[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-81748-en":3,"doc-seo-81748-105":30,"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":21,"is_downloadable":21,"audit_status":21,"page_count":22,"language":23,"language_code":24,"site_id":25,"html_lang":24,"table_of_contents":26,"faqs":27,"seo_title":13,"seo_description":14,"update_tm":28,"read_time":29},81748,4398048949847,"Eliana","https://ap-avatar.wpscdn.com/avatar/400002536579ef2da7f?_k=1778318612642679267",8,"Research & Report","Representation as a Bottleneck for Mechanistic Interpretability: The Manifestation Unit Protocol","Mechanistic interpretability produces component-level analyses, but their outputs are difficult to reuse across studies, limiting composability, natural-language querying, and downstream auditing or intervention. This work treats the representation layer between analyses and use as a bottleneck and introduces Manifestation Units, a typed tuple protocol (E, S, R, D, G) with transformer attention-head primitives (T). Across β-VAE, CNN, and GPT-2, typed structure improves retrieval, and CNN retrievals satisfy causal sufficiency and necessity under matched budgets.","Representation as a Bottleneck for Mechanistic Interpretability: The Manifestation Unit Protocol  \nHussein Chouman 1 Wataru Sasaki 1 Tomokazu Matsui 1 Hirohiko Suwa 1 Keiichi Yasumoto 1  \narXiv :2607 .00089v1 [ cs .LG] 30 Jun 2026  \nAbstract  \nMechanistic interpretability has produced a rich inventory of component-level analyses that characterise what neural-network components encode and how they interact. Their outputs, however, are not easily reusable: selectivity tables, circuit diagrams, and feature lists remain locked in per-study notebooks—non-composable, not queryable in natural language, and not directly actionable for downstream audit or intervention. We study the representation layer that sits between these analyses and downstream use as a bottleneck that can be evaluated independently and introduce Manifestation Units, a typed tuple protocol (E, S, R, D, G) extended with attentionhead primitives (T) for transformer architectures organising per-component statistics into structured fields populated automatically and queried through hybrid retrieval. Instantiated across generative vision (β-VAE), discriminative vision (CNN), and language (GPT-2), the protocol supports two findings: typed structure substantially outperforms unstructured baselines on retrieval, and CNN filters retrieved by the schema satisfy causal sufficiency and necessity criteria under matched-budget controls. The schema absorbs attention-head primitives without modification, set-recovers known IOI circuit members under retrieval-budget-matched controls, and reveals an irreducible two-field core (S+R) with remaining fields either redundant or actively interfering. We present this as schema infrastructure for mechanistic interpretability rather than frontier-scale validation. 1  \n1Ubiquitous Computing Systems Laboratory, Nara Institute of Science and Technology, Nara, Japan. Correspondence to: Hussein Chouman \u003C[chouman.hussein.ck5@naist.ac.jp](chouman.hussein.ck5@naist.ac.jp) >.  \nPreprint. July 2, 2026.  \n1Transformer and CNN demos are available at  \n[https://manifestation-xai.github.io/manifestation-transformers/](https://manifestation-xai.github.io/manifestation-transformers/)[ ](https://manifestation-xai.github.io/manifestation-transformers/)and [https://manifestation-xai.github.io/manifestation-cnn/](https://manifestation-xai.github.io/manifestation-cnn/) .  \n1. Introduction  \nMechanistic interpretability has produced a rich inventory of analysis techniques for characterising what neural-network components encode and how they interact: Network Dissection, probing classifiers, sparse-autoencoder decompositions, and circuit analyses (Bau et al., 2017 ; Belinkov, 2022 ; Lieberum et al., 2024 ; Elhage et al., 2021 ; Wanget al., 2023 ; Olsson et al., 2022 ; Conmy et al., 2023) . Their outputs, however, are not easily reusable. Selectivity tables, circuit diagrams, and SAE feature lists remain in per-study notebooks and scripts: not composable, not queryable in natural language, and not directly actionable for downstream audit or intervention.  \nA practitioner asks of a trained CNN: “which filters encode ‘dog’, and what happens if I amplify them?” The needed statistics—selectivity, activation correlations, ablation effects—already exist, but there is no grounded interface to ask. Semantic search fails on numerical identifiers (Li et al., 2023), exact-match fails on conceptual queries, and free-form prose loses the typing that separates“which filters detect dog” from “what does filter 81 detect.”  \nWe study the representation layer that sits between component-level analyses and downstream use, holding the analyses themselves fixed. We introduce Manifestation Units, a typed tuple protocol (E, S, R, D, G, T) that organises per-component statistics into six fields aligned with distinct query primitives: entity identity (E), semantic associations (S), cross-component relations (R), quantified dynamics (D), intervention guidance (G), and attention-head primit","cbCaijhKPCgl7A3F","https://ap.wps.com/l/cbCaijhKPCgl7A3F","pdf",8223321,2,1,65,"English","en",105,"# Introduction\n## Representation bottleneck and motivation\n## Manifestation Units: typed tuple protocol\n## Retrieval and hybrid matching\n## Evaluation hypotheses (H1, H2)\n## Empirical findings across architectures","[{\"question\":\"What problem does the paper identify in existing mechanistic interpretability outputs?