How effectively can LLM-based AI assistants utilize their memory (context) to perform various tasks? Traditional data benchmarks, which are often manually crafted, suffer from several limitations: they are static, susceptible to overfitting, difficult to interpret, and lack actionable insights–failing to pinpoint the specific capabilities a model lacks when it does not pass a test. In this paper, we present a framework for automatically generating a comprehensive set of tests to evaluate models’ abilities to use their memory effectively. Our framework extends the range of capability tests beyond the commonly explored (passkey, key-value, needle in the haystack) search, a dominant focus in the literature. Specifically, we evaluate models on atomic tasks such as searching, recalling, editing, matching, comparing information in context memory, and performing basic operations when inputs are structured into distinct blocks, simulating real-world data. Additionally, we design composite tests to investigate the models’ ability to maintain state while operating on memory. Our benchmark enables an interpretable, detailed assessment of memory capabilities of LLMs.<\/p>\n","protected":false},"excerpt":{"rendered":"
How effectively can LLM-based AI assistants utilize their memory (context) to perform various tasks? Traditional data benchmarks, which are often manually crafted, suffer from several limitations: they are static, susceptible to overfitting, difficult to interpret, and lack actionable insights–failing to pinpoint the specific capabilities a model lacks when it does not pass a test. In […]<\/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":null,"msr_publishername":"","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":"ICML 2025","msr_doi":"","msr_arxiv_id":"","msr_s2_paper_id":"","msr_mag_id":"","msr_pubmed_id":"","msr_other_authors":"","msr_other_contributors":"","msr_speaker":"","msr_award":"","msr_affiliation":"","msr_institution":"","msr_host":"","msr_version":"","msr_duration":"","msr_original_fields_of_study":null,"msr_release_tracker_id":"","msr_s2_match_type":"","msr_citation_count_updated":"","msr_published_date":"2025-7-13","msr_highlight_text":"","msr_notes":"","msr_longbiography":"","msr_publicationurl":"","msr_external_url":"","msr_secondary_video_url":"","msr_conference_url":"","msr_journal_url":"","msr_s2_pdf_url":"","msr_year":0,"msr_citation_count":0,"msr_influential_citations":0,"msr_reference_count":0,"msr_s2_match_confidence":0,"msr_microsoftintellectualproperty":true,"msr_s2_open_access":false,"msr_s2_author_ids":[],"msr_pub_ids":[],"msr_hide_image_in_river":null,"footnotes":""},"msr-research-highlight":[],"research-area":[13556],"msr-publication-type":[193716],"msr-publisher":[],"msr-focus-area":[],"msr-locale":[268875],"msr-post-option":[269148,269142],"msr-field-of-study":[246691,268089],"msr-conference":[260284],"msr-journal":[],"msr-impact-theme":[],"msr-pillar":[],"class_list":["post-1140724","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-computer-science","msr-field-of-study-large-language-models"],"msr_publishername":"","msr_edition":"","msr_affiliation":"","msr_published_date":"2025-7-13","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":"doi","viewUrl":"false","id":"false","title":"https:\/\/doi.org\/10.48550\/arXiv.2502.03358","label_id":"243106","label":0},{"type":"url","viewUrl":"false","id":"false","title":"https:\/\/arxiv.org\/abs\/2502.03358","label_id":"243109","label":0},{"type":"url","viewUrl":"false","id":"false","title":"https:\/\/dblp.org\/rec\/journals\/corr\/abs-2502-03358.html","label_id":"243109","label":0}],"msr_related_uploader":"","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":[],"msr-author-ordering":[{"type":"user_nicename","value":"Molly Xia","user_id":41943,"rest_url":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/microsoft-research\/v1\/researchers?person=Molly Xia"},{"type":"user_nicename","value":"Victor Ruehle","user_id":41027,"rest_url":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/microsoft-research\/v1\/researchers?person=Victor Ruehle"},{"type":"user_nicename","value":"Saravan Rajmohan","user_id":41039,"rest_url":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/microsoft-research\/v1\/researchers?person=Saravan Rajmohan"},{"type":"text","value":"Reza Shokri","user_id":0,"rest_url":false}],"msr_impact_theme":[],"msr_research_lab":[],"msr_event":[1140057],"msr_group":[1145968],"msr_project":[1150291],"publication":[],"video":[],"msr-tool":[],"msr_publication_type":"inproceedings","related_content":{"projects":[{"ID":1150291,"post_title":"Efficient AI applications: context engineering and agents","post_name":"efficient-ai-applications-context-engineering-and-agents","post_type":"msr-project","post_date":"2025-10-22 10:21:05","post_modified":"2025-11-07 05:25:17","post_status":"publish","permalink":"https:\/\/www.microsoft.com\/en-us\/research\/project\/efficient-ai-applications-context-engineering-and-agents\/","post_excerpt":"Modern AI systems face a dual challenge: delivering high\u2011quality outputs while staying cost- and latency\u2011efficient. Every token processed and every millisecond of compute impacts scalability, user experience, and sustainability. Efficiency isn\u2019t just an optimisation, it\u2019s a design principle that makes AI applications feasible and scalable. Efficient AI applications start with the right context. Identifying relevant information, reducing redundancy, and maintaining long-term memory are key to effective performance. Our research combines structured and unstructured context pruning,…","_links":{"self":[{"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-project\/1150291"}]}}]},"_links":{"self":[{"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-research-item\/1140724","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-research-item"}],"about":[{"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/types\/msr-research-item"}],"version-history":[{"count":1,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-research-item\/1140724\/revisions"}],"predecessor-version":[{"id":1140730,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-research-item\/1140724\/revisions\/1140730"}],"wp:attachment":[{"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/media?parent=1140724"}],"wp:term":[{"taxonomy":"msr-research-highlight","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-research-highlight?post=1140724"},{"taxonomy":"msr-research-area","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/research-area?post=1140724"},{"taxonomy":"msr-publication-type","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-publication-type?post=1140724"},{"taxonomy":"msr-publisher","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-publisher?post=1140724"},{"taxonomy":"msr-focus-area","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-focus-area?post=1140724"},{"taxonomy":"msr-locale","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-locale?post=1140724"},{"taxonomy":"msr-post-option","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-post-option?post=1140724"},{"taxonomy":"msr-field-of-study","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-field-of-study?post=1140724"},{"taxonomy":"msr-conference","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-conference?post=1140724"},{"taxonomy":"msr-journal","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-journal?post=1140724"},{"taxonomy":"msr-impact-theme","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-impact-theme?post=1140724"},{"taxonomy":"msr-pillar","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-pillar?post=1140724"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}