[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-81944-en":3,"doc-seo-81944-105":31,"detail-sidebar-cat-0-en-105":92},{"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":28,"seo_description":14,"update_tm":29,"read_time":30},81944,8796095461610,"Oliver","https://ap-avatar.wpscdn.com/davatar_276721f389ce27ea32af1340a28f341c",8,"Research & Report","AbICL: In-Context Learning for Antigen-Specific Antibody Affinity Ranking","Accurate ranking of antibody candidates by binding affinity is crucial for therapeutic antibody discovery, yet many existing methods compare affinities in isolation and fail to exploit context from other labeled comparisons. With only a small set of experimentally characterized affinity comparisons available for many target antigens, this work investigates whether a model can infer antigen-specific ranking patterns from these demonstrations. The proposed AbICL uses a pretrained structural encoder and a context ranking head, trained with episodic meta-training for test-time adaptation without gradient updates. Experiments on AbRank show consistent improvements over ranking baselines, with larger gains under distribution shift and fine-grained discrimination.","AbICL: In-Context Learning for Antigen-Specific  \nAntibody Affinity Ranking  \nZhiyuan Chen† Henlius Shanghai, China  \nZhiyuan [Chen@henlius.com](Chen@henlius.com)  \nXinyi Yang  \nHenlius  \nShanghai, China [Connie](Connie Yang@henlius.com)[ ](Connie Yang@henlius.com)[Yang@henlius.com](Connie Yang@henlius.com)  \nJing Hu†  \nHenlius Shanghai, China  \nJing [Hu1@henlius.com](Hu1@henlius.com)  \nJunzhe Wang  \nHenlius Shanghai, China [Junzhe](Junzhe Wang@henlius.com)[ ](Junzhe Wang@henlius.com)[Wang@henlius.com](Junzhe Wang@henlius.com)  \nYueyang Huang  \nHenlius Shanghai, China  \nYueyang [Huang@henlius.com](Huang@henlius.com)  \nZhaoyang Wang  \nHenlius Shanghai, China [John](John Wang1@henlius.com)[ ](John Wang1@henlius.com)[Wang1@henlius.com](John Wang1@henlius.com)  \nFeng Zhu  \nHenlius Shanghai, China  \nShawn [Zhu@henlius.com](Zhu@henlius.com)  \narXiv :2607 .05846v 1 [ cs .LG] 7 Jul 2026  \nAbstract—Accurate ranking of antibody candidates according to their binding affinity is essential for therapeutic antibody discovery. However, existing methods treat affinity comparisons independently and ignore the contextual information encoded in other labeled comparisons, limiting their ability to capture antigen-specific binding landscapes. For many target antigens, a small number of experimentally characterized affinity comparisons are often available. An important question is whether the model can exploit these existing comparisons to infer antigenspecific ranking patterns that facilitate subsequent affinity ranking. This form of learning from labeled demonstrations closely resembles the paradigm of In-Context Learning, motivating us torevisit antibody affinity ranking from an ICL perspective. To this end, we propose AbICL, an ICL framework for antigen-specific antibody affinity ranking. AbICL combines a pretrained structural encoder with a context ranking head and is trained with an episodic meta-training strategy that enables the model to leverage support demonstrations for test-time adaptation without gradient updates. Experiments on the AbRank benchmark demonstrate that AbICL consistently outperforms existing ranking baselines across almost all data splits and evaluation benchmarks. Further analysis shows that the value of contextual demonstrations depends on how well they match the target inference task, and becomes increasingly pronounced under distribution shift and fine-grained affinity discrimination. These findings highlight the potential of ICL as an effective paradigm for antigen-specific antibody affinity ranking, particularly in challenging settings where a single global ranking function is insufficient.  \nIndex Terms—In-Context Learning, Meta-Episodic Learning, Antibody Affinity Ranking.  \nI. INTRODUCTION  \nPredicting antibody-antigen binding affinity is central to therapeutic antibody discovery, yet direct affinity prediction remains challenging because affinity measurements are often noisy, assay-dependent, and frequently censored. Consequently, affinity ranking, which prioritizes candidates according to their relative binding strengths rather than absolute affinity values, has emerged as a robust alternative.  \nExisting approaches for antibody affinity prediction can be broadly categorized into three paradigms: affinity regres-  \n†Corresponding author  \nsion methods [1, 1, 2, 3, 4, 5, 6, 7, 8, 9] , ranking-based methods [10, 11], and indirect affinity scoring methods based on pretrained foundation models [12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23] . Affinity regression methods directly predict experimental binding affinities but are often affected by noisy and heterogeneous measurements. Ranking-based methods instead learn relative binding preferences, providing a more robust objective, yet typically rely on a contextindependent ranking function that treats each comparison as an isolated prediction problem. Indirect affinity scoring methods estimate binding quality using sequence likelihoods, structural confidence, or pr","cbCaimACyiIFefJ9","https://ap.wps.com/l/cbCaimACyiIFefJ9","pdf",623351,2,1,10,"English","en",105,"# Introduction\n## Antibody affinity prediction and ranking motivation\n## Related paradigms and limitations\n## Antigen-specific evidence and ICL framing\n# AbICL framework\n## Model components: structural encoder and context ranking head\n## Episodic meta-training and test-time adaptation\n# Experiments and findings\n## Overall performance on AbRank\n## Context demonstration effects under shift and fine-grained discrimination","[{\"question\":\"Why is antibody affinity ranking important in therapeutic antibody discovery?