Self-Discover: Large Language Models Self-Compose Reasoning Structures Article Swipe
YOU?
·
· 2024
· Open Access
·
· DOI: https://doi.org/10.48550/arxiv.2402.03620
We introduce SELF-DISCOVER, a general framework for LLMs to self-discover the task-intrinsic reasoning structures to tackle complex reasoning problems that are challenging for typical prompting methods. Core to the framework is a self-discovery process where LLMs select multiple atomic reasoning modules such as critical thinking and step-by-step thinking, and compose them into an explicit reasoning structure for LLMs to follow during decoding. SELF-DISCOVER substantially improves GPT-4 and PaLM 2's performance on challenging reasoning benchmarks such as BigBench-Hard, grounded agent reasoning, and MATH, by as much as 32% compared to Chain of Thought (CoT). Furthermore, SELF-DISCOVER outperforms inference-intensive methods such as CoT-Self-Consistency by more than 20%, while requiring 10-40x fewer inference compute. Finally, we show that the self-discovered reasoning structures are universally applicable across model families: from PaLM 2-L to GPT-4, and from GPT-4 to Llama2, and share commonalities with human reasoning patterns.
Related Topics
- Type
- preprint
- Language
- en
- Landing Page
- http://arxiv.org/abs/2402.03620
- https://arxiv.org/pdf/2402.03620
- OA Status
- green
- Cited By
- 9
- Related Works
- 10
- OpenAlex ID
- https://openalex.org/W4391631614
Raw OpenAlex JSON
- OpenAlex ID
-
https://openalex.org/W4391631614Canonical identifier for this work in OpenAlex
- DOI
-
https://doi.org/10.48550/arxiv.2402.03620Digital Object Identifier
- Title
-
Self-Discover: Large Language Models Self-Compose Reasoning StructuresWork title
- Type
-
preprintOpenAlex work type
- Language
-
enPrimary language
- Publication year
-
2024Year of publication
- Publication date
-
2024-02-06Full publication date if available
- Authors
-
Pei Zhou, Jay Pujara, Xiang Ren, Xinyun Chen, Heng-Tze Cheng, Quoc V. Le, Ed H., Denny Zhou, Swaroop Mishra, Huaixiu ZhengList of authors in order
- Landing page
-
https://arxiv.org/abs/2402.03620Publisher landing page
- PDF URL
-
https://arxiv.org/pdf/2402.03620Direct 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/2402.03620Direct OA link when available
- Concepts
-
Computer science, Natural language processing, Artificial intelligence, Linguistics, Cognitive science, Psychology, PhilosophyTop concepts (fields/topics) attached by OpenAlex
- Cited by
-
9Total citation count in OpenAlex
- Citations by year (recent)
-
2025: 5, 2024: 2, 2023: 2Per-year citation counts (last 5 years)
- Related works (count)
-
10Other works algorithmically related by OpenAlex
Full payload
| id | https://openalex.org/W4391631614 |
|---|---|
| doi | https://doi.org/10.48550/arxiv.2402.03620 |
| ids.doi | https://doi.org/10.48550/arxiv.2402.03620 |
| ids.openalex | https://openalex.org/W4391631614 |
| fwci | |
| type | preprint |
| title | Self-Discover: Large Language Models Self-Compose Reasoning Structures |
| 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 | 0.9115999937057495 |
| 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 |
| is_xpac | False |
| apc_list | |
| apc_paid | |
| concepts[0].id | https://openalex.org/C41008148 |
| concepts[0].level | 0 |
| concepts[0].score | 0.5543417930603027 |
| concepts[0].wikidata | https://www.wikidata.org/wiki/Q21198 |
| concepts[0].display_name | Computer science |
| concepts[1].id | https://openalex.org/C204321447 |
| concepts[1].level | 1 |
| concepts[1].score | 0.4658079743385315 |
| concepts[1].wikidata | https://www.wikidata.org/wiki/Q30642 |
| concepts[1].display_name | Natural language processing |
| concepts[2].id | https://openalex.org/C154945302 |
| concepts[2].level | 1 |
| concepts[2].score | 0.3966338038444519 |
| concepts[2].wikidata | https://www.wikidata.org/wiki/Q11660 |
