RLDBF: Enhancing LLMs Via Reinforcement Learning With DataBase FeedBack Article Swipe

PDF

Weichen Dai , Ying Pan , X Li , Xi Li , Yi Zhou , Jiang Wu ·

YOU? · · 2025 · Open Access · · DOI: https://doi.org/10.48550/arxiv.2504.03713

While current large language models (LLMs) demonstrate remarkable linguistic capabilities through training on massive unstructured text corpora, they remain inadequate in leveraging structured scientific data (e.g., chemical molecular properties in databases) that encapsulate centuries of accumulated scientific expertise. These structured datasets hold strategic significance for advancing AI for Science yet current approaches merely treat them as auxiliary supplements to unstructured text. This study pioneers a systematic investigation into enhancing LLMs with structured scientific data, using chemical molecular science as a testbed. We investigate the impact of incorporating molecular property data on LLM across distinct training phases, including continual pre-training, supervised fine-tuning, and reinforcement learning. Notably, to address the inherent limitation of numerical insensitivity in large models, we propose an innovative methodology termed "Reinforcement Learning with Database Feedback" (RLDBF). Experimental evaluations demonstrate the efficacy of the proposed approach, with the model exhibiting remarkable generalization capabilities on previously unseen data and other chemical tasks. The results substantiate the potential of our method in advancing the field of structured scientific data processing within LLMs.

Related Topics

Truth And Reconciliation Commission Of Canada

Alanis Morissette

2025 Nba Draft

28 Years Later

Reich Ministry Of Public Enlightenment And Propaganda

Mahmood Mamdani

Concepts

No concepts available.

Metadata

Type: preprint
Language: en
Landing Page: http://arxiv.org/abs/2504.03713
PDF: https://arxiv.org/pdf/2504.03713
OA Status: green
OpenAlex ID: https://openalex.org/W4416123725

All OpenAlex metadata

Raw OpenAlex JSON

OpenAlex ID: https://openalex.org/W4416123725

Canonical identifier for this work in OpenAlex
DOI: https://doi.org/10.48550/arxiv.2504.03713

Digital Object Identifier
Title: RLDBF: Enhancing LLMs Via Reinforcement Learning With DataBase FeedBack

Work title
Type: preprint

OpenAlex work type
Language: en

Primary language
Publication year: 2025

Year of publication
Publication date: 2025-03-28

Full publication date if available
Authors: Weichen Dai, Ying Pan, X Li, Xi Li, Yi Zhou, Jiang Wu

