ProtoReasoning: Prototypes as the Foundation for Generalizable Reasoning in LLMs Article Swipe
YOU?
·
· 2025
· Open Access
·
· DOI: https://doi.org/10.48550/arxiv.2506.15211
Recent advances in Large Reasoning Models (LRMs) trained with Long Chain-of-Thought (Long CoT) reasoning have demonstrated remarkable cross-domain generalization capabilities. However, the underlying mechanisms supporting such transfer remain poorly understood. We hypothesize that cross-domain generalization arises from shared abstract reasoning prototypes -- fundamental reasoning patterns that capture the essence of problems across domains. These prototypes minimize the nuances of the representation, revealing that seemingly diverse tasks are grounded in shared reasoning structures.Based on this hypothesis, we propose ProtoReasoning, a framework that enhances the reasoning ability of LLMs by leveraging scalable and verifiable prototypical representations (Prolog for logical reasoning, PDDL for planning).ProtoReasoning features: (1) an automated prototype construction pipeline that transforms problems into corresponding prototype representations; (2) a comprehensive verification system providing reliable feedback through Prolog/PDDL interpreters; (3) the scalability to synthesize problems arbitrarily within prototype space while ensuring correctness. Extensive experiments show that ProtoReasoning achieves 4.7% improvement over baseline models on logical reasoning (Enigmata-Eval), 6.3% improvement on planning tasks, 4.0% improvement on general reasoning (MMLU) and 1.0% on mathematics (AIME24). Significantly, our ablation studies confirm that learning in prototype space also demonstrates enhanced generalization to structurally similar problems compared to training solely on natural language representations, validating our hypothesis that reasoning prototypes serve as the foundation for generalizable reasoning in large language models.
Related Topics
- Type
- preprint
- Language
- en
- Landing Page
- http://arxiv.org/abs/2506.15211
- https://arxiv.org/pdf/2506.15211
- OA Status
- green
- OpenAlex ID
- https://openalex.org/W4415333566
Raw OpenAlex JSON
- OpenAlex ID
-
https://openalex.org/W4415333566Canonical identifier for this work in OpenAlex
- DOI
-
https://doi.org/10.48550/arxiv.2506.15211Digital Object Identifier
- Title
-
ProtoReasoning: Prototypes as the Foundation for Generalizable Reasoning in LLMsWork title
- Type
-
preprintOpenAlex work type
- Language
-
enPrimary language
- Publication year
-
2025Year of publication
- Publication date
-
2025-06-18Full publication date if available
- Authors
-
He Feng, Zijun Chen, Xiaoyi Liang, Ma Tingting, Yicheng Qiu, Shuangzhi Wu, Junchi YanList of authors in order
- Landing page
-
https://arxiv.org/abs/2506.15211Publisher landing page
- PDF URL
-
https://arxiv.org/pdf/2506.15211Direct 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/2506.15211Direct OA link when available
- Cited by
-
0Total citation count in OpenAlex
Full payload
| id | https://openalex.org/W4415333566 |
|---|---|
| doi | https://doi.org/10.48550/arxiv.2506.15211 |
| ids.doi | https://doi.org/10.48550/arxiv.2506.15211 |
| ids.openalex | https://openalex.org/W4415333566 |
| fwci | 0.0 |
| type | preprint |
| title | ProtoReasoning: Prototypes as the Foundation for Generalizable Reasoning in LLMs |
| biblio.issue | |
| biblio.volume | |
| biblio.last_page | |
| biblio.first_page | |
| topics[0].id | https://openalex.org/T10181 |
| topics[0].field.id | https://openalex.org/fields/17 |
| topics[0].field.display_name | Computer Science |
| topics[0].score | 0.9697999954223633 |
| 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 | Natural Language Processing Techniques |
| topics[1].id | https://openalex.org/T10215 |
