Behavior cloning (BC) is a practical offline imitation learning method, but it often fails when expert demonstrations are limited. Recent works have introduced a class of architectures named predictive inverse dynamics models (PIDM) that combine a future state predictor with an inverse dynamics model (IDM). While PIDM often outperforms BC, the reasons behind its benefits remain unclear. In this paper, we provide a theoretical explanation: PIDM introduces a bias-variance tradeoff. While predicting the future state introduces bias, conditioning the IDM on the prediction can significantly reduce variance. We establish conditions on the state predictor bias for PIDM to achieve lower prediction error and higher sample efficiency than BC, with the gap widening when additional data sources are available. We validate the theoretical insights empirically in 2D navigation tasks, where BC requires up to five times (three times on average) more demonstrations than PIDM to reach comparable performance; and in a complex 3D environment in a modern video game with high-dimensional visual inputs and stochastic transitions, where BC requires over 66% more samples than PIDM.<\/p>\n\n\n\n
<\/p>\n","protected":false},"excerpt":{"rendered":"
Behavior cloning (BC) is a practical offline imitation learning method, but it often fails when expert demonstrations are limited. Recent works have introduced a class of architectures named predictive inverse dynamics models (PIDM) that combine a future state predictor with an inverse dynamics model (IDM). While PIDM often outperforms BC, the reasons behind its benefits […]<\/p>\n","protected":false},"featured_media":0,"template":"","meta":{"msr-url-field":"","msr-podcast-episode":"","msrModifiedDate":"","msrModifiedDateEnabled":false,"ep_exclude_from_search":false,"_classifai_error":"","msr-author-ordering":[{"type":"user_nicename","value":"Lukas Schäfer","user_id":"43602"},{"type":"user_nicename","value":"Pallavi Choudhury","user_id":"33184"},{"type":"user_nicename","value":"Abdelhak Lemkhenter","user_id":"43059"},{"type":"user_nicename","value":"Chris Lovett","user_id":"36027"},{"type":"user_nicename","value":"Somjit Nath","user_id":"43869"},{"type":"user_nicename","value":"Luis Fran\u00e7a","user_id":"40414"},{"type":"text","value":"Matheus Ribeiro Furtado de Mendonça","user_id":0},{"type":"user_nicename","value":"Alex Lamb","user_id":"40822"},{"type":"user_nicename","value":"Riashat Islam","user_id":"44046"},{"type":"user_nicename","value":"Siddhartha Sen","user_id":"33656"},{"type":"user_nicename","value":"John Langford","user_id":"32204"},{"type":"user_nicename","value":"Katja Hofmann","user_id":"32468"},{"type":"user_nicename","value":"Sergio Valcarcel Macua","user_id":"42507"}],"msr_publishername":"Proceedings of Machine Learning Research","msr_publisher_other":"","msr_booktitle":"","msr_chapter":"","msr_edition":"","msr_editors":"","msr_how_published":"","msr_isbn":"","msr_issue":"","msr_journal":"","msr_number":"","msr_organization":"","msr_pages_string":"","msr_page_range_start":"","msr_page_range_end":"","msr_series":"","msr_volume":"","msr_copyright":"","msr_conference_name":"","msr_doi":"","msr_arxiv_id":"","msr_mag_id":"","msr_other_authors":"","msr_other_contributors":"","msr_speaker":"","msr_award":"","msr_affiliation":"","msr_institution":"","msr_host":"","msr_version":"","msr_duration":"","msr_release_tracker_id":"","msr_highlight_type":"","msr_date_display_format":"","msr_main_download_label":"","msr_external_link_label":"","msr_doi_label":"","msr_published_date":"2026","msr_startdate":"","msr_presentation_date":"","msr_highlight_text":"","msr_notes":"","msr_longbiography":"","msr_publicationurl":"","msr_external_url":"","msr_secondary_video_url":"","msr_conference_url":"https:\/\/icml.cc\/","msr_journal_url":"","msr_year":2026,"msr_month":0,"msr_day":0,"msr_microsoftintellectualproperty":true,"msr_pub_id":"","msr_publication_uploader":[{"type":"url","title":"https:\/\/arxiv.org\/pdf\/2601.21718","label":0,"label_id":243109,"id":false,"viewUrl":"http:\/\/false"}],"msr_related_uploader":[{"type":"file","title":"predictive_inverse_dynamics","label":0,"label_id":243112,"id":1161170,"viewUrl":"https:\/\/find.codeghost.online\/en-us\/research\/wp-content\/uploads\/2026\/01\/predictive_inverse_dynamics.pdf"}],"msr_original_fields_of_study":[],"msr_s2_paper_id":"","msr_s2_pdf_url":"","msr_citation_count_updated":"","msr_citation_count":0,"msr_influential_citations":0,"msr_reference_count":0,"msr_s2_open_access":false,"msr_s2_author_ids":[],"msr_pub_ids":[],"msr_hide_image_in_river":0,"footnotes":""},"msr-research-highlight":[],"research-area":[13556],"msr-publication-type":[193716],"msr-publisher":[],"msr-publication-cta":[],"msr-focus-area":[],"msr-locale":[268875],"msr-post-option":[269148,269142],"msr-field-of-study":[246694,246685],"msr-conference":[260284],"msr-journal":[],"msr-impact-theme":[],"msr-pillar":[],"class_list":["post-1161169","msr-research-item","type-msr-research-item","status-publish","hentry","msr-research-area-artificial-intelligence","msr-locale-en_us","msr-post-option-approved-for-river","msr-post-option-include-in-river","msr-field-of-study-artificial-intelligence","msr-field-of-study-machine-learning"],"msr_publishername":"Proceedings of Machine Learning Research","msr_edition":"","msr_affiliation":"","msr_published_date":"2026","msr_host":"","msr_duration":"","msr_version":"","msr_speaker":"","msr_other_contributors":"","msr_booktitle":"","msr_pages_string":"","msr_chapter":"","msr_isbn":"","msr_journal":"","msr_volume":"","msr_number":"","msr_editors":"","msr_series":"","msr_issue":"","msr_organization":"","msr_how_published":"","msr_notes":"","msr_highlight_text":"","msr_release_tracker_id":"","msr_original_fields_of_study":"","msr_download_urls":"","msr_external_url":"","msr_secondary_video_url":"","msr_longbiography":"","msr_microsoftintellectualproperty":1,"msr_main_download":"","msr_publicationurl":"","msr_doi":"","msr_publication_uploader":[{"type":"url","title":"https:\/\/arxiv.org\/pdf\/2601.21718","label":0,"label_id":243109,"id":false,"viewUrl":"http:\/\/false"}],"msr_related_uploader":[{"type":"file","title":"predictive_inverse_dynamics","label":0,"label_id":243112,"id":1161170,"viewUrl":"https:\/\/find.codeghost.online\/en-us\/research\/wp-content\/uploads\/2026\/01\/predictive_inverse_dynamics.pdf"}],"msr_citation_count":0,"msr_citation_count_updated":"","msr_s2_paper_id":"","msr_influential_citations":0,"msr_reference_count":0,"msr_arxiv_id":"","msr_s2_author_ids":[],"msr_s2_open_access":false,"msr_s2_pdf_url":null,"msr_attachments":[{"id":1161170,"url":"https:\/\/find.codeghost.online\/en-us\/research\/wp-content\/uploads\/2026\/01\/predictive_inverse_dynamics.pdf"}],"msr-author-ordering":[{"type":"user_nicename","value":"Lukas Schäfer","user_id":43602,"rest_url":"https:\/\/find.codeghost.online\/en-us\/research\/wp-json\/microsoft-research\/v1\/researchers?person=Lukas Schäfer"},{"type":"user_nicename","value":"Pallavi Choudhury","user_id":33184,"rest_url":"https:\/\/find.codeghost.online\/en-us\/research\/wp-json\/microsoft-research\/v1\/researchers?person=Pallavi Choudhury"},{"type":"user_nicename","value":"Abdelhak Lemkhenter","user_id":43059,"rest_url":"https:\/\/find.codeghost.online\/en-us\/research\/wp-json\/microsoft-research\/v1\/researchers?person=Abdelhak Lemkhenter"},{"type":"user_nicename","value":"Chris Lovett","user_id":36027,"rest_url":"https:\/\/find.codeghost.online\/en-us\/research\/wp-json\/microsoft-research\/v1\/researchers?person=Chris Lovett"},{"type":"user_nicename","value":"Somjit Nath","user_id":43869,"rest_url":"https:\/\/find.codeghost.online\/en-us\/research\/wp-json\/microsoft-research\/v1\/researchers?person=Somjit Nath"},{"type":"user_nicename","value":"Luis Fran\u00e7a","user_id":40414,"rest_url":"https:\/\/find.codeghost.online\/en-us\/research\/wp-json\/microsoft-research\/v1\/researchers?person=Luis Fran\u00e7a"},{"type":"text","value":"Matheus Ribeiro Furtado de Mendonça","user_id":0,"rest_url":false},{"type":"user_nicename","value":"Alex Lamb","user_id":40822,"rest_url":"https:\/\/find.codeghost.online\/en-us\/research\/wp-json\/microsoft-research\/v1\/researchers?person=Alex Lamb"},{"type":"user_nicename","value":"Riashat Islam","user_id":44046,"rest_url":"https:\/\/find.codeghost.online\/en-us\/research\/wp-json\/microsoft-research\/v1\/researchers?person=Riashat Islam"},{"type":"user_nicename","value":"Siddhartha Sen","user_id":33656,"rest_url":"https:\/\/find.codeghost.online\/en-us\/research\/wp-json\/microsoft-research\/v1\/researchers?person=Siddhartha Sen"},{"type":"user_nicename","value":"John Langford","user_id":32204,"rest_url":"https:\/\/find.codeghost.online\/en-us\/research\/wp-json\/microsoft-research\/v1\/researchers?person=John Langford"},{"type":"user_nicename","value":"Katja