Reflect before Act: Proactive Error Correction in Language Models Article Swipe

PDF

Qiuhai Zeng , S. Rajkumar , Narendra Gyanchandani , Wenbo Yan ·

YOU? · · 2025 · Open Access · · DOI: https://doi.org/10.48550/arxiv.2509.18607

Large Language Models (LLMs) have demonstrated remarkable capabilities in interactive decision-making tasks, but existing methods often struggle with error accumulation and lack robust self-correction mechanisms. We introduce "Reflect before Act" (REBACT), a novel approach that enhances LLM-based decision-making by introducing a critical reflect step prior to taking the next action. This approach allows for immediate error correction, ensuring smooth action path and adaptibity to environment feedback. We evaluate REBACT on three diverse interactive environments: ALFWorld, WebShop, and TextCraft. Our results demonstrate that REBACT significantly outperforms strong baselines, improving success rates by up to 24% on WebShop (achieving 61%), 6.72% on ALFWorld (achieving 98.51%), and 0.5% on TextCraft (achieving 99.5%) using Claude3.5-sonnet as the underlying LLM. Further analysis reveals that REBACT's performance improvements are achieved with only a few modification steps, demonstrating its computational efficiency.

Related Topics

Me Before You (Film)

Before Sunrise

The Nightmare Before Christmas

The Land Before Time

Before Trilogy

The Land Before Time (Film)

Before Sunset

The Storm Before The Calm

To All The Boys I'Ve Loved Before (Film)

Before The Devil Knows You'Re Dead

1001 Albums You Must Hear Before You Die

Earnings Before Interest And Taxes

Concepts

No concepts available.

Metadata

Type: preprint
Language: en
Landing Page: http://arxiv.org/abs/2509.18607
PDF: https://arxiv.org/pdf/2509.18607
OA Status: green
OpenAlex ID: https://openalex.org/W4416258148

All OpenAlex metadata

Raw OpenAlex JSON

OpenAlex ID: https://openalex.org/W4416258148

Canonical identifier for this work in OpenAlex
DOI: https://doi.org/10.48550/arxiv.2509.18607

Digital Object Identifier
Title: Reflect before Act: Proactive Error Correction in Language Models

Work title
Type: preprint

OpenAlex work type
Language: en

Primary language
Publication year: 2025

Year of publication
Publication date: 2025-09-23

Full publication date if available
Authors: Qiuhai Zeng, S. Rajkumar, Narendra Gyanchandani, Wenbo Yan