List of authors in order
Landing page: https://arxiv.org/abs/2504.03713

Publisher landing page
PDF URL: https://arxiv.org/pdf/2504.03713

Direct link to full text PDF
Open access: Yes

Whether a free full text is available
OA status: green

Open access status per OpenAlex
OA URL: https://arxiv.org/pdf/2504.03713

Direct OA link when available
Cited by: 0

Total citation count in OpenAlex

Full payload

id	https://openalex.org/W4416123725
doi	https://doi.org/10.48550/arxiv.2504.03713
ids.doi	https://doi.org/10.48550/arxiv.2504.03713
ids.openalex	https://openalex.org/W4416123725
fwci
type	preprint
title	RLDBF: Enhancing LLMs Via Reinforcement Learning With DataBase FeedBack
biblio.issue
biblio.volume
biblio.last_page
biblio.first_page
is_xpac	False
apc_list
apc_paid
language	en
locations[0].id	pmh:oai:arXiv.org:2504.03713
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/2504.03713
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/2504.03713
locations[1].id	doi:10.48550/arxiv.2504.03713
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.2504.03713
indexed_in	arxiv, datacite
authorships[0].author.id	https://openalex.org/A5042276536
authorships[0].author.orcid	https://orcid.org/0000-0002-5019-2755
authorships[0].author.display_name	Weichen Dai
authorships[0].author_position	first
authorships[0].raw_author_name	Dai, Weichen
authorships[0].is_corresponding	False
authorships[1].author.id	https://openalex.org/A5060891066
authorships[1].author.orcid	https://orcid.org/0000-0002-8203-3943
authorships[1].author.display_name	Ying Pan
authorships[1].author_position	middle
authorships[1].raw_author_name	Pan, Yixuan
authorships[1].is_corresponding	False
authorships[2].author.id	https://openalex.org/A5009258949
authorships[2].author.orcid	https://orcid.org/0000-0001-6709-0723
authorships[2].author.display_name	X Li
authorships[2].author_position	middle
authorships[2].raw_author_name	Li, Xinhe
authorships[2].is_corresponding	False
authorships[3].author.id	https://openalex.org/A5100407749
authorships[3].author.orcid	https://orcid.org/0009-0003-7871-5401
authorships[3].author.display_name	Xi Li
authorships[3].author_position	last
authorships[3].raw_author_name	Li, Xi
authorships[3].is_corresponding	False
authorships[4].author.id	https://openalex.org/A5037840405
authorships[4].author.orcid	https://orcid.org/0000-0002-4774-2234
authorships[4].author.display_name	Yi Zhou
authorships[4].author_position	middle
authorships[4].raw_author_name	Zhou, Yi
authorships[4].is_corresponding	False
authorships[5].author.id	https://openalex.org/A5065134331
authorships[5].author.orcid	https://orcid.org/0000-0003-4628-7660
authorships[5].author.display_name	Jiang Wu
authorships[5].author_position	middle
authorships[5].raw_author_name	Jiang, Wu
authorships[5].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/2504.03713
open_access.oa_status	green
open_access.any_repository_has_fulltext	False
created_date	2025-10-10T00:00:00
display_name	RLDBF: Enhancing LLMs Via Reinforcement Learning With DataBase FeedBack
has_fulltext	False
is_retracted	False
updated_date	2025-11-28T05:42:52.777296
primary_topic
cited_by_count	0
locations_count	2
best_oa_location.id	pmh:oai:arXiv.org:2504.03713
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/2504.03713
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/2504.03713
primary_location.id	pmh:oai:arXiv.org:2504.03713
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/2504.03713
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/2504.03713
publication_date	2025-03-28
publication_year	2025
referenced_works_count	0
abstract_inverted_index.a	64, 79
abstract_inverted_index.AI	46
abstract_inverted_index.We	81
abstract_inverted_index.an	118
abstract_inverted_index.as	55, 78
abstract_inverted_index.in	20, 29, 113, 160
abstract_inverted_index.of	34, 85, 110, 133, 157, 164
abstract_inverted_index.on	12, 90, 144
abstract_inverted_index.to	58, 105
abstract_inverted_index.we	116
abstract_inverted_index.LLM	91
abstract_inverted_index.The	152
abstract_inverted_index.and	101, 148
abstract_inverted_index.for	44, 47
abstract_inverted_index.our	158
abstract_inverted_index.the	83, 107, 131, 134, 138, 155, 162
abstract_inverted_index.yet	49
abstract_inverted_index.LLMs	69
abstract_inverted_index.This	61
abstract_inverted_index.data	24, 89, 147, 167
abstract_inverted_index.hold	41
abstract_inverted_index.into	67
abstract_inverted_index.text	15
abstract_inverted_index.that	31
abstract_inverted_index.them	54
abstract_inverted_index.they	17
abstract_inverted_index.with	70, 124, 137
abstract_inverted_index.LLMs.	170
abstract_inverted_index.These	38
abstract_inverted_index.While	0
abstract_inverted_index.data,	73
abstract_inverted_index.field	163
abstract_inverted_index.large	2, 114
abstract_inverted_index.model	139
abstract_inverted_index.other	149
abstract_inverted_index.study	62
abstract_inverted_index.text.	60
abstract_inverted_index.treat	53
abstract_inverted_index.using	74
abstract_inverted_index.(LLMs)	5
abstract_inverted_index.(e.g.,	25
abstract_inverted_index.across	92
abstract_inverted_index.impact	84
abstract_inverted_index.merely	52
abstract_inverted_index.method	159
abstract_inverted_index.models	4
abstract_inverted_index.remain	18
abstract_inverted_index.tasks.	151
abstract_inverted_index.termed	121
abstract_inverted_index.unseen	146
abstract_inverted_index.within	169
abstract_inverted_index.Science	48
abstract_inverted_index.address	106
abstract_inverted_index.current	1, 50
abstract_inverted_index.massive	13
abstract_inverted_index.models,	115
abstract_inverted_index.phases,	95
abstract_inverted_index.propose	117
abstract_inverted_index.results	153
abstract_inverted_index.science	77
abstract_inverted_index.through	10
abstract_inverted_index.(RLDBF).	127
abstract_inverted_index.Database	125
abstract_inverted_index.Learning	123
abstract_inverted_index.Notably,	104
abstract_inverted_index.chemical	26, 75, 150
abstract_inverted_index.corpora,	16
abstract_inverted_index.datasets	40
abstract_inverted_index.distinct	93
abstract_inverted_index.efficacy	132
abstract_inverted_index.inherent	108
abstract_inverted_index.language	3
abstract_inverted_index.pioneers	63
abstract_inverted_index.property	88
abstract_inverted_index.proposed	135
abstract_inverted_index.testbed.	80
abstract_inverted_index.training	11, 94
abstract_inverted_index.Feedback"	126
abstract_inverted_index.advancing	45, 161
abstract_inverted_index.approach,	136
abstract_inverted_index.auxiliary	56
abstract_inverted_index.centuries	33
abstract_inverted_index.continual	97
abstract_inverted_index.enhancing	68
abstract_inverted_index.including	96
abstract_inverted_index.learning.	103
abstract_inverted_index.molecular	27, 76, 87
abstract_inverted_index.numerical	111
abstract_inverted_index.potential	156
abstract_inverted_index.strategic	42
abstract_inverted_index.approaches	51
abstract_inverted_index.databases)	30
abstract_inverted_index.exhibiting	140
abstract_inverted_index.expertise.	37
abstract_inverted_index.inadequate	19
abstract_inverted_index.innovative	119
abstract_inverted_index.leveraging	21
abstract_inverted_index.limitation	109
abstract_inverted_index.linguistic	8
abstract_inverted_index.previously	145
abstract_inverted_index.processing	168
abstract_inverted_index.properties	28
abstract_inverted_index.remarkable	7, 141
abstract_inverted_index.scientific	23, 36, 72, 166
abstract_inverted_index.structured	22, 39, 71, 165
abstract_inverted_index.supervised	99
abstract_inverted_index.systematic	65
abstract_inverted_index.accumulated	35
abstract_inverted_index.demonstrate	6, 130
abstract_inverted_index.encapsulate	32
abstract_inverted_index.evaluations	129
abstract_inverted_index.investigate	82
abstract_inverted_index.methodology	120
abstract_inverted_index.supplements	57
abstract_inverted_index.Experimental	128
abstract_inverted_index.capabilities	9, 143
abstract_inverted_index.fine-tuning,	100
abstract_inverted_index.significance	43
abstract_inverted_index.substantiate	154
abstract_inverted_index.unstructured	14, 59
abstract_inverted_index.incorporating	86
abstract_inverted_index.insensitivity	112
abstract_inverted_index.investigation	66
abstract_inverted_index.pre-training,	98
abstract_inverted_index.reinforcement	102
abstract_inverted_index."Reinforcement	122
abstract_inverted_index.generalization	142
cited_by_percentile_year
countries_distinct_count	0
institutions_distinct_count	6
citation_normalized_percentile