| topics[1].field.id | https://openalex.org/fields/17 |
| topics[1].field.display_name | Computer Science |
| topics[1].score | 0.96670001745224 |
| 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 | Semantic Web and Ontologies |
| is_xpac | False |
| apc_list | |
| apc_paid | |
| language | en |
| locations[0].id | pmh:oai:arXiv.org:2506.15211 |
| 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/2506.15211 |
| 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/2506.15211 |
| locations[1].id | doi:10.48550/arxiv.2506.15211 |
| 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 | |
| locations[1].pdf_url | |
| locations[1].version | |
| locations[1].raw_type | article |
| locations[1].license_id | |
| 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.2506.15211 |
| indexed_in | arxiv, datacite |
| authorships[0].author.id | https://openalex.org/A5041475984 |
| authorships[0].author.orcid | |
| authorships[0].author.display_name | He Feng |
| authorships[0].author_position | first |
| authorships[0].raw_author_name | He, Feng |
| authorships[0].is_corresponding | False |
| authorships[1].author.id | https://openalex.org/A5101609780 |
| authorships[1].author.orcid | https://orcid.org/0000-0001-8642-1604 |
| authorships[1].author.display_name | Zijun Chen |
| authorships[1].author_position | middle |
| authorships[1].raw_author_name | Chen, Zijun |
| authorships[1].is_corresponding | False |
| authorships[2].author.id | https://openalex.org/A5102577010 |
| authorships[2].author.orcid | |
| authorships[2].author.display_name | Xiaoyi Liang |
| authorships[2].author_position | middle |
| authorships[2].raw_author_name | Liang, Xinnian |
| authorships[2].is_corresponding | False |
| authorships[3].author.id | https://openalex.org/A5101020623 |
| authorships[3].author.orcid | |
| authorships[3].author.display_name | Ma Tingting |
| authorships[3].author_position | middle |
| authorships[3].raw_author_name | Ma, Tingting |
| authorships[3].is_corresponding | False |
| authorships[4].author.id | https://openalex.org/A5067796056 |
| authorships[4].author.orcid | https://orcid.org/0009-0006-8841-8648 |
| authorships[4].author.display_name | Yicheng Qiu |
| authorships[4].author_position | middle |
| authorships[4].raw_author_name | Qiu, Yunqi |
| authorships[4].is_corresponding | False |
| authorships[5].author.id | https://openalex.org/A5091839641 |
| authorships[5].author.orcid | https://orcid.org/0000-0002-6371-9720 |
| authorships[5].author.display_name | Shuangzhi Wu |
| authorships[5].author_position | middle |
| authorships[5].raw_author_name | Wu, Shuangzhi |
| authorships[5].is_corresponding | False |
| authorships[6].author.id | https://openalex.org/A5087158377 |
| authorships[6].author.orcid | https://orcid.org/0000-0001-9639-7679 |
| authorships[6].author.display_name | Junchi Yan |
| authorships[6].author_position | last |
| authorships[6].raw_author_name | Yan, Junchi |
| authorships[6].is_corresponding | False |
| has_content.pdf | True |
| has_content.grobid_xml | True |
| is_paratext | False |
| open_access.is_oa | True |
| open_access.oa_url | https://arxiv.org/pdf/2506.15211 |
| open_access.oa_status | green |
| open_access.any_repository_has_fulltext | False |
| created_date | 2025-10-19T00:00:00 |
| display_name | ProtoReasoning: Prototypes as the Foundation for Generalizable Reasoning in LLMs |
| has_fulltext | False |
| is_retracted | False |
| updated_date | 2025-11-06T06:51:31.235846 |
| primary_topic.id | https://openalex.org/T10181 |
| primary_topic.field.id | https://openalex.org/fields/17 |
| primary_topic.field.display_name | Computer Science |
| primary_topic.score | 0.9697999954223633 |
| 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 | Natural Language Processing Techniques |
| cited_by_count | 0 |
