{"id":1161796,"date":"2026-02-10T10:32:57","date_gmt":"2026-02-10T18:32:57","guid":{"rendered":"https:\/\/cm-edgetun.pages.dev\/en-us\/research\/?post_type=msr-research-item&#038;p=1161796"},"modified":"2026-02-10T10:45:44","modified_gmt":"2026-02-10T18:45:44","slug":"reducing-the-costs-of-proof-synthesis-on-rust-systems-by-scaling-up-a-seed-training-set","status":"publish","type":"msr-research-item","link":"https:\/\/cm-edgetun.pages.dev\/en-us\/research\/publication\/reducing-the-costs-of-proof-synthesis-on-rust-systems-by-scaling-up-a-seed-training-set\/","title":{"rendered":"Reducing the Costs of Proof Synthesis on Rust Systems by Scaling Up a Seed Training Set"},"content":{"rendered":"<p>Large Language Models (LLMs) are widely used for code generation. However, the correctness of code generated by LLMs remains a concern. A potential remedy to this concern is to have LLMs generate formal correctness proofs along with such code. However, compared with code generation, code-proof generation requires much higher reasoning capability and has much less existing data to learn from. In this paper, we present VeruSyn, a data synthesis pipeline for Verus, a state-of-the-art verification tool for system software written in Rust. Through self-synthesis and tutorial-based synthesis, VeruSyn achieves much larger scale and Verus-feature coverage than previous data-synthesis techniques designed for Verus; VeruSyn also supplements its dataset with long-chain-of-thought (CoT) data through agent trajectory synthesis. With VeruSyn, we synthesize the largest set of Verus verified programs: 6.9 million Rust programs, each with a formal specification and a proof that it meets that specification. This dataset lets us create a fine-tuned Qwen2.5-Coder-32B-Instruct model with appealing cost-proof tradeoff compared with state-of-the-art commercial models like Claude Sonnet 4.5. It also significantly outperforms models like o4-mini and previously proposed research models.<\/p>\n<p>&nbsp;<\/p>\n<p>The training data set produced by this work can be accessed at <a class=\"msr-external-link glyph-append glyph-append-open-in-new-tab glyph-append-xsmall\" rel=\"noopener noreferrer\" target=\"_blank\" href=\"https:\/\/huggingface.co\/datasets\/microsoft\/Verus_Training_Data\">microsoft\/Verus_Training_Data \u00b7 Datasets at Hugging Face<span class=\"sr-only\"> (opens in new tab)<\/span><\/a><\/p>\n","protected":false},"excerpt":{"rendered":"<p>Large Language Models (LLMs) are widely used for code generation. However, the correctness of code generated by LLMs remains a concern. A potential remedy to this concern is to have LLMs generate formal correctness proofs along with such code. However, compared with code generation, code-proof generation requires much higher reasoning capability and has much less [&hellip;]<\/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":"","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":"","msr_release_tracker_id":"","msr_s2_match_type":"","msr_citation_count_updated":"","msr_published_date":"2026-2-4","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":[13556,13560,13547],"msr-publication-type":[193724],"msr-publisher":[],"msr-focus-area":[],"msr-locale":[268875],"msr-post-option":[269148,269142],"msr-field-of-study":[],"msr-conference":[],"msr-journal":[],"msr-impact-theme":[],"msr-pillar":[],"class_list":["post-1161796","msr-research-item","type-msr-research-item","status-publish","hentry","msr-research-area-artificial-intelligence","msr-research-area-programming-languages-software-engineering","msr-research-area-systems-and-networking","msr-locale-en_us","msr-post-option-approved-for-river","msr-post-option-include-in-river"],"msr_publishername":"","msr_edition":"","msr_affiliation":"","msr_published_date":"2026-2-4","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","viewUrl":"false","id":"false","title":"https:\/\/arxiv.org\/abs\/2602.04910","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":"text","value":"Nongyu Di","user_id":0,"rest_url":false},{"type":"user_nicename","value":"Tianyu Chen","user_id":44066,"rest_url":"https:\/\/cm-edgetun.pages.dev\/en-us\/research\/wp-json\/microsoft-research\/v1\/researchers?person=Tianyu Chen"},{"type":"user_nicename","value":"Shan Lu","user_id":43215,"rest_url":"https:\/\/cm-edgetun.pages.dev\/en-us\/research\/wp-json\/microsoft-research\/v1\/researchers?person=Shan Lu"},{"type":"text","value":"Shuai Lu","user_id":0,"rest_url":false},{"type":"user_nicename","value":"Yeyun