List of authors in order
Landing page: https://arxiv.org/abs/2509.18607

Publisher landing page
PDF URL: https://arxiv.org/pdf/2509.18607

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/2509.18607

Direct OA link when available
Cited by: 0

Total citation count in OpenAlex

Full payload

id	https://openalex.org/W4416258148
doi	https://doi.org/10.48550/arxiv.2509.18607
ids.doi	https://doi.org/10.48550/arxiv.2509.18607
ids.openalex	https://openalex.org/W4416258148
fwci
type	preprint
title	Reflect before Act: Proactive Error Correction in Language Models
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:2509.18607
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/2509.18607
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/2509.18607
locations[1].id	doi:10.48550/arxiv.2509.18607
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.2509.18607
indexed_in	arxiv, datacite
authorships[0].author.id	https://openalex.org/A5037089279
authorships[0].author.orcid	https://orcid.org/0000-0002-9956-8397
authorships[0].author.display_name	Qiuhai Zeng
authorships[0].author_position	first
authorships[0].raw_author_name	Zeng, Qiuhai
authorships[0].is_corresponding	False
authorships[1].author.id	https://openalex.org/A5108882709
authorships[1].author.orcid
authorships[1].author.display_name	S. Rajkumar
authorships[1].author_position	middle
authorships[1].raw_author_name	Rajkumar, Sarvesh
authorships[1].is_corresponding	False
authorships[2].author.id	https://openalex.org/A5120280750
authorships[2].author.orcid
authorships[2].author.display_name	Narendra Gyanchandani
authorships[2].author_position	last
authorships[2].raw_author_name	Gyanchandani, Narendra
authorships[2].is_corresponding	False
authorships[3].author.id	https://openalex.org/A5050783869
authorships[3].author.orcid	https://orcid.org/0000-0002-6977-6283
authorships[3].author.display_name	Wenbo Yan
authorships[3].author_position	middle
authorships[3].raw_author_name	Yan, Wenbo
authorships[3].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/2509.18607
open_access.oa_status	green
open_access.any_repository_has_fulltext	False
created_date	2025-10-10T00:00:00
display_name	Reflect before Act: Proactive Error Correction in Language Models
has_fulltext	False
is_retracted	False
updated_date	2025-11-28T09:34:26.226595
primary_topic
cited_by_count	0
locations_count	2
best_oa_location.id	pmh:oai:arXiv.org:2509.18607
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/2509.18607
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/2509.18607
primary_location.id	pmh:oai:arXiv.org:2509.18607
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/2509.18607
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/2509.18607
publication_date	2025-09-23
publication_year	2025
referenced_works_count	0
abstract_inverted_index.a	31, 40, 126
abstract_inverted_index.We	25, 66
abstract_inverted_index.as	111
abstract_inverted_index.by	38, 90
abstract_inverted_index.in	8
abstract_inverted_index.on	69, 94, 99, 105
abstract_inverted_index.to	45, 63, 92
abstract_inverted_index.up	91
abstract_inverted_index.24%	93
abstract_inverted_index.Our	78
abstract_inverted_index.and	20, 61, 76, 103
abstract_inverted_index.are	122
abstract_inverted_index.but	12
abstract_inverted_index.few	127
abstract_inverted_index.for	53
abstract_inverted_index.its	131
abstract_inverted_index.the	47, 112
abstract_inverted_index.0.5%	104
abstract_inverted_index.Act"	29
abstract_inverted_index.LLM.	114
abstract_inverted_index.This	50
abstract_inverted_index.have	4
abstract_inverted_index.lack	21
abstract_inverted_index.next	48
abstract_inverted_index.only	125
abstract_inverted_index.path	60
abstract_inverted_index.step	43
abstract_inverted_index.that	34, 81, 118
abstract_inverted_index.with	17, 124
abstract_inverted_index.6.72%	98
abstract_inverted_index.61%),	97
abstract_inverted_index.Large	0
abstract_inverted_index.error	18, 55
abstract_inverted_index.novel	32
abstract_inverted_index.often	15
abstract_inverted_index.prior	44
abstract_inverted_index.rates	89
abstract_inverted_index.three	70
abstract_inverted_index.using	109
abstract_inverted_index.(LLMs)	3
abstract_inverted_index.99.5%)	108
abstract_inverted_index.Models	2
abstract_inverted_index.REBACT	68, 82
abstract_inverted_index.action	59
abstract_inverted_index.allows	52
abstract_inverted_index.before	28
abstract_inverted_index.robust	22
abstract_inverted_index.smooth	58
abstract_inverted_index.steps,	129
abstract_inverted_index.strong	85
abstract_inverted_index.taking	46
abstract_inverted_index.tasks,	11
abstract_inverted_index.Further	115
abstract_inverted_index.WebShop	95
abstract_inverted_index.action.	49
abstract_inverted_index.diverse	71
abstract_inverted_index.methods	14
abstract_inverted_index.reflect	42
abstract_inverted_index.results	79
abstract_inverted_index.reveals	117
abstract_inverted_index.success	88
abstract_inverted_index."Reflect	27
abstract_inverted_index.98.51%),	102
abstract_inverted_index.ALFWorld	100
abstract_inverted_index.Language	1
abstract_inverted_index.REBACT's	119
abstract_inverted_index.WebShop,	75
abstract_inverted_index.achieved	123
abstract_inverted_index.analysis	116
abstract_inverted_index.approach	33, 51
abstract_inverted_index.critical	41
abstract_inverted_index.enhances	35
abstract_inverted_index.ensuring	57
abstract_inverted_index.evaluate	67
abstract_inverted_index.existing	13
abstract_inverted_index.struggle	16
abstract_inverted_index.(REBACT),	30
abstract_inverted_index.ALFWorld,	74
abstract_inverted_index.LLM-based	36
abstract_inverted_index.TextCraft	106
abstract_inverted_index.feedback.	65
abstract_inverted_index.immediate	54
abstract_inverted_index.improving	87
abstract_inverted_index.introduce	26
abstract_inverted_index.(achieving	96, 101, 107
abstract_inverted_index.TextCraft.	77
abstract_inverted_index.adaptibity	62
abstract_inverted_index.baselines,	86
abstract_inverted_index.remarkable	6
abstract_inverted_index.underlying	113
abstract_inverted_index.correction,	56
abstract_inverted_index.demonstrate	80
abstract_inverted_index.efficiency.	133
abstract_inverted_index.environment	64
abstract_inverted_index.interactive	9, 72
abstract_inverted_index.introducing	39
abstract_inverted_index.mechanisms.	24
abstract_inverted_index.outperforms	84
abstract_inverted_index.performance	120
abstract_inverted_index.accumulation	19
abstract_inverted_index.capabilities	7
abstract_inverted_index.demonstrated	5
abstract_inverted_index.improvements	121
abstract_inverted_index.modification	128
abstract_inverted_index.computational	132
abstract_inverted_index.demonstrating	130
abstract_inverted_index.environments:	73
abstract_inverted_index.significantly	83
abstract_inverted_index.decision-making	10, 37
abstract_inverted_index.self-correction	23
abstract_inverted_index.Claude3.5-sonnet	110
cited_by_percentile_year
countries_distinct_count	0
institutions_distinct_count	4
citation_normalized_percentile