Hofmann","user_id":32468,"rest_url":"https:\/\/find.codeghost.online\/en-us\/research\/wp-json\/microsoft-research\/v1\/researchers?person=Katja Hofmann"},{"type":"user_nicename","value":"Sergio Valcarcel Macua","user_id":42507,"rest_url":"https:\/\/find.codeghost.online\/en-us\/research\/wp-json\/microsoft-research\/v1\/researchers?person=Sergio Valcarcel Macua"}],"msr_impact_theme":[],"msr_research_lab":[199561,199565],"msr_event":[1173513],"msr_group":[],"msr_project":[1173527],"publication":[1179004],"video":[],"msr-tool":[],"msr_publication_type":"inproceedings","related_content":{"projects":[{"ID":1173527,"post_title":"PIDM: Predictive Inverse Dynamic Models","post_name":"pidm-predictive-inverse-dynamic-models","post_type":"msr-project","post_date":"2026-07-10 08:30:50","post_modified":"2026-07-10 08:44:40","post_status":"publish","permalink":"https:\/\/find.codeghost.online\/en-us\/research\/project\/pidm-predictive-inverse-dynamic-models\/","post_excerpt":"Imitation learning enables agents to learn complex behaviour from demonstrations, but in practice it often requires large datasets that are costly or impractical to collect. Our project studies how to make imitation learning significantly more data-efficient, enabling agents to learn effectively from limited demonstrations in complex, high-dimensional environments. We use real-world problems such as modern video games as testbeds to develop and study these methods under realistic conditions. Our agents are able to mimic complex…","_links":{"self":[{"href":"https:\/\/find.codeghost.online\/en-us\/research\/wp-json\/wp\/v2\/msr-project\/1173527"}]}}]},"_links":{"self":[{"href":"https:\/\/find.codeghost.online\/en-us\/research\/wp-json\/wp\/v2\/msr-research-item\/1161169","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/find.codeghost.online\/en-us\/research\/wp-json\/wp\/v2\/msr-research-item"}],"about":[{"href":"https:\/\/find.codeghost.online\/en-us\/research\/wp-json\/wp\/v2\/types\/msr-research-item"}],"version-history":[{"count":3,"href":"https:\/\/find.codeghost.online\/en-us\/research\/wp-json\/wp\/v2\/msr-research-item\/1161169\/revisions"}],"predecessor-version":[{"id":1179012,"href":"https:\/\/find.codeghost.online\/en-us\/research\/wp-json\/wp\/v2\/msr-research-item\/1161169\/revisions\/1179012"}],"wp:attachment":[{"href":"https:\/\/find.codeghost.online\/en-us\/research\/wp-json\/wp\/v2\/media?parent=1161169"}],"wp:term":[{"taxonomy":"msr-research-highlight","embeddable":true,"href":"https:\/\/find.codeghost.online\/en-us\/research\/wp-json\/wp\/v2\/msr-research-highlight?post=1161169"},{"taxonomy":"msr-research-area","embeddable":true,"href":"https:\/\/find.codeghost.online\/en-us\/research\/wp-json\/wp\/v2\/research-area?post=1161169"},{"taxonomy":"msr-publication-type","embeddable":true,"href":"https:\/\/find.codeghost.online\/en-us\/research\/wp-json\/wp\/v2\/msr-publication-type?post=1161169"},{"taxonomy":"msr-publisher","embeddable":true,"href":"https:\/\/find.codeghost.online\/en-us\/research\/wp-json\/wp\/v2\/msr-publisher?post=1161169"},{"taxonomy":"msr-publication-cta","embeddable":true,"href":"https:\/\/find.codeghost.online\/en-us\/research\/wp-json\/wp\/v2\/msr-publication-cta?post=1161169"},{"taxonomy":"msr-focus-area","embeddable":true,"href":"https:\/\/find.codeghost.online\/en-us\/research\/wp-json\/wp\/v2\/msr-focus-area?post=1161169"},{"taxonomy":"msr-locale","embeddable":true,"href":"https:\/\/find.codeghost.online\/en-us\/research\/wp-json\/wp\/v2\/msr-locale?post=1161169"},{"taxonomy":"msr-post-option","embeddable":true,"href":"https:\/\/find.codeghost.online\/en-us\/research\/wp-json\/wp\/v2\/msr-post-option?post=1161169"},{"taxonomy":"msr-field-of-study","embeddable":true,"href":"https:\/\/find.codeghost.online\/en-us\/research\/wp-json\/wp\/v2\/msr-field-of-study?post=1161169"},{"taxonomy":"msr-conference","embeddable":true,"href":"https:\/\/find.codeghost.online\/en-us\/research\/wp-json\/wp\/v2\/msr-conference?post=1161169"},{"taxonomy":"msr-journal","embeddable":true,"href":"https:\/\/find.codeghost.online\/en-us\/research\/wp-json\/wp\/v2\/msr-journal?post=1161169"},{"taxonomy":"msr-impact-theme","embeddable":true,"href":"https:\/\/find.codeghost.online\/en-us\/research\/wp-json\/wp\/v2\/msr-impact-theme?post=1161169"},{"taxonomy":"msr-pillar","embeddable":true,"href":"https:\/\/find.codeghost.online\/en-us\/research\/wp-json\/wp\/v2\/msr-pillar?post=1161169"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}