| concepts[2].display_name | Artificial intelligence |
| concepts[3].id | https://openalex.org/C41895202 |
| concepts[3].level | 1 |
| concepts[3].score | 0.32760822772979736 |
| concepts[3].wikidata | https://www.wikidata.org/wiki/Q8162 |
| concepts[3].display_name | Linguistics |
| concepts[4].id | https://openalex.org/C188147891 |
| concepts[4].level | 1 |
| concepts[4].score | 0.32304781675338745 |
| concepts[4].wikidata | https://www.wikidata.org/wiki/Q147638 |
| concepts[4].display_name | Cognitive science |
| concepts[5].id | https://openalex.org/C15744967 |
| concepts[5].level | 0 |
| concepts[5].score | 0.19602042436599731 |
| concepts[5].wikidata | https://www.wikidata.org/wiki/Q9418 |
| concepts[5].display_name | Psychology |
| concepts[6].id | https://openalex.org/C138885662 |
| concepts[6].level | 0 |
| concepts[6].score | 0.08266562223434448 |
| concepts[6].wikidata | https://www.wikidata.org/wiki/Q5891 |
| concepts[6].display_name | Philosophy |
| keywords[0].id | https://openalex.org/keywords/computer-science |
| keywords[0].score | 0.5543417930603027 |
| keywords[0].display_name | Computer science |
| keywords[1].id | https://openalex.org/keywords/natural-language-processing |
| keywords[1].score | 0.4658079743385315 |
| keywords[1].display_name | Natural language processing |
| keywords[2].id | https://openalex.org/keywords/artificial-intelligence |
| keywords[2].score | 0.3966338038444519 |
| keywords[2].display_name | Artificial intelligence |
| keywords[3].id | https://openalex.org/keywords/linguistics |
| keywords[3].score | 0.32760822772979736 |
| keywords[3].display_name | Linguistics |
| keywords[4].id | https://openalex.org/keywords/cognitive-science |
| keywords[4].score | 0.32304781675338745 |
| keywords[4].display_name | Cognitive science |
| keywords[5].id | https://openalex.org/keywords/psychology |
| keywords[5].score | 0.19602042436599731 |
| keywords[5].display_name | Psychology |
| keywords[6].id | https://openalex.org/keywords/philosophy |
| keywords[6].score | 0.08266562223434448 |
| keywords[6].display_name | Philosophy |
| language | en |
| locations[0].id | pmh:oai:arXiv.org:2402.03620 |
| 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/2402.03620 |
| 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/2402.03620 |
| locations[1].id | doi:10.48550/arxiv.2402.03620 |
| 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 |
| 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.2402.03620 |
| indexed_in | arxiv, datacite |
| authorships[0].author.id | https://openalex.org/A5070613964 |
| authorships[0].author.orcid | https://orcid.org/0000-0003-1631-3637 |
| authorships[0].author.display_name | Pei Zhou |
| authorships[0].author_position | first |
| authorships[0].raw_author_name | Zhou, Pei |
| authorships[0].is_corresponding | False |
| authorships[1].author.id | https://openalex.org/A5073443117 |
| authorships[1].author.orcid | https://orcid.org/0000-0001-6921-1744 |
| authorships[1].author.display_name | Jay Pujara |
| authorships[1].author_position | middle |
| authorships[1].raw_author_name | Pujara, Jay |
| authorships[1].is_corresponding | False |
| authorships[2].author.id | https://openalex.org/A5009408707 |
| authorships[2].author.orcid | https://orcid.org/0000-0001-8655-663X |
| authorships[2].author.display_name | Xiang Ren |
| authorships[2].author_position | middle |
| authorships[2].raw_author_name | Ren, Xiang |
| authorships[2].is_corresponding | False |
| authorships[3].author.id | https://openalex.org/A5101977008 |
| authorships[3].author.orcid | https://orcid.org/0009-0001-9700-8951 |
| authorships[3].author.display_name | Xinyun Chen |
| authorships[3].author_position | middle |
| authorships[3].raw_author_name | Chen, Xinyun |
| authorships[3].is_corresponding | False |
| authorships[4].author.id | https://openalex.org/A5054880622 |
