Full-Step-DPO: Self-Supervised Preference Optimization with Step-wise Rewards for Mathematical Reasoning Article Swipe
YOU?
·
· 2025
· Open Access
·
· DOI: https://doi.org/10.48550/arxiv.2502.14356
Direct Preference Optimization (DPO) often struggles with long-chain mathematical reasoning. Existing approaches, such as Step-DPO, typically improve this by focusing on the first erroneous step in the reasoning chain. However, they overlook all other steps and rely heavily on humans or GPT-4 to identify erroneous steps. To address these issues, we propose Full-Step-DPO, a novel DPO framework tailored for mathematical reasoning. Instead of optimizing only the first erroneous step, it leverages step-wise rewards from the entire reasoning chain. This is achieved by training a self-supervised process reward model, which automatically scores each step, providing rewards while avoiding reliance on external signals. Furthermore, we introduce a novel step-wise DPO loss, which dynamically updates gradients based on these step-wise rewards. This endows stronger reasoning capabilities to language models. Extensive evaluations on both in-domain and out-of-domain mathematical reasoning benchmarks across various base language models, demonstrate that Full-Step-DPO achieves superior performance compared to state-of-the-art baselines.
Related Topics
- Type
- preprint
- Language
- en
- Landing Page
- http://arxiv.org/abs/2502.14356
- https://arxiv.org/pdf/2502.14356
- OA Status
- green
- Related Works
- 10
- OpenAlex ID
- https://openalex.org/W4407806679
Raw OpenAlex JSON
- OpenAlex ID
-
https://openalex.org/W4407806679Canonical identifier for this work in OpenAlex
- DOI
-
https://doi.org/10.48550/arxiv.2502.14356Digital Object Identifier
- Title
-
Full-Step-DPO: Self-Supervised Preference Optimization with Step-wise Rewards for Mathematical ReasoningWork title
- Type
-
preprintOpenAlex work type
- Language
-
enPrimary language
- Publication year
-
2025Year of publication
- Publication date
-
2025-02-20Full publication date if available
- Authors
-
Huimin Xu, Xin Mao, Feng-Lin Li, Xiaobao Wu, Wang Chen, Wei Zhang, Anh Tuan LuuList of authors in order
- Landing page
-
https://arxiv.org/abs/2502.14356Publisher landing page
- PDF URL
-
https://arxiv.org/pdf/2502.14356Direct 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/2502.14356Direct OA link when available
- Concepts
-
Preference, Computer science, Artificial intelligence, Mathematical optimization, Mathematics, StatisticsTop concepts (fields/topics) attached by OpenAlex
- Cited by
-
0Total citation count in OpenAlex
- Related works (count)
-
10Other works algorithmically related by OpenAlex
Full payload
| id | https://openalex.org/W4407806679 |
|---|---|
| doi | https://doi.org/10.48550/arxiv.2502.14356 |
| ids.doi | https://doi.org/10.48550/arxiv.2502.14356 |
| ids.openalex | https://openalex.org/W4407806679 |
| fwci | |
| type | preprint |
| title | Full-Step-DPO: Self-Supervised Preference Optimization with Step-wise Rewards for Mathematical Reasoning |
| biblio.issue | |
| biblio.volume | |
| biblio.last_page | |
| biblio.first_page | |
| topics[0].id | https://openalex.org/T11106 |
| topics[0].field.id | https://openalex.org/fields/17 |
| topics[0].field.display_name | Computer Science |
| topics[0].score | 0.9657999873161316 |
| topics[0].domain.id | https://openalex.org/domains/3 |
| topics[0].domain.display_name | Physical Sciences |
| topics[0].subfield.id | https://openalex.org/subfields/1711 |
| topics[0].subfield.display_name | Signal Processing |
| topics[0].display_name | Data Management and Algorithms |
| topics[1].id | https://openalex.org/T11596 |
| topics[1].field.id | https://openalex.org/fields/17 |
| topics[1].field.display_name | Computer Science |
| topics[1].score | 0.9341999888420105 |
| topics[1].domain.id | https://openalex.org/domains/3 |
| topics[1].domain.display_name | Physical Sciences |
| topics[1].subfield.id | https://openalex.org/subfields/1705 |
| topics[1].subfield.display_name | Computer Networks and Communications |
