{"id":103947,"date":"2026-07-13T19:15:40","date_gmt":"2026-07-13T19:15:40","guid":{"rendered":"https:\/\/youzum.net\/stanford-researchers-introduce-trace-a-capability-targeted-agentic-training-system-that-turns-recurrent-agent-failures-into-synthetic-rl-environment\/"},"modified":"2026-07-13T19:15:40","modified_gmt":"2026-07-13T19:15:40","slug":"stanford-researchers-introduce-trace-a-capability-targeted-agentic-training-system-that-turns-recurrent-agent-failures-into-synthetic-rl-environment","status":"publish","type":"post","link":"https:\/\/youzum.net\/it\/stanford-researchers-introduce-trace-a-capability-targeted-agentic-training-system-that-turns-recurrent-agent-failures-into-synthetic-rl-environment\/","title":{"rendered":"Stanford Researchers Introduce TRACE: A Capability-Targeted Agentic Training System That Turns Recurrent Agent Failures Into Synthetic RL Environment"},"content":{"rendered":"<p class=\"wp-block-paragraph\">Agentic LLMs often fail the same way, again and again. A Stanford research team traced this to missing, reusable capabilities. Their system, <strong>TRACE<\/strong>, diagnoses those gaps and trains for them directly.<\/p>\n<p class=\"wp-block-paragraph\">TRACE stands for <strong>T<\/strong>urning <strong>R<\/strong>ecurrent <strong>A<\/strong>gent failures into <strong>C<\/strong>apability-targeted training <strong>E<\/strong>nvironments. It was released open-source under an MIT license.<\/p>\n<h2 class=\"wp-block-heading\"><strong>What problem does TRACE solve?<\/strong><\/h2>\n<p class=\"wp-block-paragraph\">To understand the design, first consider why agents fail. They lack specific skills that tasks demand, like retrieving the right record or verifying a precondition.<\/p>\n<p class=\"wp-block-paragraph\">Two mainstream fixes spend compute poorly. Direct RL or SFT gives sparse rewards that never say which skill was missing. Broad synthetic data is untargeted, so budget flows to skills the model already has.<\/p>\n<p class=\"wp-block-paragraph\">However, TRACE observes that failures are not random. A small set of deficits accounts for most failed trajectories. Therefore, each recurring deficit can become its own dense, verifiable training signal.<\/p>\n<h2 class=\"wp-block-heading\"><strong>How does TRACE work?<\/strong><\/h2>\n<p class=\"wp-block-paragraph\">Given that findings, TRACE runs an automated four-step pipeline. Each step is driven by an LLM agent following a markdown prompt.<\/p>\n<h4 class=\"wp-block-heading\"><strong>Step 1: Contrastive capability analysis<\/strong><\/h4>\n<p class=\"wp-block-paragraph\">The base agent generates rollouts in the target environment. An analysis agent splits them into successful and failed sets. It then labels every trajectory-capability pair as <code>NA<\/code>, <code>PRESENT<\/code>, or <code>LACKING<\/code>.<\/p>\n<p class=\"wp-block-paragraph\">A capability is retained only when it is contrastive and high-coverage. Specifically, its contrastive gap must clear \u03b4 = 0.20 and coverage must clear \u03c1 = 0.10. Consequently, the pipeline keeps skills whose absence concentrates in failures.<\/p>\n<h4 class=\"wp-block-heading\"><strong>Step 2: Targeted environment synthesis<\/strong><\/h4>\n<p class=\"wp-block-paragraph\">Next, a generation agent builds one synthetic environment per retained capability. Each environment isolates a single capability while preserving the target\u2019s tool schemas and format.<\/p>\n<p class=\"wp-block-paragraph\">Task instances are procedurally generated from random seeds. Because generation and verification are algorithmic, rewards need no human labels or LLM judge.