author.orcid | https://orcid.org/0000-0002-4706-5499 |
| authorships[23].author.display_name | Zhengping Zhou |
| authorships[23].author_position | middle |
| authorships[23].raw_author_name | Zhou, Zhengping |
| authorships[23].is_corresponding | False |
| authorships[24].author.id | https://openalex.org/A5071542832 |
| authorships[24].author.orcid | |
| authorships[24].author.display_name | Najoung Kim |
| authorships[24].author_position | middle |
| authorships[24].raw_author_name | Kim, Najoung |
| authorships[24].is_corresponding | False |
| authorships[25].author.id | https://openalex.org/A5112713734 |
| authorships[25].author.orcid | |
| authorships[25].author.display_name | Samuel R. Bowman |
| authorships[25].author_position | middle |
| authorships[25].raw_author_name | Bowman, Samuel R. |
| authorships[25].is_corresponding | False |
| authorships[26].author.id | https://openalex.org/A5091112967 |
| authorships[26].author.orcid | |
| authorships[26].author.display_name | Ethan Perez |
| authorships[26].author_position | last |
| authorships[26].raw_author_name | Perez, Ethan |
| authorships[26].is_corresponding | False |
| has_content.pdf | False |
| has_content.grobid_xml | False |
| is_paratext | False |
| open_access.is_oa | True |
| open_access.oa_url | https://arxiv.org/pdf/2306.09479 |
| open_access.oa_status | green |
| open_access.any_repository_has_fulltext | False |
| created_date | 2025-10-10T00:00:00 |
| display_name | Inverse Scaling: When Bigger Isn't Better |
| has_fulltext | False |
| is_retracted | False |
| updated_date | 2025-11-06T06:51:31.235846 |
| primary_topic.id | https://openalex.org/T10028 |
| primary_topic.field.id | https://openalex.org/fields/17 |
| primary_topic.field.display_name | Computer Science |
| primary_topic.score | 0.9994000196456909 |
| primary_topic.domain.id | https://openalex.org/domains/3 |
| primary_topic.domain.display_name | Physical Sciences |
| primary_topic.subfield.id | https://openalex.org/subfields/1702 |
| primary_topic.subfield.display_name | Artificial Intelligence |
| primary_topic.display_name | Topic Modeling |
| related_works | https://openalex.org/W4387497383, https://openalex.org/W3183948672, https://openalex.org/W3173606202, https://openalex.org/W3110381201, https://openalex.org/W2948807893, https://openalex.org/W2778153218, https://openalex.org/W2758277628, https://openalex.org/W1531601525, https://openalex.org/W4308245303, https://openalex.org/W4389519396 |
| cited_by_count | 22 |
| counts_by_year[0].year | 2025 |
| counts_by_year[0].cited_by_count | 2 |
| counts_by_year[1].year | 2024 |
| counts_by_year[1].cited_by_count | 10 |
| counts_by_year[2].year | 2023 |
| counts_by_year[2].cited_by_count | 10 |
| locations_count | 2 |
| best_oa_location.id | pmh:oai:arXiv.org:2306.09479 |
| best_oa_location.is_oa | True |
| best_oa_location.source.id | https://openalex.org/S4306400194 |
| best_oa_location.source.issn | |
| best_oa_location.source.type | repository |
| best_oa_location.source.is_oa | True |
| best_oa_location.source.issn_l | |
| best_oa_location.source.is_core | False |
| best_oa_location.source.is_in_doaj | False |
| best_oa_location.source.display_name | arXiv (Cornell University) |
| best_oa_location.source.host_organization | https://openalex.org/I205783295 |
| best_oa_location.source.host_organization_name | Cornell University |
| best_oa_location.source.host_organization_lineage | https://openalex.org/I205783295 |
| best_oa_location.license | |
| best_oa_location.pdf_url | https://arxiv.org/pdf/2306.09479 |
| best_oa_location.version | submittedVersion |
| best_oa_location.raw_type | text |
| best_oa_location.license_id | |
| best_oa_location.is_accepted | False |
| best_oa_location.is_published | False |
| best_oa_location.raw_source_name | |
| best_oa_location.landing_page_url | http://arxiv.org/abs/2306.09479 |
| primary_location.id | pmh:oai:arXiv.org:2306.09479 |
| primary_location.is_oa | True |
| primary_location.source.id | https://openalex.org/S4306400194 |
| primary_location.source.issn | |
| primary_location.source.type | repository |
| primary_location.source.is_oa | True |
| primary_location.source.issn_l | |