| topics[1].display_name | Constraint Satisfaction and Optimization |
| topics[2].id | https://openalex.org/T10820 |
| topics[2].field.id | https://openalex.org/fields/17 |
| topics[2].field.display_name | Computer Science |
| topics[2].score | 0.9122999906539917 |
| 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 | Fuzzy Logic and Control Systems |
| is_xpac | False |
| apc_list | |
| apc_paid | |
| concepts[0].id | https://openalex.org/C2781249084 |
| concepts[0].level | 2 |
| concepts[0].score | 0.7135095596313477 |
| concepts[0].wikidata | https://www.wikidata.org/wiki/Q908656 |
| concepts[0].display_name | Preference |
| concepts[1].id | https://openalex.org/C41008148 |
| concepts[1].level | 0 |
| concepts[1].score | 0.49958348274230957 |
| concepts[1].wikidata | https://www.wikidata.org/wiki/Q21198 |
| concepts[1].display_name | Computer science |
| concepts[2].id | https://openalex.org/C154945302 |
| concepts[2].level | 1 |
| concepts[2].score | 0.49809789657592773 |
| concepts[2].wikidata | https://www.wikidata.org/wiki/Q11660 |
| concepts[2].display_name | Artificial intelligence |
| concepts[3].id | https://openalex.org/C126255220 |
| concepts[3].level | 1 |
| concepts[3].score | 0.34632599353790283 |
| concepts[3].wikidata | https://www.wikidata.org/wiki/Q141495 |
| concepts[3].display_name | Mathematical optimization |
| concepts[4].id | https://openalex.org/C33923547 |
| concepts[4].level | 0 |
| concepts[4].score | 0.32005250453948975 |
| concepts[4].wikidata | https://www.wikidata.org/wiki/Q395 |
| concepts[4].display_name | Mathematics |
| concepts[5].id | https://openalex.org/C105795698 |
| concepts[5].level | 1 |
| concepts[5].score | 0.12946867942810059 |
| concepts[5].wikidata | https://www.wikidata.org/wiki/Q12483 |
| concepts[5].display_name | Statistics |
| keywords[0].id | https://openalex.org/keywords/preference |
| keywords[0].score | 0.7135095596313477 |
| keywords[0].display_name | Preference |
| keywords[1].id | https://openalex.org/keywords/computer-science |
| keywords[1].score | 0.49958348274230957 |
| keywords[1].display_name | Computer science |
| keywords[2].id | https://openalex.org/keywords/artificial-intelligence |
| keywords[2].score | 0.49809789657592773 |
| keywords[2].display_name | Artificial intelligence |
| keywords[3].id | https://openalex.org/keywords/mathematical-optimization |
| keywords[3].score | 0.34632599353790283 |
| keywords[3].display_name | Mathematical optimization |
| keywords[4].id | https://openalex.org/keywords/mathematics |
| keywords[4].score | 0.32005250453948975 |
| keywords[4].display_name | Mathematics |
| keywords[5].id | https://openalex.org/keywords/statistics |
| keywords[5].score | 0.12946867942810059 |
| keywords[5].display_name | Statistics |
| language | en |
| locations[0].id | pmh:oai:arXiv.org:2502.14356 |
| 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/2502.14356 |
| 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/2502.14356 |
| locations[1].id | doi:10.48550/arxiv.2502.14356 |
| 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.2502.14356 |
| indexed_in | arxiv, datacite |
| authorships[0].author.id | https://openalex.org/A5000871321 |
| authorships[0].author.orcid | |
| authorships[0].author.display_name | Huimin Xu |
| authorships[0].author_position | first |
| authorships[0].raw_author_name | Xu, Huimin |
| authorships[0].is_corresponding | False |
| authorships[1].author.id | https://openalex.org/A5102756403 |
| authorships[1].author.orcid | https://orcid.org/0000-0002-3740-8355 |
| authorships[1].author.display_name | Xin Mao |
| authorships[1].author_position | middle |
| authorships[1].raw_author_name | Mao, Xin |
| authorships[1].is_corresponding | False |
| authorships[2].author.id | https://openalex.org/A5101808283 |
| authorships[2].author.orcid | |
| authorships[2].author.display_name | Feng-Lin Li |
| authorships[2].author_position | middle |
| authorships[2].raw_author_name | Li, Feng-Lin |
| authorships[2].is_corresponding | False |
| authorships[3].author.id | https://openalex.org/A5011376608 |
| authorships[3].author.orcid | https://orcid.org/0000-0003-0076-3924 |
| authorships[3].author.display_name | Xiaobao Wu |
| authorships[3].author_position | middle |
| authorships[3].raw_author_name | Wu, Xiaobao |
| authorships[3].is_corresponding | False |
| authorships[4].author.id | https://openalex.org/A5100340740 |
| authorships[4].author.orcid | https://orcid.org/0000-0001-7892-3575 |
| authorships[4].author.display_name | Wang Chen |
| authorships[4].author_position | middle |
| authorships[4].raw_author_name | Chen, Wang |
| authorships[4].is_corresponding | False |
| authorships[5].author.id | https://openalex.org/A5101555891 |
| authorships[5].author.orcid | https://orcid.org/0009-0001-5847-663X |
| authorships[5].author.display_name | Wei Zhang |
| authorships[5].author_position | middle |
| authorships[5].raw_author_name | Zhang, Wei |
| authorships[5].is_corresponding | False |
| authorships[6].author.id | https://openalex.org/A5050386762 |
| authorships[6].author.orcid | https://orcid.org/0000-0002-1927-9895 |
| authorships[6].author.display_name | Anh Tuan Luu |
| authorships[6].author_position | last |
| authorships[6].raw_author_name | Luu, Anh Tuan |
| authorships[6].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/2502.14356 |
| open_access.oa_status | green |
| open_access.any_repository_has_fulltext | False |
| created_date | 2025-10-10T00:00:00 |
| display_name | Full-Step-DPO: Self-Supervised Preference Optimization with Step-wise Rewards for Mathematical Reasoning |
| has_fulltext | False |
| is_retracted | False |
| updated_date | 2025-11-06T06:51:31.235846 |
| primary_topic.id | https://openalex.org/T11106 |
| primary_topic.field.id | https://openalex.org/fields/17 |
| primary_topic.field.display_name | Computer Science |
| primary_topic.score | 0.9657999873161316 |
| primary_topic.domain.id | https://openalex.org/domains/3 |
| primary_topic.domain.display_name | Physical Sciences |
| primary_topic.subfield.id | https://openalex.org/subfields/1711 |
| primary_topic.subfield.display_name | Signal Processing |
| primary_topic.display_name | Data Management and Algorithms |
| related_works | https://openalex.org/W2899084033, https://openalex.org/W2748952813, https://openalex.org/W4391375266, https://openalex.org/W1979597421, https://openalex.org/W2007980826, https://openalex.org/W2061531152, https://openalex.org/W3002753104, https://openalex.org/W2077600819, https://openalex.org/W2142036596, https://openalex.org/W2072657027 |
| cited_by_count | 0 |
| locations_count | 2 |
| best_oa_location.id | pmh:oai:arXiv.org:2502.14356 |
| 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/2502.14356 |
| 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/2502.14356 |
| primary_location.id | pmh:oai:arXiv.org:2502.14356 |
| 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/2502.14356 |
| 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/2502.14356 |
| publication_date | 2025-02-20 |
| publication_year | 2025 |
| referenced_works_count | 0 |
| abstract_inverted_index.a | 53, 83, 104 |
| abstract_inverted_index.To | 46 |
| abstract_inverted_index.as | 13 |
| abstract_inverted_index.by | 18, 81 |
| abstract_inverted_index.in | 25 |
| abstract_inverted_index.is | 79 |
| abstract_inverted_index.it | 69 |
| abstract_inverted_index.of | 62 |
| abstract_inverted_index.on | 20, 38, 98, 114, 128 |
| abstract_inverted_index.or | 40 |
| abstract_inverted_index.to | 42, 123, 148 |
| abstract_inverted_index.we | 50, 102 |
| abstract_inverted_index.DPO | 55, 107 |
| abstract_inverted_index.all | 32 |
| abstract_inverted_index.and | 35, 131 |
| abstract_inverted_index.for | 58 |
| abstract_inverted_index.the | 21, 26, 65, 74 |
| abstract_inverted_index.This | 78, 118 |
| abstract_inverted_index.base | 138 |
| abstract_inverted_index.both | 129 |
| abstract_inverted_index.each | 91 |
| abstract_inverted_index.from | 73 |
| abstract_inverted_index.only | 64 |
| abstract_inverted_index.rely | 36 |
| abstract_inverted_index.step | 24 |
| abstract_inverted_index.such | 12 |