<\/p>\n<h4 class=\"wp-block-heading\"><strong>Step 3: Capability adapter training<\/strong><\/h4>\n<p class=\"wp-block-paragraph\">Then each capability gets one LoRA (Low-Rank Adaptation) adapter, trained on its synthetic environment. The training algorithm is GRPO (Group Relative Policy Optimization). The base model stays frozen throughout.<\/p>\n<p class=\"wp-block-paragraph\">GRPO groups rollouts by shared seed, so scenarios are identical within a group. Rewards are then normalized within each group to isolate the policy\u2019s contribution.<\/p>\n<h4 class=\"wp-block-heading\"><strong>Step 4: MoE composition with token-level routing<\/strong><\/h4>\n<p class=\"wp-block-paragraph\">Finally, TRACE composes the adapters into a Mixture-of-Experts (MoE) model. The backbone and adapters stay frozen, and only lightweight token-level gates are trained.<\/p>\n<p class=\"wp-block-paragraph\">At inference, each token is routed top-1 to a single capability adapter. This lets the model switch experts mid-trajectory.<\/p>\n<p class=\"wp-block-paragraph\">\n<\/p><p><title>How TRACE Works \u2014 Interactive Explainer<\/title><\/p>\n<div>\n<div class=\"te-eyebrow\">Interactive Explainer<\/div>\n<h1>How TRACE Turns Agent Failures Into Targeted Training<\/h1>\n<p class=\"te-sub\">TRACE diagnoses the capabilities an agent lacks, builds one verifiable environment per gap, trains a LoRA expert for each, then routes tokens across experts. Step through the pipeline below.<\/p>\n<div class=\"te-steps\"><\/div>\n<div class=\"te-stage\">\n    <!-- STEP 1 -->\n<div class=\"te-view\" data-view=\"0\">\n<div class=\"te-caption\"><span class=\"te-dot\"><\/span>\n<div>\n<h3>1 \u00b7 Contrastive Capability Analysis<\/h3>\n<p>Split rollouts into pass \/ fail, then keep gaps that separate them.<\/p>\n<\/div>\n<\/div>\n<div class=\"te-traj-wrap\">\n<div>\n<div class=\"te-pool\">\n<h4>Passed \u00a0(D\u207a)<\/h4>\n<div><\/div>\n<\/div>\n<div class=\"te-pool\">\n<h4>Failed \u00a0(D\u207b)<\/h4>\n<div><\/div>\n<\/div><\/div>\n<div class=\"te-caps\">\n<h4>Retained if \u0394 \u2265 0.20 and Cov \u2265 0.10<\/h4>\n<div><\/div>\n<\/div>\n<\/div>\n<\/div>\n<p>    <!-- STEP 2 --><\/p>\n<div class=\"te-view\" data-view=\"1\">\n<div class=\"te-caption\"><span class=\"te-dot\"><\/span>\n<div>\n<h3>2 \u00b7 Targeted Environment Synthesis<\/h3>\n<p>One seeded, auto-verifiable environment is generated per capability.<\/p>\n<\/div>\n<\/div>\n<div class=\"te-envgrid\"><\/div>\n<\/div>\n<p>    <!-- STEP 3 --><\/p>\n<div class=\"te-view\" data-view=\"2\">\n<div class=\"te-caption\"><span class=\"te-dot\"><\/span>\n<div>\n<h3>3 \u00b7 Capability Adapter Training (GRPO)<\/h3>\n<p>Rollouts share a seed; rewards are normalized within the group.<\/p>\n<\/div>\n<\/div>\n<div class=\"te-grpo\">\n<div class=\"te-group\"><\/div>\n<div class=\"te-adapter\">\n<div class=\"te-lora\">\n<div class=\"te-fill\"><\/div>\n<p><b>0%<\/b><span>LoRA \u0394c \u00a0(~5.3%)<\/span><\/p><\/div>\n<div class=\"te-frozen\"><img decoding=\"async\" src=\"https:\/\/s.w.org\/images\/core\/emoji\/17.0.2\/72x72\/1f512.png\" alt=\"\ud83d\udd12\" class=\"wp-smiley\" \/> Base model frozen \u00b7 only \u0394c updates<\/div>\n<\/div>\n<\/div>\n<\/div>\n<p>    <!-- STEP 4 --><\/p>\n<div class=\"te-view\" data-view=\"3\">\n<div class=\"te-caption\"><span class=\"te-dot\"><\/span>\n<div>\n<h3>4 \u00b7 MoE Composition \u00b7 Token-Level Routing<\/h3>\n<p>A learned gate routes each token top-1 to a single capability expert.