| primary_location.source.is_core | False |
| primary_location.source.is_in_doaj | False |
| primary_location.source.display_name | arXiv (Cornell University) |
| primary_location.source.host_organization | https://openalex.org/I205783295 |
| primary_location.source.host_organization_name | Cornell University |
| primary_location.source.host_organization_lineage | https://openalex.org/I205783295 |
| primary_location.license | |
| primary_location.pdf_url | https://arxiv.org/pdf/2306.09479 |
| primary_location.version | submittedVersion |
| primary_location.raw_type | text |
| primary_location.license_id | |
| primary_location.is_accepted | False |
| primary_location.is_published | False |
| primary_location.raw_source_name | |
| primary_location.landing_page_url | http://arxiv.org/abs/2306.09479 |
| publication_date | 2023-06-15 |
| publication_year | 2023 |
| referenced_works_count | 0 |
| abstract_inverted_index.a | 69, 77 |
| abstract_inverted_index.11 | 64 |
| abstract_inverted_index.We | 56, 149 |
| abstract_inverted_index.an | 124, 178 |
| abstract_inverted_index.at | 154, 189 |
| abstract_inverted_index.by | 67 |
| abstract_inverted_index.go | 225 |
| abstract_inverted_index.in | 50, 91, 117 |
| abstract_inverted_index.of | 60, 83, 99, 114, 146, 161, 171, 193 |
| abstract_inverted_index.on | 1, 63 |
| abstract_inverted_index.or | 39 |
| abstract_inverted_index.to | 14, 48, 104, 156, 216, 224 |
| abstract_inverted_index.we | 27, 94 |
| abstract_inverted_index.(i) | 102 |
| abstract_inverted_index.LMs | 34, 129 |
| abstract_inverted_index.Our | 164 |
| abstract_inverted_index.and | 24, 54, 139, 173, 218, 229 |
| abstract_inverted_index.are | 186, 205 |
| abstract_inverted_index.but | 142 |
| abstract_inverted_index.due | 47 |
| abstract_inverted_index.for | 30, 158, 207, 231 |
| abstract_inverted_index.has | 4 |
| abstract_inverted_index.may | 35, 213 |
| abstract_inverted_index.not | 214 |
| abstract_inverted_index.on, | 132 |
| abstract_inverted_index.our | 200 |
| abstract_inverted_index.the | 31, 51, 72, 84, 92, 118, 135, 147, 151, 169, 191, 227 |
| abstract_inverted_index.(ii) | 112 |
| abstract_inverted_index.(iv) | 140 |
| abstract_inverted_index.Work | 0 |
| abstract_inverted_index.data | 228 |
| abstract_inverted_index.easy | 125 |
| abstract_inverted_index.four | 96 |
| abstract_inverted_index.have | 166 |
| abstract_inverted_index.into | 226 |
| abstract_inverted_index.laws | 3 |
| abstract_inverted_index.lead | 215 |
| abstract_inverted_index.less | 187 |
| abstract_inverted_index.loss | 16 |
| abstract_inverted_index.more | 220 |
| abstract_inverted_index.over | 108 |
| abstract_inverted_index.real | 137 |
| abstract_inverted_index.show | 11, 36 |
| abstract_inverted_index.task | 41, 127 |
| abstract_inverted_index.than | 134, 196 |
| abstract_inverted_index.that | 6, 33, 183, 203, 219 |
| abstract_inverted_index.with | 17, 43, 76, 87 |
| abstract_inverted_index.(LMs) | 10 |
| abstract_inverted_index.(iii) | 121 |
| abstract_inverted_index.Here, | 26 |
| abstract_inverted_index.allow | 157 |
| abstract_inverted_index.alone | 212 |
| abstract_inverted_index.along | 86 |
| abstract_inverted_index.claim | 32 |
| abstract_inverted_index.could | 130 |
| abstract_inverted_index.data, | 23, 120 |
| abstract_inverted_index.data. | 55 |
| abstract_inverted_index.drive | 168 |
| abstract_inverted_index.e.g., | 46 |
| abstract_inverted_index.flaws | 49 |
| abstract_inverted_index.focus | 131 |
| abstract_inverted_index.found | 5, 90 |
| abstract_inverted_index.large | 7 |
| abstract_inverted_index.model | 210 |
| abstract_inverted_index.needs | 223 |
| abstract_inverted_index.other | 88 |
| abstract_inverted_index.pool. | 80 |
| abstract_inverted_index.prize | 79 |
| abstract_inverted_index.scale | 19, 211 |
| abstract_inverted_index.size, | 21 |
| abstract_inverted_index.task, | 138 |
| abstract_inverted_index.task. | 148 |
| abstract_inverted_index.tasks | 122, 165, 206 |
| abstract_inverted_index.there | 204 |
| abstract_inverted_index.trend | 180 |
| abstract_inverted_index.where | 177 |
| abstract_inverted_index.which | 128, 208 |
| abstract_inverted_index.worse | 40 |
