[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-85497-en":3,"doc-seo-85497-105":29,"detail-sidebar-cat-0-en-105":91},{"code":4,"msg":5,"data":6},0,"success",{"doc_id":7,"user_id":8,"nickname":9,"user_avatar":10,"doc_module":4,"category_id":11,"category_name":12,"doc_title":13,"doc_description":14,"doc_content":15,"file_id":16,"file_url":17,"file_type":18,"file_size":19,"view_count":20,"is_deleted":4,"is_public":20,"is_downloadable":20,"audit_status":20,"page_count":21,"language":22,"language_code":23,"site_id":24,"html_lang":23,"table_of_contents":25,"faqs":26,"seo_title":13,"seo_description":14,"update_tm":27,"read_time":28},85497,962075006959,"Anda","https://ap-avatar.wpscdn.com/avatar/e0002397efbe92a78e?_k=1776741047341049297",8,"Research & Report","PRiSM Benchmarking Phone Realization in Speech Models","Phone recognition (PR) provides a language-agnostic interface for cross-lingual speech processing and phonetic analysis by transcribing speech into phonetic units. Current evaluations often rely on surface transcription accuracy and miss blind spots in phonetic perception. PRiSM introduces an open-source benchmark with both intrinsic and extrinsic probing, standardizing transcription-based evaluation and measuring downstream utility across clinical, educational, and multilingual settings. Results show training language exposure, encoder-CTC stability, and specialized PR models outperforming Large Audio Language Models.","PRiSM: Benchmarking Phone Realization in Speech Models  \nShikhar Bharadwaj* 1 Chin-Jou Li∗1 Yoonjae Kim∗1,2 Kwanghee Choi3 Eunjung Yeo3  \nRyan Soh-Eun Shim4 Hanyu Zhou1 Brendon Boldt1 Karen Rosero Jacome1 Kalvin Chang5 Darsh Agrawal1 Keer Xu1 Chao-Han Huck Yang6 Jian Zhu7 Shinji Watanabe1 David R. Mortensen1  \n1 CMU 2 GIST 3UT Austin 4LMU Munich  \n5UC Berkeley 6NVIDIA 7UBC  \n{sbharad2,chinjoul,[dmortens}@andrew.cmu.edu](dmortens}@andrew.cmu.edu) , [rladbswo12@gm.gist.ac.kr](rladbswo12@gm.gist.ac.kr)  \narXiv :2601 . 14046v2 [ cs .CL] 13 Jul 2026  \nAbstract  \nPhone recognition (PR) serves as the atomic interface for language-agnostic modeling for cross-lingual speech processing and phonetic analysis. Despite prolonged efforts in developing PR systems, current evaluations only measure surface-level transcription accuracy. We introduce PRiSM, the first open-source benchmark designed to expose blind spots in phonetic perception through intrinsic and extrinsic evaluation of PR systems. PRiSM standardizes transcription-based evaluation and assesses downstream utility in clinical, educational, and multilingual settings with transcription and representation probes. We find that diverse language exposure during training is key to PR performance, encoder-CTC models are the most stable, and specialized PR models still outperform Large Audio Language Models. PRiSM releases code, recipes, and datasets to move the field toward multilingual speech models with robust phonetic ability 1.  \n1 Introduction  \nPhone recognition (PR) entails transcribing speech into phonetic units that capture the physical realization of sounds, independent of languagespecific phonological constraints. By preserving acoustic nuances often abstracted away by wordor phoneme-level models2 , PR provides a robust foundation for cross-lingual speech processing (Liet al., 2022 ; Yusuyin et al., 2025) and downstream applications in clinical (Shriberg et al., 2025 ; Choi et al., 2025) and educational settings (Tu et al., 2018 ; Inceoglu et al., 2023) .  \nPR models have scaled substantially to cover diverse linguistic settings (see § 2.1), yet existing evaluations remain difficult to compare across  \n*Equal contribution.  \n1 [https://github.com/changelinglab/prism](https://github.com/changelinglab/prism)  \n2For example, tell may be transcribed as [thEë] in Mainstream American English and [thEl] in Scottish English, while the phonemic form of tell is consistently /tEl/ .  \n Intrinsic: Core Capability  \n|  | \u003Cbr>→\u003Cbr>Evaluating transcription\u003Cbr> |  |\n| --- | --- | --- |\n\n Extrinsic: Downstream Utility  \nPR system Transcription Representation Lightweight Module Task-specific target  \nFigure 1: PRiSM is the first open-source benchmark for phone recognition systems, covering intrinsic and extrinsic evaluations, i.e., transcription task and downstream task performance.  \nstudies. For example, models often differ in language coverage and phone inventories (Zhu et al., 2025), and evaluation metrics are not standardized (Li et al., 2025) . A common response has been to fix a metric (Taguchi et al., 2023 ; Li et al., 2025) and expand the number of test datasets to mitigate bias (Zhu et al., 2025) . Yet this approach scales poorly due to the scarcity of phonetically transcribed data. Moreover, transcription error rates do not necessarily reflect a model’s phonetic capabilities or practical utility. Error rates in PR are inherently noisier than in ASR, as phones, unlike lexical units, correspond to a lower-level, articulatorily defined abstraction of the acoustic signal.  \nFurthermore, the link between transcription accuracy and downstream performance is often assumed rather than empirically proven. In practice, models leverage phonetic information via two channels: explicit transcriptions and latent internal representations. The latter are especially potent, as they encode rich phonetic cues (see § 2.2) . Consequently, metrics based solely on transcription error fail to capture the full","cbCaiqhFkrtFzBkr","https://ap.wps.com/l/cbCaiqhFkrtFzBkr","pdf",1951712,1,20,"English","en",105,"# Abstract\n# Introduction\n# Background\n## Phone Recognition Systems","[{\"question\":\"What problem does PRiSM target in phone recognition evaluation?