| locations_count | 2 |
| best_oa_location.id | pmh:oai:arXiv.org:2506.15211 |
| 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/2506.15211 |
| 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/2506.15211 |
| primary_location.id | pmh:oai:arXiv.org:2506.15211 |
| 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/2506.15211 |
| 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/2506.15211 |
| publication_date | 2025-06-18 |
| publication_year | 2025 |
| referenced_works_count | 0 |
| abstract_inverted_index.a | 78, 116 |
| abstract_inverted_index.-- | 41 |
| abstract_inverted_index.We | 30 |
| abstract_inverted_index.an | 103 |
| abstract_inverted_index.as | 203 |
| abstract_inverted_index.by | 87 |
| abstract_inverted_index.in | 2, 68, 177, 209 |
| abstract_inverted_index.of | 49, 58, 85 |
| abstract_inverted_index.on | 72, 150, 156, 161, 167, 192 |
| abstract_inverted_index.to | 129, 184, 189 |
| abstract_inverted_index.we | 75 |
| abstract_inverted_index.(1) | 102 |
| abstract_inverted_index.(2) | 115 |
| abstract_inverted_index.(3) | 126 |
| abstract_inverted_index.and | 90, 165 |
| abstract_inverted_index.are | 66 |
| abstract_inverted_index.for | 95, 99, 206 |
| abstract_inverted_index.our | 171, 197 |
| abstract_inverted_index.the | 21, 47, 56, 59, 82, 127, 204 |
| abstract_inverted_index.1.0% | 166 |
| abstract_inverted_index.4.0% | 159 |
| abstract_inverted_index.4.7% | 145 |
| abstract_inverted_index.6.3% | 154 |
| abstract_inverted_index.CoT) | 12 |
| abstract_inverted_index.LLMs | 86 |
| abstract_inverted_index.Long | 9 |
| abstract_inverted_index.PDDL | 98 |
| abstract_inverted_index.also | 180 |
| abstract_inverted_index.from | 36 |
| abstract_inverted_index.have | 14 |
| abstract_inverted_index.into | 111 |
| abstract_inverted_index.over | 147 |
| abstract_inverted_index.show | 141 |
| abstract_inverted_index.such | 25 |
| abstract_inverted_index.that | 32, 45, 62, 80, 108, 142, 175, 199 |
| abstract_inverted_index.this | 73 |
| abstract_inverted_index.with | 8 |
| abstract_inverted_index.(Long | 11 |
| abstract_inverted_index.Large | 3 |
| abstract_inverted_index.These | 53 |
| abstract_inverted_index.large | 210 |
| abstract_inverted_index.serve | 202 |
| abstract_inverted_index.space | 135, 179 |
| abstract_inverted_index.tasks | 65 |
| abstract_inverted_index.while | 136 |
| abstract_inverted_index.(LRMs) | 6 |
| abstract_inverted_index.(MMLU) | 164 |
| abstract_inverted_index.Models | 5 |
| abstract_inverted_index.Recent | 0 |
| abstract_inverted_index.across | 51 |
| abstract_inverted_index.arises | 35 |
| abstract_inverted_index.models | 149 |
| abstract_inverted_index.poorly | 28 |
| abstract_inverted_index.remain | 27 |
| abstract_inverted_index.shared | 37, 69 |
| abstract_inverted_index.solely | 191 |
| abstract_inverted_index.system | 119 |
| abstract_inverted_index.tasks, | 158 |
| abstract_inverted_index.within | 133 |
| abstract_inverted_index.(Prolog | 94 |
| abstract_inverted_index.ability | 84 |
| abstract_inverted_index.capture | 46 |
| abstract_inverted_index.confirm | 174 |
| abstract_inverted_index.diverse | 64 |
| abstract_inverted_index.essence | 48 |
| abstract_inverted_index.general | 162 |
| abstract_inverted_index.logical | 96, 151 |
| abstract_inverted_index.models. | 212 |
| abstract_inverted_index.natural | 193 |
| abstract_inverted_index.nuances | 57 |
| abstract_inverted_index.propose | 76 |
| abstract_inverted_index.similar | 186 |
| abstract_inverted_index.studies | 173 |
| abstract_inverted_index.through | 123 |
| abstract_inverted_index.trained | 7 |
| abstract_inverted_index.However, | 20 |
| abstract_inverted_index.ablation | 172 |
| abstract_inverted_index.abstract | 38 |
| abstract_inverted_index.achieves | 144 |
| abstract_inverted_index.advances | 1 |
| abstract_inverted_index.baseline | 148 |
| abstract_inverted_index.compared | 188 |
| abstract_inverted_index.domains. | 52 |
| abstract_inverted_index.enhanced | 182 |
| abstract_inverted_index.enhances | 81 |
| abstract_inverted_index.ensuring | 137 |
| abstract_inverted_index.feedback | 122 |
| abstract_inverted_index.grounded | 67 |
| abstract_inverted_index.language | 194, 211 |
| abstract_inverted_index.learning | 176 |
| abstract_inverted_index.minimize | 55 |
| abstract_inverted_index.patterns | 44 |
| abstract_inverted_index.pipeline | 107 |
| abstract_inverted_index.planning | 157 |
| abstract_inverted_index.problems | 50, 110, 131, 187 |
| abstract_inverted_index.reliable | 121 |
| abstract_inverted_index.scalable | 89 |
| abstract_inverted_index.training | 190 |
| abstract_inverted_index.transfer | 26 |
| abstract_inverted_index.(AIME24). | 169 |
| abstract_inverted_index.Extensive | 139 |
| abstract_inverted_index.Reasoning | 4 |
| abstract_inverted_index.automated | 104 |
| abstract_inverted_index.features: | 101 |
| abstract_inverted_index.framework | 79 |
| abstract_inverted_index.prototype | 105, 113, 134, 178 |
| abstract_inverted_index.providing | 120 |
| abstract_inverted_index.reasoning | 13, 39, 43, 70, 83, 152, 163, 200, 208 |
| abstract_inverted_index.revealing | 61 |
| abstract_inverted_index.seemingly | 63 |
| abstract_inverted_index.foundation | 205 |
| abstract_inverted_index.hypothesis | 198 |
| abstract_inverted_index.leveraging | 88 |
| abstract_inverted_index.mechanisms | 23 |
| abstract_inverted_index.prototypes | 40, 54, 201 |
| abstract_inverted_index.reasoning, | 97 |
| abstract_inverted_index.remarkable | 16 |
| abstract_inverted_index.supporting | 24 |
| abstract_inverted_index.synthesize | 130 |
| abstract_inverted_index.transforms | 109 |
| abstract_inverted_index.underlying | 22 |
| abstract_inverted_index.validating | 196 |
| abstract_inverted_index.verifiable | 91 |
| abstract_inverted_index.Prolog/PDDL | 124 |
| abstract_inverted_index.arbitrarily | 132 |
| abstract_inverted_index.experiments | 140 |
| abstract_inverted_index.fundamental | 42 |
| abstract_inverted_index.hypothesis, | 74 |
| abstract_inverted_index.hypothesize | 31 |
| abstract_inverted_index.improvement | 146, 155, 160 |
| abstract_inverted_index.mathematics | 168 |
| abstract_inverted_index.scalability | 128 |
| abstract_inverted_index.understood. | 29 |
| abstract_inverted_index.construction | 106 |
| abstract_inverted_index.correctness. | 138 |
| abstract_inverted_index.cross-domain | 17, 33 |
| abstract_inverted_index.demonstrated | 15 |
| abstract_inverted_index.demonstrates | 181 |
| abstract_inverted_index.prototypical | 92 |
| abstract_inverted_index.structurally | 185 |
| abstract_inverted_index.verification | 118 |
| abstract_inverted_index.capabilities. | 19 |
| abstract_inverted_index.comprehensive | 117 |
| abstract_inverted_index.corresponding | 112 |
| abstract_inverted_index.generalizable | 207 |
| abstract_inverted_index.interpreters; | 125 |
| abstract_inverted_index.ProtoReasoning | 143 |
| abstract_inverted_index.Significantly, | 170 |
| abstract_inverted_index.generalization | 18, 34, 183 |
| abstract_inverted_index.ProtoReasoning, | 77 |
| abstract_inverted_index.representation, | 60 |
| abstract_inverted_index.representations | 93 |
| abstract_inverted_index.(Enigmata-Eval), | 153 |
| abstract_inverted_index.Chain-of-Thought | 10 |
| abstract_inverted_index.representations, | 195 |
| abstract_inverted_index.representations; | 114 |
| abstract_inverted_index.structures.Based | 71 |
| abstract_inverted_index.planning).ProtoReasoning | 100 |
| cited_by_percentile_year | |
| countries_distinct_count | 0 |
| institutions_distinct_count | 7 |
| citation_normalized_percentile |