Gong","user_id":39186,"rest_url":"https:\/\/cm-edgetun.pages.dev\/en-us\/research\/wp-json\/microsoft-research\/v1\/researchers?person=Yeyun Gong"},{"type":"user_nicename","value":"Peng Cheng","user_id":33225,"rest_url":"https:\/\/cm-edgetun.pages.dev\/en-us\/research\/wp-json\/microsoft-research\/v1\/researchers?person=Peng Cheng"},{"type":"user_nicename","value":"Jay Lorch","user_id":32732,"rest_url":"https:\/\/cm-edgetun.pages.dev\/en-us\/research\/wp-json\/microsoft-research\/v1\/researchers?person=Jay Lorch"},{"type":"text","value":"Yuan Yao","user_id":0,"rest_url":false},{"type":"text","value":"Xiaoxing Ma","user_id":0,"rest_url":false}],"msr_impact_theme":[],"msr_research_lab":[199560,199565],"msr_event":[],"msr_group":[144927,920469],"msr_project":[1097610],"publication":[],"video":[],"msr-tool":[],"msr_publication_type":"miscellaneous","related_content":{"projects":[{"ID":1097610,"post_title":"Practical System Verification","post_name":"practical-system-verification","post_type":"msr-project","post_date":"2024-10-25 14:55:19","post_modified":"2025-12-23 11:58:56","post_status":"publish","permalink":"https:\/\/cm-edgetun.pages.dev\/en-us\/research\/project\/practical-system-verification\/","post_excerpt":"Formal verification is a promising approach to eliminate bugs at compile time, before software ships. Unfortunately, verifying the correctness of system software traditionally requires heroic developer effort.&nbsp;In this project, we aim to enable accessible, faster, cheaper verification of rich properties for realistic systems written in Rust using Verus. Verus is an SMT-based tool for formally verifying Rust programs. With Verus, programmers express proofs and specifications using the Rust language, with no need to learn a&hellip;","_links":{"self":[{"href":"https:\/\/cm-edgetun.pages.dev\/en-us\/research\/wp-json\/wp\/v2\/msr-project\/1097610"}]}}]},"_links":{"self":[{"href":"https:\/\/cm-edgetun.pages.dev\/en-us\/research\/wp-json\/wp\/v2\/msr-research-item\/1161796","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/cm-edgetun.pages.dev\/en-us\/research\/wp-json\/wp\/v2\/msr-research-item"}],"about":[{"href":"https:\/\/cm-edgetun.pages.dev\/en-us\/research\/wp-json\/wp\/v2\/types\/msr-research-item"}],"version-history":[{"count":2,"href":"https:\/\/cm-edgetun.pages.dev\/en-us\/research\/wp-json\/wp\/v2\/msr-research-item\/1161796\/revisions"}],"predecessor-version":[{"id":1161799,"href":"https:\/\/cm-edgetun.pages.dev\/en-us\/research\/wp-json\/wp\/v2\/msr-research-item\/1161796\/revisions\/1161799"}],"wp:attachment":[{"href":"https:\/\/cm-edgetun.pages.dev\/en-us\/research\/wp-json\/wp\/v2\/media?parent=1161796"}],"wp:term":[{"taxonomy":"msr-research-highlight","embeddable":true,"href":"https:\/\/cm-edgetun.pages.dev\/en-us\/research\/wp-json\/wp\/v2\/msr-research-highlight?post=1161796"},{"taxonomy":"msr-research-area","embeddable":true,"href":"https:\/\/cm-edgetun.pages.dev\/en-us\/research\/wp-json\/wp\/v2\/research-area?post=1161796"},{"taxonomy":"msr-publication-type","embeddable":true,"href":"https:\/\/cm-edgetun.pages.dev\/en-us\/research\/wp-json\/wp\/v2\/msr-publication-type?post=1161796"},{"taxonomy":"msr-publisher","embeddable":true,"href":"https:\/\/cm-edgetun.pages.dev\/en-us\/research\/wp-json\/wp\/v2\/msr-publisher?post=1161796"},{"taxonomy":"msr-focus-area","embeddable":true,"href":"https:\/\/cm-edgetun.pages.dev\/en-us\/research\/wp-json\/wp\/v2\/msr-focus-area?post=1161796"},{"taxonomy":"msr-locale","embeddable":true,"href":"https:\/\/cm-edgetun.pages.dev\/en-us\/research\/wp-json\/wp\/v2\/msr-locale?post=1161796"},{"taxonomy":"msr-post-option","embeddable":true,"href":"https:\/\/cm-edgetun.pages.dev\/en-us\/research\/wp-json\/wp\/v2\/msr-post-option?post=1161796"},{"taxonomy":"msr-field-of-study","embeddable":true,"href":"https:\/\/cm-edgetun.pages.dev\/en-us\/research\/wp-json\/wp\/v2\/msr-field-of-study?post=1161796"},{"taxonomy":"msr-conference","embeddable":true,"href":"https:\/\/cm-edgetun.pages.dev\/en-us\/research\/wp-json\/wp\/v2\/msr-conference?post=1161796"},{"taxonomy":"msr-journal","embeddable":true,"href":"https:\/\/cm-edgetun.pages.dev\/en-us\/research\/wp-json\/wp\/v2\/msr-journal?post=1161796"},{"taxonomy":"msr-impact-theme","embeddable":true,"href":"https:\/\/cm-edgetun.pages.dev\/en-us\/research\/wp-json\/wp\/v2\/msr-impact-theme?post=1161796"},{"taxonomy":"msr-pillar","embeddable":true,"href":"https:\/\/cm-edgetun.pages.dev\/en-us\/research\/wp-json\/wp\/v2\/msr-pillar?post=1161796"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}