| authorships[4].author.orcid | https://orcid.org/0009-0007-3845-8796 |
| authorships[4].author.display_name | Heng-Tze Cheng |
| authorships[4].author_position | middle |
| authorships[4].raw_author_name | Cheng, Heng-Tze |
| authorships[4].is_corresponding | False |
| authorships[5].author.id | https://openalex.org/A5088551093 |
| authorships[5].author.orcid | https://orcid.org/0000-0002-1087-2844 |
| authorships[5].author.display_name | Quoc V. Le |
| authorships[5].author_position | middle |
| authorships[5].raw_author_name | Le, Quoc V. |
| authorships[5].is_corresponding | False |
| authorships[6].author.id | https://openalex.org/A5028125399 |
| authorships[6].author.orcid | https://orcid.org/0000-0003-3230-5338 |
| authorships[6].author.display_name | Ed H. |
| authorships[6].author_position | middle |
| authorships[6].raw_author_name | Chi, Ed H. |
| authorships[6].is_corresponding | False |
| authorships[7].author.id | https://openalex.org/A5061512999 |
| authorships[7].author.orcid | |
| authorships[7].author.display_name | Denny Zhou |
| authorships[7].author_position | middle |
| authorships[7].raw_author_name | Zhou, Denny |
| authorships[7].is_corresponding | False |
| authorships[8].author.id | https://openalex.org/A5063722751 |
| authorships[8].author.orcid | https://orcid.org/0009-0001-6413-7001 |
| authorships[8].author.display_name | Swaroop Mishra |
| authorships[8].author_position | middle |
| authorships[8].raw_author_name | Mishra, Swaroop |
| authorships[8].is_corresponding | False |
| authorships[9].author.id | https://openalex.org/A5003807260 |
| authorships[9].author.orcid | |
| authorships[9].author.display_name | Huaixiu Zheng |
| authorships[9].author_position | last |
| authorships[9].raw_author_name | Zheng, Huaixiu Steven |
| authorships[9].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/2402.03620 |
| open_access.oa_status | green |
| open_access.any_repository_has_fulltext | False |
| created_date | 2025-10-10T00:00:00 |
| display_name | Self-Discover: Large Language Models Self-Compose Reasoning Structures |
| 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 | 0.9115999937057495 |
| 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/W2748952813, https://openalex.org/W2390279801, https://openalex.org/W2358668433, https://openalex.org/W2376932109, https://openalex.org/W2001405890, https://openalex.org/W2382290278, https://openalex.org/W2478288626, https://openalex.org/W2350741829, https://openalex.org/W2530322880, https://openalex.org/W3204019825 |
| cited_by_count | 9 |
| counts_by_year[0].year | 2025 |
| counts_by_year[0].cited_by_count | 5 |
| counts_by_year[1].year | 2024 |
| counts_by_year[1].cited_by_count | 2 |
| counts_by_year[2].year | 2023 |
| counts_by_year[2].cited_by_count | 2 |
| locations_count | 2 |
| best_oa_location.id | pmh:oai:arXiv.org:2402.03620 |
| 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/2402.03620 |
| 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/2402.03620 |
| primary_location.id | pmh:oai:arXiv.org:2402.03620 |
| 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/2402.03620 |
| 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/2402.03620 |
| publication_date | 2024-02-06 |
| publication_year | 2024 |
| referenced_works_count | 0 |
| abstract_inverted_index.a | 3, 31 |
| abstract_inverted_index.We | 0 |
| abstract_inverted_index.an | 52 |
| abstract_inverted_index.as | 42, 75, 83, 85, 99 |
| abstract_inverted_index.by | 82, 101 |
| abstract_inverted_index.is | 30 |
| abstract_inverted_index.of | 90 |
| abstract_inverted_index.on | 70 |
| abstract_inverted_index.to | 8, 14, 27, 58, 88, 128, 133 |
| abstract_inverted_index.we | 112 |
| abstract_inverted_index.2's | 68 |
| abstract_inverted_index.2-L | 127 |
| abstract_inverted_index.32% | 86 |
| abstract_inverted_index.and | 45, 48, 66, 80, 130, 135 |
| abstract_inverted_index.are | 20, 119 |