| abstract_inverted_index.(model | 20 |
| abstract_inverted_index.Prize, | 75 |
| abstract_inverted_index.causes | 98 |
| abstract_inverted_index.harder | 136 |
| abstract_inverted_index.helped | 167 |
| abstract_inverted_index.models | 9, 195 |
| abstract_inverted_index.public | 70 |
| abstract_inverted_index.rather | 133 |
| abstract_inverted_index.repeat | 105 |
| abstract_inverted_index.scale, | 45 |
| abstract_inverted_index.trends | 185 |
| abstract_inverted_index.Inverse | 73 |
| abstract_inverted_index.Scaling | 74 |
| abstract_inverted_index.Through | 81 |
| abstract_inverted_index.careful | 221 |
| abstract_inverted_index.correct | 141 |
| abstract_inverted_index.further | 159 |
| abstract_inverted_index.initial | 179 |
| abstract_inverted_index.inverse | 37, 61, 100, 162 |
| abstract_inverted_index.models. | 234 |
| abstract_inverted_index.overall | 15 |
| abstract_inverted_index.present | 28, 57 |
| abstract_inverted_index.release | 150 |
| abstract_inverted_index.results | 201 |
| abstract_inverted_index.running | 68 |
| abstract_inverted_index.scaling | 2, 62, 175, 184 |
| abstract_inverted_index.suggest | 202 |
| abstract_inverted_index.thought | 222 |
| abstract_inverted_index.trends, | 176 |
| abstract_inverted_index.winning | 152 |
| abstract_inverted_index.Overall, | 199 |
| abstract_inverted_index.U-shaped | 172 |
| abstract_inverted_index.analysis | 82 |
| abstract_inverted_index.behavior | 192 |
| abstract_inverted_index.contest, | 71 |
| abstract_inverted_index.datasets | 65, 153 |
| abstract_inverted_index.evidence | 29, 59 |
| abstract_inverted_index.examples | 89 |
| abstract_inverted_index.few-shot | 144 |
| abstract_inverted_index.identify | 95 |
| abstract_inverted_index.language | 8, 233 |
| abstract_inverted_index.patterns | 116 |
| abstract_inverted_index.reliable | 188 |
| abstract_inverted_index.scaling, | 38 |
| abstract_inverted_index.scaling. | 163 |
| abstract_inverted_index.scaling: | 101 |
| abstract_inverted_index.training | 22, 52, 119, 232 |
| abstract_inverted_index.collected | 66 |
| abstract_inverted_index.compute). | 25 |
| abstract_inverted_index.datasets, | 85 |
| abstract_inverted_index.discovery | 170 |
| abstract_inverted_index.empirical | 58 |
| abstract_inverted_index.following | 109 |
| abstract_inverted_index.imitation | 113 |
| abstract_inverted_index.increased | 18, 44, 209 |
| abstract_inverted_index.memorized | 106 |
| abstract_inverted_index.objective | 53 |
| abstract_inverted_index.potential | 97 |
| abstract_inverted_index.progress, | 217 |
| abstract_inverted_index.reverses, | 181 |
| abstract_inverted_index.sequences | 107 |
| abstract_inverted_index.containing | 123 |
| abstract_inverted_index.distractor | 126 |
| abstract_inverted_index.in-context | 110 |
| abstract_inverted_index.inverted-U | 174 |
| abstract_inverted_index.misleading | 143 |
| abstract_inverted_index.objectives | 230 |
| abstract_inverted_index.predicting | 190 |
| abstract_inverted_index.preference | 103 |
| abstract_inverted_index.previously | 197 |
| abstract_inverted_index.suggesting | 182 |
| abstract_inverted_index.literature, | 93 |
| abstract_inverted_index.performance | 42 |
| abstract_inverted_index.predictable | 12 |
| abstract_inverted_index.substantial | 78 |
| abstract_inverted_index.understood. | 198 |
| abstract_inverted_index.undesirable | 115 |
| abstract_inverted_index.improvements | 13 |
| abstract_inverted_index.larger-scale | 194 |
| abstract_inverted_index.instructions, | 111 |
| abstract_inverted_index.investigation | 160 |
| abstract_inverted_index.demonstrations | 145 |
| abstract_inverted_index.https://inversescaling.com/data | 155 |
| cited_by_percentile_year.max | 99 |
| cited_by_percentile_year.min | 95 |
| countries_distinct_count | 0 |
| institutions_distinct_count | 27 |
| sustainable_development_goals[0].id | https://metadata.un.org/sdg/4 |
| sustainable_development_goals[0].score | 0.5699999928474426 |
| sustainable_development_goals[0].display_name | Quality Education |
| citation_normalized_percentile.value | 0.95679144 |
| citation_normalized_percentile.is_in_top_1_percent | False |
| citation_normalized_percentile.is_in_top_10_percent | True |