{"id":"https://openalex.org/W7154490662","doi":"https://doi.org/10.48550/arxiv.2604.12652","title":"PromptEcho: Annotation-Free Reward from Vision-Language Models for Text-to-Image Reinforcement Learning","display_name":"PromptEcho: Annotation-Free Reward from Vision-Language Models for Text-to-Image Reinforcement Learning","publication_year":2026,"publication_date":"2026-04-14","ids":{"openalex":"https://openalex.org/W7154490662","doi":"https://doi.org/10.48550/arxiv.2604.12652"},"language":null,"primary_location":{"id":"doi:10.48550/arxiv.2604.12652","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2604.12652","pdf_url":null,"source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":null,"issn":null,"is_oa":true,"is_in_doaj":false,"is_core":false,"host_organization":"https://openalex.org/I205783295","host_organization_name":"Cornell University","host_organization_lineage":["https://openalex.org/I205783295"],"host_organization_lineage_names":[],"type":"repository"},"license":null,"license_id":null,"version":null,"is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"article"},"type":"preprint","indexed_in":["datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://doi.org/10.48550/arxiv.2604.12652","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5133727653","display_name":"Jinlong Liu","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Liu, Jinlong","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5133679438","display_name":"Wanggui He","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"He, Wanggui","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5133655726","display_name":"Peng Zhang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zhang, Peng","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5039853684","display_name":"Mushui Liu","orcid":"https://orcid.org/0000-0002-2909-7702"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Liu, Mushui","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5133666519","display_name":"Hao Jiang","orcid":"https://orcid.org/0000-0003-0856-5818"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Jiang, Hao","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5059615376","display_name":"Pipei Huang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Huang, Pipei","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]}],"institutions":[],"countries_distinct_count":0,"institutions_distinct_count":6,"corresponding_author_ids":[],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":null,"has_fulltext":false,"cited_by_count":0,"citation_normalized_percentile":null,"cited_by_percentile_year":null,"biblio":{"volume":null,"issue":null,"first_page":null,"last_page":null},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T11714","display_name":"Multimodal Machine Learning Applications","score":0.8700000047683716,"subfield":{"id":"https://openalex.org/subfields/1707","display_name":"Computer Vision and Pattern Recognition"},"field":{"id":"https://openalex.org/fields/17","display_name":"Computer Science"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}},"topics":[{"id":"https://openalex.org/T11714","display_name":"Multimodal Machine Learning Applications","score":0.8700000047683716,"subfield":{"id":"https://openalex.org/subfields/1707","display_name":"Computer Vision and Pattern Recognition"},"field":{"id":"https://openalex.org/fields/17","display_name":"Computer Science"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}},{"id":"https://openalex.org/T11636","display_name":"Artificial Intelligence in Healthcare and Education","score":0.01360000018030405,"subfield":{"id":"https://openalex.org/subfields/2718","display_name":"Health Informatics"},"field":{"id":"https://openalex.org/fields/27","display_name":"Medicine"},"domain":{"id":"https://openalex.org/domains/4","display_name":"Health Sciences"}},{"id":"https://openalex.org/T10775","display_name":"Generative Adversarial Networks and Image Synthesis","score":0.01269999984651804,"subfield":{"id":"https://openalex.org/subfields/1707","display_name":"Computer Vision and Pattern Recognition"},"field":{"id":"https://openalex.org/fields/17","display_name":"Computer Science"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}}],"keywords":[{"id":"https://openalex.org/keywords/reinforcement-learning","display_name":"Reinforcement learning","score":0.7497000098228455},{"id":"https://openalex.org/keywords/benchmark","display_name":"Benchmark (surveying)","score":0.5983999967575073},{"id":"https://openalex.org/keywords/quality","display_name":"Quality (philosophy)","score":0.44609999656677246},{"id":"https://openalex.org/keywords/annotation","display_name":"Annotation","score":0.42800000309944153},{"id":"https://openalex.org/keywords/preference","display_name":"Preference","score":0.4068000018596649},{"id":"https://openalex.org/keywords/reinforcement","display_name":"Reinforcement","score":0.37389999628067017}],"concepts":[{"id":"https://openalex.org/C97541855","wikidata":"https://www.wikidata.org/wiki/Q830687","display_name":"Reinforcement learning","level":2,"score":0.7497000098228455},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.7246000170707703},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.6780999898910522},{"id":"https://openalex.org/C185798385","wikidata":"https://www.wikidata.org/wiki/Q1161707","display_name":"Benchmark (surveying)","level":2,"score":0.5983999967575073},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.585099995136261},{"id":"https://openalex.org/C2779530757","wikidata":"https://www.wikidata.org/wiki/Q1207505","display_name":"Quality (philosophy)","level":2,"score":0.44609999656677246},{"id":"https://openalex.org/C2776321320","wikidata":"https://www.wikidata.org/wiki/Q857525","display_name":"Annotation","level":2,"score":0.42800000309944153},{"id":"https://openalex.org/C2781249084","wikidata":"https://www.wikidata.org/wiki/Q908656","display_name":"Preference","level":2,"score":0.4068000018596649},{"id":"https://openalex.org/C67203356","wikidata":"https://www.wikidata.org/wiki/Q1321905","display_name":"Reinforcement","level":2,"score":0.37389999628067017},{"id":"https://openalex.org/C176217482","wikidata":"https://www.wikidata.org/wiki/Q860554","display_name":"Metric (unit)","level":2,"score":0.2980000078678131},{"id":"https://openalex.org/C196340769","wikidata":"https://www.wikidata.org/wiki/Q7698910","display_name":"Temporal difference learning","level":3,"score":0.2662999927997589},{"id":"https://openalex.org/C108583219","wikidata":"https://www.wikidata.org/wiki/Q197536","display_name":"Deep learning","level":2,"score":0.25589999556541443}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.48550/arxiv.2604.12652","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2604.12652","pdf_url":null,"source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":null,"issn":null,"is_oa":true,"is_in_doaj":false,"is_core":false,"host_organization":"https://openalex.org/I205783295","host_organization_name":"Cornell University","host_organization_lineage":["https://openalex.org/I205783295"],"host_organization_lineage_names":[],"type":"repository"},"license":null,"license_id":null,"version":null,"is_accepted":false,"is_published":null,"raw_source_name":null,"raw_type":"article"}],"best_oa_location":{"id":"doi:10.48550/arxiv.2604.12652","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2604.12652","pdf_url":null,"source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":null,"issn":null,"is_oa":true,"is_in_doaj":false,"is_core":false,"host_organization":"https://openalex.org/I205783295","host_organization_name":"Cornell University","host_organization_lineage":["https://openalex.org/I205783295"],"host_organization_lineage_names":[],"type":"repository"},"license":null,"license_id":null,"version":null,"is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"article"},"sustainable_development_goals":[],"awards":[],"funders":[],"has_content":{"pdf":false,"grobid_xml":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"Reinforcement":[0],"learning":[1],"(RL)":[2],"can":[3],"improve":[4],"the":[5,65,74,78,82,169,183,187],"prompt":[6,76,119],"following":[7,120],"capability":[8],"of":[9,69,112],"text-to-image":[10],"(T2I)":[11],"models,":[12],"yet":[13],"obtaining":[14],"high-quality":[15],"reward":[16,27,43,52,91,174],"signals":[17],"remains":[18],"challenging:":[19],"CLIP":[20],"Score":[21],"is":[22,92],"too":[23],"coarse-grained,":[24],"while":[25],"VLM-based":[26],"models":[28,128,185],"(e.g.,":[29],"RewardDance)":[30],"require":[31],"costly":[32],"human-annotated":[33],"preference":[34],"data":[35],"and":[36,50,59,96,130,153,172,186],"additional":[37],"fine-tuning.":[38],"We":[39,180],"propose":[40],"PromptEcho,":[41],"a":[42,56,60,70,110],"construction":[44],"method":[45],"that":[46,133,162,173],"requires":[47],"\\emph{no}":[48,51],"annotation":[49],"model":[53],"training.":[54,158],"Given":[55],"generated":[57],"image":[58],"guiding":[61],"query,":[62],"PromptEcho":[63,134,163],"computes":[64],"token-level":[66],"cross-entropy":[67],"loss":[68],"frozen":[71],"VLM":[72,88,178],"with":[73,147,168,177],"original":[75],"as":[77,99],"label,":[79],"directly":[80],"extracting":[81],"image-text":[83],"alignment":[84],"knowledge":[85],"encoded":[86],"during":[87],"pretraining.":[89],"The":[90],"deterministic,":[93],"computationally":[94],"efficient,":[95],"improves":[97],"automatically":[98],"stronger":[100],"open-source":[101,182],"VLMs":[102],"become":[103],"available.":[104],"For":[105],"evaluation,":[106],"we":[107],"develop":[108],"DenseAlignBench,":[109],"benchmark":[111],"concept-rich":[113],"dense":[114],"captions":[115],"for":[116],"rigorously":[117],"testing":[118],"capability.":[121],"Experimental":[122],"results":[123],"on":[124,138,150],"two":[125],"state-of-the-art":[126],"T2I":[127],"(Z-Image":[129],"QwenImage-2512)":[131],"demonstrate":[132],"achieves":[135],"substantial":[136],"improvements":[137],"DenseAlignBench":[139],"(+26.8pp":[140],"/":[141],"+16.2pp":[142],"net":[143],"win":[144],"rate),":[145],"along":[146],"consistent":[148],"gains":[149],"GenEval,":[151],"DPG-Bench,":[152],"TIIFBench":[154],"without":[155],"any":[156],"task-specific":[157],"Ablation":[159],"studies":[160],"confirm":[161],"comprehensively":[164],"outperforms":[165],"inference-based":[166],"scoring":[167],"same":[170],"VLM,":[171],"quality":[175],"scales":[176],"size.":[179],"will":[181],"trained":[184],"DenseAlignBench.":[188]},"counts_by_year":[],"updated_date":"2026-06-11T09:08:48.828518","created_date":"2026-04-16T00:00:00"}