<\/p>\n<\/div>\n<\/div>\n<div class=\"te-route\">\n<div class=\"te-tasks\"><\/div>\n<div class=\"te-tokline\"><span class=\"te-hint\">Pick a task above to route its tokens.<\/span><\/div>\n<div class=\"te-experts\"><\/div>\n<\/div>\n<\/div>\n<div class=\"te-ctrl\">\n      <button class=\"te-btn primary\"><img decoding=\"async\" src=\"https:\/\/s.w.org\/images\/core\/emoji\/17.0.2\/72x72\/25b6.png\" alt=\"\u25b6\" class=\"wp-smiley\" \/> Play step<\/button><br \/>\n      <button class=\"te-btn\">Next step \u2192<\/button><br \/>\n      <span class=\"te-hint\"><\/span>\n    <\/div>\n<\/div>\n<p>  <!-- RESULTS --><\/p>\n<div class=\"te-results\">\n<h3>Result \u00b7 \u03c4\u00b2-Bench overall pass rate (Qwen3-30B-A3B)<\/h3>\n<p class=\"te-rp\">Targeted training and MoE composition beat prompt optimization and single-adapter baselines.<\/p>\n<div><\/div>\n<\/div>\n<p class=\"te-foot\">Built from <a class=\"te-src\" href=\"https:\/\/arxiv.org\/abs\/2604.05336\" target=\"_blank\" rel=\"noopener\">arXiv:2604.05336<\/a> \u00b7 <a class=\"te-src\" href=\"https:\/\/github.com\/ScalingIntelligence\/TRACE\" target=\"_blank\" rel=\"noopener\">code<\/a>. Numbers are from the paper. \u00a0\u2022\u00a0 <b>Marktechpost<\/b><\/p>\n<\/div>","protected":false},"excerpt":{"rendered":"<p>Agentic LLMs often fail the same way, again and again. A Stanford research team traced this to missing, reusable capabilities. Their system, TRACE, diagnoses those gaps and trains for them directly. TRACE stands for Turning Recurrent Agent failures into Capability-targeted training Environments. It was released open-source under an MIT license. What problem does TRACE solve? To understand the design, first consider why agents fail. They lack specific skills that tasks demand, like retrieving the right record or verifying a precondition. Two mainstream fixes spend compute poorly. Direct RL or SFT gives sparse rewards that never say which skill was missing. Broad synthetic data is untargeted, so budget flows to skills the model already has. However, TRACE observes that failures are not random. A small set of deficits accounts for most failed trajectories. Therefore, each recurring deficit can become its own dense, verifiable training signal. How does TRACE work? Given that findings, TRACE runs an automated four-step pipeline. Each step is driven by an LLM agent following a markdown prompt. Step 1: Contrastive capability analysis The base agent generates rollouts in the target environment. An analysis agent splits them into successful and failed sets. It then labels every trajectory-capability pair as NA, PRESENT, or LACKING. A capability is retained only when it is contrastive and high-coverage. Specifically, its contrastive gap must clear \u03b4 = 0.20 and coverage must clear \u03c1 = 0.10. Consequently, the pipeline keeps skills whose absence concentrates in failures. Step 2: Targeted environment synthesis Next, a generation agent builds one synthetic environment per retained capability. Each environment isolates a single capability while preserving the target\u2019s tool schemas and format. Task instances are procedurally generated from random seeds. Because generation and verification are algorithmic, rewards need no human labels or LLM judge. Step 3: Capability adapter training Then each capability gets one LoRA (Low-Rank Adaptation) adapter, trained on its synthetic environment. The training algorithm is GRPO (Group Relative Policy Optimization). The base model stays frozen throughout. GRPO groups rollouts by shared seed, so scenarios are identical within a group. Rewards are then normalized within each group to isolate the policy\u2019s