[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-85683-en":3,"doc-seo-85683-105":29,"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":20,"is_downloadable":20,"audit_status":20,"page_count":21,"language":22,"language_code":23,"site_id":24,"html_lang":23,"table_of_contents":25,"faqs":26,"seo_title":13,"seo_description":14,"update_tm":27,"read_time":28},85683,549758252649,"Ivy","https://ap-avatar.wpscdn.com/avatar/8000253669c5317157?_k=1778319167496531819",8,"Research & Report","YUKTI: From Natural-Language Situations to Robust, Verifiable Decisions","Language models increasingly convert natural-language situations into numeric plans, yet dominant pipelines commit to a single objective and point-valued coefficients, then solve once—making confidence a failure mode for real budgets, field effort, or clinical attention. Every objectified number becomes an assumption, yielding fragile recommendations that look optimized without guaranteeing substance. YUKTI re-targets autoformulation using a typed-proposition graph with coefficient uncertainty, distributional multi-stage Pareto hand-offs, and Assumption-Robust Pareto Frontiers to measure decision survival via ρ, enabling regret-bounded, auditable recommendations.","arXiv :2607 .09706v 1 [ cs .AI] 22 Jun 2026  \nYUKTI: From Natural-Language Situations to Robust, Verifiable Decisions  \nAn Uncertainty-Typed Proposition IR, Assumption-Robust Pareto Frontiers, and a Regret Certificate—why a language model should  \nformulate, not solve  \nSuyash Mishra  \nAI Researcher  \n{[zurich.suyash@gmail.com}](zurich.suyash@gmail.com})  \nJune 2026  \nAbstract  \nLanguage models increasingly turn a worded situation into a numeric plan—and the dominant pipelines (NL4Opt, OptiMUS, ORLM/LLMOPT, OR-LLM-Agent) commit to a single objective and point-valued coefficients, then solve once. For decisions that allocate real budget, real field effort, or real clinical attention, that confidence is the failure mode: every objectified number is an assumption, and a plan that is optimal only if the guesses are exactly right is silently fragile. We name the risk mimicry of computation—output with the form of an optimized recommendation but none of its substance.  \nWe present YUKTI, which changes the target of autoformulation. Its intermediate representation is a Typed Proposition graph in which every relationship carries a shape prior, a coefficient uncertainty distribution, and provenance. On it, YUKTI (i) routes each stage to an exact, nonlinear, or evolutionary solver by probing structure; (ii) couples stages through a distributional Pareto hand-off; (iii) introduces Assumption-Robust Pareto Frontiers (ARPF), resampling the assumptions—including structural, ε-contamination misspecification—to score how often each action survives (ρ); and (iv) when no data exist, synthesizes a benchmark-faithful foundation (SRJANA) . We prove a regret bound making ρ an exact factor of decision regret, and we surface decision traceability—which segments compose an action and which constraint binds—so a recommendation is auditable, not merely persuasive.  \nWe validate the central claim three ways. Under controlled structural misspecification, the robust compromise cuts mean and tail regret by over 90% versus a naive point plan. On a regulated commercial decision (an oncology brand during loss of exclusivity), we optimize inside a lawful—MLR/consent/frequency/KAM —action space against a prescribing-proxy objective and price the downside in euros. Finally, on a real public dataset of 41 , 188 marketing decisions we did not generate, an out-of-sample backtest shows the robust rule beating the logged status quo by 34% and the naive point rule by 4% while measurably reducing the optimizer’s curse. The solvers are standard and we make no benchmark-SOTA claim; the contribution is the uncertainty-typed IR, ARPF and its regret bound, the distributional multi-stage hand-off, decision-traceability reporting, and a deployment posture as a decision stresstesting layer—not a system of record. A head-to-head locates the limit of LLM reasoning precisely: an LRM handed the correct numbers, and classical single-objective optimization, both incur ∼47× the held-out regret of YUKTI’s robust compromise—an LRM is a formulator, not a solver. Finally we show where the formalism itself must change: under long-range causal coupling between actions, the forward multi-stage hand-off becomes unsound—its regret grows with coupling and its per-stage certificate turns optimistic—locating the boundary at which a hand-off must become a backward-induction causal policy.  \nPart I | The Problem  \n1 Introduction  \nA recurring request from decision-makers is deceptively simple: “here is my situation in words—tell me what todo.” Underneath it sits an optimization problem that  \nnobody has written down. The decision-maker knows the levers they control, the signals they care about, and the limits they must respect, but not the variables, objective, and constraints of a formal program. Recent work has shown that language models can bridge this gap: given a description, they can extract variables, constraints, and an objective and emit solver code [1, 3 , 5] .  \nYet the prevailing formul","cbCaisR5tUlQore5","https://ap.wps.com/l/cbCaisR5tUlQore5","pdf",1996323,1,19,"English","en",105,"# Abstract\n# Part I | The Problem\n## Introduction","[{\"question\":\"Why are single-objective, point-valued autoformulations fragile in high-stakes decisions?\",\"answer\":\"Because invented coefficients are assumptions; a plan optimized only for those assumed values can be silently brittle, offering recommendations that look optimal without exposing their uncertainty.\"},{\"question\":\"What is YUKTI’s core idea for changing the autoformulation target?\",\"answer\":\"YUKTI formulates an intermediate Typed Proposition representation that carries coefficient uncertainty, provenance, and shape priors, then uses uncertainty-aware multi-stage hand-offs and assumption-robust evaluation.\"},{\"question\":\"How does YUKTI make recommendations auditable rather than merely persuasive?\",\"answer\":\"It reports decision traceability by segmenting which components compose an action and which constraint binds, linking the recommendation to survivability metrics under the assumed uncertainty (ρ).\"}]",1784205582,48,{"code":4,"msg":30,"data":31},"ok",{"site_id":24,"language":23,"slug":32,"title":13,"keywords":33,"description":14,"schema_data":34,"social_meta":86,"head_meta":88,"extra_data":90,"updated_unix":27},"yukti-from-natural-language-situations-to-robust-verifiable-decisions","",{"@graph":35,"@context":85},[36,53,68],{"@type":37,"itemListElement":38},"BreadcrumbList",[39,43,47,50],{"item":40,"name":41,"@type":42,"position":20},"https://docshare.wps.com","Home","ListItem",{"item":44,"name":45,"@type":42,"position":46},"https://docshare.wps.com/document/","Document",2,{"item":48,"name":12,"@type":42,"position":49},"https://docshare.wps.com/document/research-report/",3,{"item":51,"name":13,"@type":42,"position":52},"https://docshare.wps.com/document/yukti-from-natural-language-situations-to-robust-verifiable-decisions/85683/",4,{"url":51,"name":13,"@type":54,"author":55,"headline":13,"publisher":57,"fileFormat":60,"inLanguage":23,"description":14,"dateModified":61,"datePublished":62,"encodingFormat":60,"isAccessibleForFree":63,"interactionStatistic":64},"DigitalDocument",{"name":9,"@type":56},"Person",{"url":40,"name":58,"@type":59},"DocShare","Organization","application/pdf","2026-07-17","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},"Why are single-objective, point-valued autoformulations fragile in high-stakes decisions?","Question",{"text":75,"@type":76},"Because invented coefficients are assumptions; a plan optimized only for those assumed values can be silently brittle, offering recommendations that look optimal without exposing their uncertainty.","Answer",{"name":78,"@type":73,"acceptedAnswer":79},"What is YUKTI’s core idea for changing the autoformulation target?",{"text":80,"@type":76},"YUKTI formulates an intermediate Typed Proposition representation that carries coefficient uncertainty, provenance, and shape priors, then uses uncertainty-aware multi-stage hand-offs and assumption-robust evaluation.",{"name":82,"@type":73,"acceptedAnswer":83},"How does YUKTI make recommendations auditable rather than merely persuasive?",{"text":84,"@type":76},"It reports decision traceability by segmenting which components compose an action and which constraint binds, linking the recommendation to survivability metrics under the assumed uncertainty (ρ).","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":24},{"code":4,"msg":5,"data":92},[93,97,101,105,110,115,120,123,128,131,135],{"id":20,"doc_module":4,"doc_module_name":45,"category_name":94,"show_sort_weight":95,"slug":96},"Story & Novel",90,"story-novel",{"id":46,"doc_module":4,"doc_module_name":45,"category_name":98,"show_sort_weight":99,"slug":100},"Literature",80,"literature",{"id":52,"doc_module":4,"doc_module_name":45,"category_name":102,"show_sort_weight":103,"slug":104},"Exam",70,"exam",{"id":106,"doc_module":4,"doc_module_name":45,"category_name":107,"show_sort_weight":108,"slug":109},5,"Comic",60,"comic",{"id":111,"doc_module":4,"doc_module_name":45,"category_name":112,"show_sort_weight":113,"slug":114},6,"Technology",50,"technology",{"id":116,"doc_module":4,"doc_module_name":45,"category_name":117,"show_sort_weight":118,"slug":119},7,"Healthcare",40,"healthcare",{"id":11,"doc_module":4,"doc_module_name":45,"category_name":12,"show_sort_weight":121,"slug":122},30,"research-report",{"id":124,"doc_module":4,"doc_module_name":45,"category_name":125,"show_sort_weight":126,"slug":127},9,"Religion & Spirituality",20,"religion-spirituality",{"id":126,"doc_module":4,"doc_module_name":45,"category_name":129,"show_sort_weight":126,"slug":130},"World Cup","world-cup",{"id":132,"doc_module":4,"doc_module_name":45,"category_name":133,"show_sort_weight":132,"slug":134},10,"Lifestyle","lifestyle",{"id":21,"doc_module":4,"doc_module_name":45,"category_name":136,"show_sort_weight":106,"slug":137},"General","general"]