| abstract_inverted_index.that | 142 |
| abstract_inverted_index.they | 30 |
| abstract_inverted_index.this | 17 |
| abstract_inverted_index.with | 6 |
| abstract_inverted_index.(DPO) | 3 |
| abstract_inverted_index.GPT-4 | 41 |
| abstract_inverted_index.based | 113 |
| abstract_inverted_index.first | 22, 66 |
| abstract_inverted_index.loss, | 108 |
| abstract_inverted_index.novel | 54, 105 |
| abstract_inverted_index.often | 4 |
| abstract_inverted_index.other | 33 |
| abstract_inverted_index.step, | 68, 92 |
| abstract_inverted_index.steps | 34 |
| abstract_inverted_index.these | 48, 115 |
| abstract_inverted_index.which | 88, 109 |
| abstract_inverted_index.while | 95 |
| abstract_inverted_index.Direct | 0 |
| abstract_inverted_index.across | 136 |
| abstract_inverted_index.chain. | 28, 77 |
| abstract_inverted_index.endows | 119 |
| abstract_inverted_index.entire | 75 |
| abstract_inverted_index.humans | 39 |
| abstract_inverted_index.model, | 87 |
| abstract_inverted_index.reward | 86 |
| abstract_inverted_index.scores | 90 |
| abstract_inverted_index.steps. | 45 |
| abstract_inverted_index.Instead | 61 |
| abstract_inverted_index.address | 47 |
| abstract_inverted_index.heavily | 37 |
| abstract_inverted_index.improve | 16 |
| abstract_inverted_index.issues, | 49 |
| abstract_inverted_index.models, | 140 |
| abstract_inverted_index.models. | 125 |
| abstract_inverted_index.process | 85 |
| abstract_inverted_index.propose | 51 |
| abstract_inverted_index.rewards | 72, 94 |
| abstract_inverted_index.updates | 111 |
| abstract_inverted_index.various | 137 |
| abstract_inverted_index.Existing | 10 |
| abstract_inverted_index.However, | 29 |
| abstract_inverted_index.achieved | 80 |
| abstract_inverted_index.achieves | 144 |
| abstract_inverted_index.avoiding | 96 |
| abstract_inverted_index.compared | 147 |
| abstract_inverted_index.external | 99 |
| abstract_inverted_index.focusing | 19 |
| abstract_inverted_index.identify | 43 |
| abstract_inverted_index.language | 124, 139 |
| abstract_inverted_index.overlook | 31 |
| abstract_inverted_index.reliance | 97 |
| abstract_inverted_index.rewards. | 117 |
| abstract_inverted_index.signals. | 100 |
| abstract_inverted_index.stronger | 120 |
| abstract_inverted_index.superior | 145 |
| abstract_inverted_index.tailored | 57 |
| abstract_inverted_index.training | 82 |
| abstract_inverted_index.Extensive | 126 |
| abstract_inverted_index.Step-DPO, | 14 |
| abstract_inverted_index.erroneous | 23, 44, 67 |
| abstract_inverted_index.framework | 56 |
| abstract_inverted_index.gradients | 112 |
| abstract_inverted_index.in-domain | 130 |
| abstract_inverted_index.introduce | 103 |
| abstract_inverted_index.leverages | 70 |
| abstract_inverted_index.providing | 93 |
| abstract_inverted_index.reasoning | 27, 76, 121, 134 |
| abstract_inverted_index.step-wise | 71, 106, 116 |
| abstract_inverted_index.struggles | 5 |
| abstract_inverted_index.typically | 15 |
| abstract_inverted_index.Preference | 1 |
| abstract_inverted_index.baselines. | 150 |
| abstract_inverted_index.benchmarks | 135 |
| abstract_inverted_index.long-chain | 7 |
| abstract_inverted_index.optimizing | 63 |
| abstract_inverted_index.reasoning. | 9, 60 |
| abstract_inverted_index.approaches, | 11 |
| abstract_inverted_index.demonstrate | 141 |
| abstract_inverted_index.dynamically | 110 |
| abstract_inverted_index.evaluations | 127 |
| abstract_inverted_index.performance | 146 |
| abstract_inverted_index.Furthermore, | 101 |
| abstract_inverted_index.Optimization | 2 |
| abstract_inverted_index.capabilities | 122 |
| abstract_inverted_index.mathematical | 8, 59, 133 |
| abstract_inverted_index.Full-Step-DPO | 143 |
| abstract_inverted_index.automatically | 89 |
| abstract_inverted_index.out-of-domain | 132 |
| abstract_inverted_index.Full-Step-DPO, | 52 |
| abstract_inverted_index.self-supervised | 84 |
| abstract_inverted_index.state-of-the-art | 149 |
| cited_by_percentile_year | |
| countries_distinct_count | 0 |
| institutions_distinct_count | 7 |
| citation_normalized_percentile |