contribution. Step 4: MoE composition with token-level routing Finally, TRACE composes the adapters into a Mixture-of-Experts (MoE) model. The backbone and adapters stay frozen, and only lightweight token-level gates are trained. At inference, each token is routed top-1 to a single capability adapter. This lets the model switch experts mid-trajectory. How TRACE Works \u2014 Interactive Explainer Interactive Explainer How TRACE Turns Agent Failures Into Targeted Training TRACE diagnoses the capabilities an agent lacks, builds one verifiable environment per gap, trains a LoRA expert for each, then routes tokens across experts. Step through the pipeline below. 1 \u00b7 Contrastive Capability Analysis Split rollouts into pass \/ fail, then keep gaps that separate them. Passed \u00a0(D\u207a) Failed \u00a0(D\u207b) Retained if \u0394 \u2265 0.20 and Cov \u2265 0.10 2 \u00b7 Targeted Environment Synthesis One seeded, auto-verifiable environment is generated per capability. 3 \u00b7 Capability Adapter Training (GRPO) Rollouts share a seed; rewards are normalized within the group. 0%LoRA \u0394c \u00a0(~5.3%) Base model frozen \u00b7 only \u0394c updates 4 \u00b7 MoE Composition \u00b7 Token-Level Routing A learned gate routes each token top-1 to a single capability expert. Pick a task above to route its tokens. Play step Next step \u2192 Result \u00b7 \u03c4\u00b2-Bench overall pass rate (Qwen3-30B-A3B) Targeted training and MoE composition beat prompt optimization and single-adapter baselines. Built from arXiv:2604.05336 \u00b7 code. Numbers are from the paper. \u00a0\u2022\u00a0 Marktechpost<\/p>","protected":false},"author":2,"featured_media":0,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"_acf_changed":false,"pmpro_default_level":"","site-sidebar-layout":"default","site-content-layout":"","ast-site-content-layout":"","site-content-style":"default","site-sidebar-style":"default","ast-global-header-display":"","ast-banner-title-visibility":"","ast-main-header-display":"","ast-hfb-above-header-display":"","ast-hfb-below-header-display":"","ast-hfb-mobile-header-display":"","site-post-title":"","ast-breadcrumbs-content":"","ast-featured-img":"","footer-sml-layout":"","theme-transparent-header-meta":"","adv-header-id-meta":"","stick-header-meta":"","header-above-stick-meta":"","header-main-stick-meta":"","header-below-stick-meta":"","astra-migrate-meta-layouts":"default","ast-page-background-enabled":"default","ast-page-background-meta":{"desktop":{"background-color":"var(--ast-global-color-4)","background-image":"","background-repeat":"repeat","background-position":"center center","background-size":"auto","background-attachment":"scroll","background-type":"","background-media":"","overlay-type":"","overlay-color":"","overlay-opacity":"","overlay-gradient":""},"tablet":{"background-color":"","background-image":"","background-repeat":"repeat","background-position":"center center","background-size":"auto","background-attachment":"scroll","background-type":"","background-media":"","overlay-type":"","overlay-color":"","overlay-opacity":"","overlay-gradient":""},"mobile":{"background-color":"","background-image":"","background-repeat":"repeat","background-position":"center center","background-size":"auto","background-attachment":"scroll","background-type":"","background-media":"","overlay-type":"","overlay-color":"","overlay-opacity":"","overlay-gradient":""}},"ast-content-background-meta":{"desktop":{"background-color":"var(--ast-global-color-5)","background-image":"","background-repeat":"repeat","background-position":"center