| abstract_inverted_index.for | 6, 22, 56 |
| abstract_inverted_index.the | 10, 28, 115 |
| abstract_inverted_index.20%, | 104 |
| abstract_inverted_index.Core | 26 |
| abstract_inverted_index.LLMs | 7, 35, 57 |
| abstract_inverted_index.PaLM | 67, 126 |
| abstract_inverted_index.from | 125, 131 |
| abstract_inverted_index.into | 51 |
| abstract_inverted_index.more | 102 |
| abstract_inverted_index.much | 84 |
| abstract_inverted_index.show | 113 |
| abstract_inverted_index.such | 41, 74, 98 |
| abstract_inverted_index.than | 103 |
| abstract_inverted_index.that | 19, 114 |
| abstract_inverted_index.them | 50 |
| abstract_inverted_index.with | 138 |
| abstract_inverted_index.Chain | 89 |
| abstract_inverted_index.GPT-4 | 65, 132 |
| abstract_inverted_index.MATH, | 81 |
| abstract_inverted_index.agent | 78 |
| abstract_inverted_index.fewer | 108 |
| abstract_inverted_index.human | 139 |
| abstract_inverted_index.model | 123 |
| abstract_inverted_index.share | 136 |
| abstract_inverted_index.where | 34 |
| abstract_inverted_index.while | 105 |
| abstract_inverted_index.(CoT). | 92 |
| abstract_inverted_index.10-40x | 107 |
| abstract_inverted_index.GPT-4, | 129 |
| abstract_inverted_index.across | 122 |
| abstract_inverted_index.atomic | 38 |
| abstract_inverted_index.during | 60 |
| abstract_inverted_index.follow | 59 |
| abstract_inverted_index.select | 36 |
| abstract_inverted_index.tackle | 15 |
| abstract_inverted_index.Llama2, | 134 |
| abstract_inverted_index.Thought | 91 |
| abstract_inverted_index.complex | 16 |
| abstract_inverted_index.compose | 49 |
| abstract_inverted_index.general | 4 |
| abstract_inverted_index.methods | 97 |
| abstract_inverted_index.modules | 40 |
| abstract_inverted_index.process | 33 |
| abstract_inverted_index.typical | 23 |
| abstract_inverted_index.Finally, | 111 |
| abstract_inverted_index.compared | 87 |
| abstract_inverted_index.compute. | 110 |
| abstract_inverted_index.critical | 43 |
| abstract_inverted_index.explicit | 53 |
| abstract_inverted_index.grounded | 77 |
| abstract_inverted_index.improves | 64 |
| abstract_inverted_index.methods. | 25 |
| abstract_inverted_index.multiple | 37 |
| abstract_inverted_index.problems | 18 |
| abstract_inverted_index.thinking | 44 |
| abstract_inverted_index.decoding. | 61 |
| abstract_inverted_index.families: | 124 |
| abstract_inverted_index.framework | 5, 29 |
| abstract_inverted_index.inference | 109 |
| abstract_inverted_index.introduce | 1 |
| abstract_inverted_index.patterns. | 141 |
| abstract_inverted_index.prompting | 24 |
| abstract_inverted_index.reasoning | 12, 17, 39, 54, 72, 117, 140 |
| abstract_inverted_index.requiring | 106 |
| abstract_inverted_index.structure | 55 |
| abstract_inverted_index.thinking, | 47 |
| abstract_inverted_index.applicable | 121 |
| abstract_inverted_index.benchmarks | 73 |
| abstract_inverted_index.reasoning, | 79 |
| abstract_inverted_index.structures | 13, 118 |
| abstract_inverted_index.challenging | 21, 71 |
| abstract_inverted_index.outperforms | 95 |
| abstract_inverted_index.performance | 69 |
| abstract_inverted_index.universally | 120 |
| abstract_inverted_index.Furthermore, | 93 |
| abstract_inverted_index.step-by-step | 46 |
| abstract_inverted_index.SELF-DISCOVER | 62, 94 |
| abstract_inverted_index.commonalities | 137 |
| abstract_inverted_index.self-discover | 9 |
| abstract_inverted_index.substantially | 63 |
| abstract_inverted_index.BigBench-Hard, | 76 |
| abstract_inverted_index.SELF-DISCOVER, | 2 |
| abstract_inverted_index.self-discovery | 32 |
| abstract_inverted_index.task-intrinsic | 11 |
| abstract_inverted_index.self-discovered | 116 |
| abstract_inverted_index.inference-intensive | 96 |
| abstract_inverted_index.CoT-Self-Consistency | 100 |
| cited_by_percentile_year | |
| countries_distinct_count | 0 |
| institutions_distinct_count | 10 |
| citation_normalized_percentile |