\",\"answer\":\"Affinity ranking prioritizes antibody candidates based on relative binding strengths, offering robustness when affinity measurements are noisy, assay-dependent, and often censored.\"},{\"question\":\"What limitation do existing affinity ranking approaches have?\",\"answer\":\"Many rely on fixed, context-independent ranking functions that treat each comparison as an isolated prediction problem, limiting adaptation to antigen-specific binding landscapes.\"},{\"question\":\"How does AbICL enable test-time adaptation without gradient updates?\",\"answer\":\"AbICL combines a pretrained structural encoder with a context ranking head and uses episodic meta-training so that contextual demonstrations guide predictions at inference through support-query reasoning alone.\"}]","AbICL: In-Context Learning for Antigen-Specific Antibody Affinity Ranking | PDF",1784177211,25,{"code":4,"msg":32,"data":33},"ok",{"site_id":25,"language":24,"slug":34,"title":13,"keywords":35,"description":14,"schema_data":36,"social_meta":87,"head_meta":89,"extra_data":91,"updated_unix":29},"abicl-in-context-learning-for-antigen-specific-antibody-affinity-ranking","",{"@graph":37,"@context":86},[38,54,69],{"@type":39,"itemListElement":40},"BreadcrumbList",[41,45,48,51],{"item":42,"name":43,"@type":44,"position":21},"https://docshare.wps.com","Home","ListItem",{"item":46,"name":47,"@type":44,"position":20},"https://docshare.wps.com/document/","Document",{"item":49,"name":12,"@type":44,"position":50},"https://docshare.wps.com/document/research-report/",3,{"item":52,"name":13,"@type":44,"position":53},"https://docshare.wps.com/document/abicl-in-context-learning-for-antigen-specific-antibody-affinity-ranking/81944/",4,{"url":52,"name":13,"@type":55,"author":56,"headline":13,"publisher":58,"fileFormat":61,"inLanguage":24,"description":14,"dateModified":62,"datePublished":63,"encodingFormat":61,"isAccessibleForFree":64,"interactionStatistic":65},"DigitalDocument",{"name":9,"@type":57},"Person",{"url":42,"name":59,"@type":60},"DocShare","Organization","application/pdf","2026-07-29","2026-07-16",true,{"@type":66,"interactionType":67,"userInteractionCount":20},"InteractionCounter",{"@type":68},"ViewAction",{"@type":70,"mainEntity":71},"FAQPage",[72,78,82],{"name":73,"@type":74,"acceptedAnswer":75},"Why is antibody affinity ranking important in therapeutic antibody discovery?","Question",{"text":76,"@type":77},"Affinity ranking prioritizes antibody candidates based on relative binding strengths, offering robustness when affinity measurements are noisy, assay-dependent, and often censored.","Answer",{"name":79,"@type":74,"acceptedAnswer":80},"What limitation do existing affinity ranking approaches have?",{"text":81,"@type":77},"Many rely on fixed, context-independent ranking functions that treat each comparison as an isolated prediction problem, limiting adaptation to antigen-specific binding landscapes.",{"name":83,"@type":74,"acceptedAnswer":84},"How does AbICL enable test-time adaptation without gradient updates?",{"text":85,"@type":77},"AbICL combines a pretrained structural encoder with a context ranking head and uses episodic meta-training so that contextual demonstrations guide predictions at inference through support-query reasoning alone.","https://schema.org",{"og:url":52,"og:type":88,"og:title":13,"og:site_name":59,"og:description":14},"article",{"robots":90,"canonical":52},"index,follow",{"doc_id":7,"site_id":25},{"code":4,"msg":5,"data":93},[94,98,102,106,111,116,121,124,129,132,135],{"id":21,"doc_module":4,"doc_module_name":47,"category_name":95,"show_sort_weight":96,"slug":97},"Story & Novel",90,"story-novel",{"id":20,"doc_module":4,"doc_module_name":47,"category_name":99,"show_sort_weight":100,"slug":101},"Literature",80,"literature",{"id":53,"doc_module":4,"doc_module_name":47,"category_name":103,"show_sort_weight":104,"slug":105},"Exam",70,"exam",{"id":107,"doc_module":4,"doc_module_name":47,"category_name":108,"show_sort_weight":109,"slug":110},5,"Comic",60,"comic",{"id":112,"doc_module":4,"doc_module_name":47,"category_name":113,"show_sort_weight":114,"slug":115},6,"Technology",50,"technology",{"id":117,"doc_module":4,"doc_module_name":47,"category_name":118,"show_sort_weight":119,"slug":120},7,"Healthcare",40,"healthcare",{"id":11,"doc_module":4,"doc_module_name":47,"category_name":12,"show_sort_weight":122,"slug":123},30,"research-report",{"id":125,"doc_module":4,"doc_module_name":47,"category_name":126,"show_sort_weight":127,"slug":128},9,"Religion & Spirituality",20,"religion-spirituality",{"id":127,"doc_module":4,"doc_module_name":47,"category_name":130,"show_sort_weight":127,"slug":131},"World Cup","world-cup",{"id":22,"doc_module":4,"doc_module_name":47,"category_name":133,"show_sort_weight":22,"slug":134},"Lifestyle","lifestyle",{"id":136,"doc_module":4,"doc_module_name":47,"category_name":137,"show_sort_weight":107,"slug":138},19,"General","general"]