\",\"answer\":\"Existing phone recognition evaluations mainly measure surface-level transcription accuracy, which does not reliably reflect phonetic capability or practical downstream utility. PRiSM is designed to expose blind spots in phonetic perception through intrinsic and extrinsic evaluation.\"},{\"question\":\"How does PRiSM evaluate phone recognition systems?\",\"answer\":\"PRiSM performs intrinsic evaluation using transcription error and extrinsic evaluation by testing downstream utility in clinical, educational, and multilingual tasks. It uses generated transcriptions and hidden representation probes.\"},{\"question\":\"What key findings does PRiSM report about model performance?\",\"answer\":\"Language exposure during training improves PR performance, encoder-CTC models are more stable, and specialized PR models still outperform Large Audio Language Models. Broader, diverse coverage also benefits results.\"}]",1784204021,50,{"code":4,"msg":30,"data":31},"ok",{"site_id":24,"language":23,"slug":32,"title":13,"keywords":33,"description":14,"schema_data":34,"social_meta":86,"head_meta":88,"extra_data":90,"updated_unix":27},"prism-benchmarking-phone-realization-in-speech-models","",{"@graph":35,"@context":85},[36,53,68],{"@type":37,"itemListElement":38},"BreadcrumbList",[39,43,47,50],{"item":40,"name":41,"@type":42,"position":20},"https://docshare.wps.com","Home","ListItem",{"item":44,"name":45,"@type":42,"position":46},"https://docshare.wps.com/document/","Document",2,{"item":48,"name":12,"@type":42,"position":49},"https://docshare.wps.com/document/research-report/",3,{"item":51,"name":13,"@type":42,"position":52},"https://docshare.wps.com/document/prism-benchmarking-phone-realization-in-speech-models/85497/",4,{"url":51,"name":13,"@type":54,"author":55,"headline":13,"publisher":57,"fileFormat":60,"inLanguage":23,"description":14,"dateModified":61,"datePublished":62,"encodingFormat":60,"isAccessibleForFree":63,"interactionStatistic":64},"DigitalDocument",{"name":9,"@type":56},"Person",{"url":40,"name":58,"@type":59},"DocShare","Organization","application/pdf","2026-07-17","2026-07-16",true,{"@type":65,"interactionType":66,"userInteractionCount":20},"InteractionCounter",{"@type":67},"ViewAction",{"@type":69,"mainEntity":70},"FAQPage",[71,77,81],{"name":72,"@type":73,"acceptedAnswer":74},"What problem does PRiSM target in phone recognition evaluation?","Question",{"text":75,"@type":76},"Existing phone recognition evaluations mainly measure surface-level transcription accuracy, which does not reliably reflect phonetic capability or practical downstream utility. PRiSM is designed to expose blind spots in phonetic perception through intrinsic and extrinsic evaluation.","Answer",{"name":78,"@type":73,"acceptedAnswer":79},"How does PRiSM evaluate phone recognition systems?",{"text":80,"@type":76},"PRiSM performs intrinsic evaluation using transcription error and extrinsic evaluation by testing downstream utility in clinical, educational, and multilingual tasks. It uses generated transcriptions and hidden representation probes.",{"name":82,"@type":73,"acceptedAnswer":83},"What key findings does PRiSM report about model performance?",{"text":84,"@type":76},"Language exposure during training improves PR performance, encoder-CTC models are more stable, and specialized PR models still outperform Large Audio Language Models. Broader, diverse coverage also benefits results.","https://schema.org",{"og:url":51,"og:type":87,"og:title":13,"og:site_name":58,"og:description":14},"article",{"robots":89,"canonical":51},"index,follow",{"doc_id":7,"site_id":24},{"code":4,"msg":5,"data":92},[93,97,101,105,110,114,119,122,126,129,133],{"id":20,"doc_module":4,"doc_module_name":45,"category_name":94,"show_sort_weight":95,"slug":96},"Story & Novel",90,"story-novel",{"id":46,"doc_module":4,"doc_module_name":45,"category_name":98,"show_sort_weight":99,"slug":100},"Literature",80,"literature",{"id":52,"doc_module":4,"doc_module_name":45,"category_name":102,"show_sort_weight":103,"slug":104},"Exam",70,"exam",{"id":106,"doc_module":4,"doc_module_name":45,"category_name":107,"show_sort_weight":108,"slug":109},5,"Comic",60,"comic",{"id":111,"doc_module":4,"doc_module_name":45,"category_name":112,"show_sort_weight":28,"slug":113},6,"Technology","technology",{"id":115,"doc_module":4,"doc_module_name":45,"category_name":116,"show_sort_weight":117,"slug":118},7,"Healthcare",40,"healthcare",{"id":11,"doc_module":4,"doc_module_name":45,"category_name":12,"show_sort_weight":120,"slug":121},30,"research-report",{"id":123,"doc_module":4,"doc_module_name":45,"category_name":124,"show_sort_weight":21,"slug":125},9,"Religion & Spirituality","religion-spirituality",{"id":21,"doc_module":4,"doc_module_name":45,"category_name":127,"show_sort_weight":21,"slug":128},"World Cup","world-cup",{"id":130,"doc_module":4,"doc_module_name":45,"category_name":131,"show_sort_weight":130,"slug":132},10,"Lifestyle","lifestyle",{"id":134,"doc_module":4,"doc_module_name":45,"category_name":135,"show_sort_weight":106,"slug":136},19,"General","general"]