\",\"answer\":\"Selectivity tables, circuit diagrams, and feature lists stay trapped in per-study notebooks, making them non-composable, not queryable in natural language, and not directly usable for downstream audit or intervention.\"},{\"question\":\"What are Manifestation Units and how do they structure information?\",\"answer\":\"Manifestation Units are a typed tuple protocol (E, S, R, D, G) that organizes per-component statistics into fields aligned with specific query primitives, extended for transformer attention-head primitives (T). The fields are extracted automatically from activations and validation-set statistics.\"},{\"question\":\"How does typed structure affect retrieval performance?\",\"answer\":\"Natural-language retrieval over component analyses requires typed structure: structured decompositions substantially outperform unstructured baselines, as shown by improved retrieval metrics on the GPT-2 schema and by field-level ablations.\"}]",1784175813,164,{"code":4,"msg":31,"data":32},"ok",{"site_id":25,"language":24,"slug":33,"title":13,"keywords":34,"description":14,"schema_data":35,"social_meta":86,"head_meta":88,"extra_data":90,"updated_unix":28},"representation-as-a-bottleneck-for-mechanistic-interpretability-the-manifestation-unit-protocol","",{"@graph":36,"@context":85},[37,53,68],{"@type":38,"itemListElement":39},"BreadcrumbList",[40,44,47,50],{"item":41,"name":42,"@type":43,"position":21},"https://docshare.wps.com","Home","ListItem",{"item":45,"name":46,"@type":43,"position":20},"https://docshare.wps.com/document/","Document",{"item":48,"name":12,"@type":43,"position":49},"https://docshare.wps.com/document/research-report/",3,{"item":51,"name":13,"@type":43,"position":52},"https://docshare.wps.com/document/representation-as-a-bottleneck-for-mechanistic-interpretability-the-manifestation-unit-protocol/81748/",4,{"url":51,"name":13,"@type":54,"author":55,"headline":13,"publisher":57,"fileFormat":60,"inLanguage":24,"description":14,"dateModified":61,"datePublished":62,"encodingFormat":60,"isAccessibleForFree":63,"interactionStatistic":64},"DigitalDocument",{"name":9,"@type":56},"Person",{"url":41,"name":58,"@type":59},"DocShare","Organization","application/pdf","2026-07-25","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 the paper identify in existing mechanistic interpretability outputs?","Question",{"text":75,"@type":76},"Selectivity tables, circuit diagrams, and feature lists stay trapped in per-study notebooks, making them non-composable, not queryable in natural language, and not directly usable for downstream audit or intervention.","Answer",{"name":78,"@type":73,"acceptedAnswer":79},"What are Manifestation Units and how do they structure information?",{"text":80,"@type":76},"Manifestation Units are a typed tuple protocol (E, S, R, D, G) that organizes per-component statistics into fields aligned with specific query primitives, extended for transformer attention-head primitives (T). The fields are extracted automatically from activations and validation-set statistics.",{"name":82,"@type":73,"acceptedAnswer":83},"How does typed structure affect retrieval performance?",{"text":84,"@type":76},"Natural-language retrieval over component analyses requires typed structure: structured decompositions substantially outperform unstructured baselines, as shown by improved retrieval metrics on the GPT-2 schema and by field-level ablations.","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":25},{"code":4,"msg":5,"data":92},[93,97,101,105,110,115,120,123,128,131,135],{"id":21,"doc_module":4,"doc_module_name":46,"category_name":94,"show_sort_weight":95,"slug":96},"Story & Novel",90,"story-novel",{"id":20,"doc_module":4,"doc_module_name":46,"category_name":98,"show_sort_weight":99,"slug":100},"Literature",80,"literature",{"id":52,"doc_module":4,"doc_module_name":46,"category_name":102,"show_sort_weight":103,"slug":104},"Exam",70,"exam",{"id":106,"doc_module":4,"doc_module_name":46,"category_name":107,"show_sort_weight":108,"slug":109},5,"Comic",60,"comic",{"id":111,"doc_module":4,"doc_module_name":46,"category_name":112,"show_sort_weight":113,"slug":114},6,"Technology",50,"technology",{"id":116,"doc_module":4,"doc_module_name":46,"category_name":117,"show_sort_weight":118,"slug":119},7,"Healthcare",40,"healthcare",{"id":11,"doc_module":4,"doc_module_name":46,"category_name":12,"show_sort_weight":121,"slug":122},30,"research-report",{"id":124,"doc_module":4,"doc_module_name":46,"category_name":125,"show_sort_weight":126,"slug":127},9,"Religion & Spirituality",20,"religion-spirituality",{"id":126,"doc_module":4,"doc_module_name":46,"category_name":129,"show_sort_weight":126,"slug":130},"World Cup","world-cup",{"id":132,"doc_module":4,"doc_module_name":46,"category_name":133,"show_sort_weight":132,"slug":134},10,"Lifestyle","lifestyle",{"id":136,"doc_module":4,"doc_module_name":46,"category_name":137,"show_sort_weight":106,"slug":138},19,"General","general"]