center","background-size":"auto","background-attachment":"scroll","background-type":"","background-media":"","overlay-type":"","overlay-color":"","overlay-opacity":"","overlay-gradient":""},"tablet":{"background-color":"var(--ast-global-color-5)","background-image":"","background-repeat":"repeat","background-position":"center center","background-size":"auto","background-attachment":"scroll","background-type":"","background-media":"","overlay-type":"","overlay-color":"","overlay-opacity":"","overlay-gradient":""},"mobile":{"background-color":"var(--ast-global-color-5)","background-image":"","background-repeat":"repeat","background-position":"center center","background-size":"auto","background-attachment":"scroll","background-type":"","background-media":"","overlay-type":"","overlay-color":"","overlay-opacity":"","overlay-gradient":""}},"_pvb_checkbox_block_on_post":false,"footnotes":""},"categories":[52,5,7,1],"tags":[],"class_list":["post-103947","post","type-post","status-publish","format-standard","hentry","category-ai-club","category-committee","category-news","category-uncategorized","pmpro-has-access"],"acf":[],"yoast_head":"<!-- This site is optimized with the Yoast SEO plugin v25.3 - https:\/\/yoast.com\/wordpress\/plugins\/seo\/ -->\n<title>Stanford Researchers Introduce TRACE: A Capability-Targeted Agentic Training System That Turns Recurrent Agent Failures Into Synthetic RL Environment - YouZum<\/title>\n<meta name=\"description\" content=\"\u0e01\u0e34\u0e08\u0e01\u0e23\u0e23\u0e21\u0e40\u0e01\u0e35\u0e48\u0e22\u0e27\u0e01\u0e31\u0e1a\u0e42\u0e14\u0e23\u0e19\" \/>\n<meta name=\"robots\" content=\"index, follow, max-snippet:-1, max-image-preview:large, max-video-preview:-1\" \/>\n<link rel=\"canonical\" href=\"https:\/\/youzum.net\/it\/stanford-researchers-introduce-trace-a-capability-targeted-agentic-training-system-that-turns-recurrent-agent-failures-into-synthetic-rl-environment\/\" \/>\n<meta property=\"og:locale\" content=\"it_IT\" \/>\n<meta property=\"og:type\" content=\"article\" \/>\n<meta property=\"og:title\" content=\"Stanford Researchers Introduce TRACE: A Capability-Targeted Agentic Training System That Turns Recurrent Agent Failures Into Synthetic RL Environment - YouZum\" \/>\n<meta property=\"og:description\" content=\"\u0e01\u0e34\u0e08\u0e01\u0e23\u0e23\u0e21\u0e40\u0e01\u0e35\u0e48\u0e22\u0e27\u0e01\u0e31\u0e1a\u0e42\u0e14\u0e23\u0e19\" \/>\n<meta property=\"og:url\" content=\"https:\/\/youzum.net\/it\/stanford-researchers-introduce-trace-a-capability-targeted-agentic-training-system-that-turns-recurrent-agent-failures-into-synthetic-rl-environment\/\" \/>\n<meta property=\"og:site_name\" content=\"YouZum\" \/>\n<meta property=\"article:publisher\" content=\"https:\/\/www.facebook.com\/DroneAssociationTH\/\" \/>\n<meta property=\"article:published_time\" content=\"2026-07-13T19:15:40+00:00\" \/>\n<meta property=\"og:image\" content=\"https:\/\/s.w.org\/images\/core\/emoji\/17.0.2\/72x72\/1f512.png\" \/>\n<meta name=\"author\" content=\"admin NU\" \/>\n<meta name=\"twitter:card\" content=\"summary_large_image\" \/>\n<meta name=\"twitter:label1\" content=\"Scritto da\" \/>\n\t<meta name=\"twitter:data1\" content=\"admin NU\" \/>\n\t<meta name=\"twitter:label2\" content=\"Tempo di lettura stimato\" \/>\n\t<meta name=\"twitter:data2\" content=\"3 minuti\" \/>\n<script type=\"application\/ld+json\" class=\"yoast-schema-graph\">{\"@context\":\"https:\/\/schema.org\",\"@graph\":[{\"@type\":\"Article\",\"@id\":\"https:\/\/youzum.net\/stanford-researchers-introduce-trace-a-capability-targeted-agentic-training-system-that-turns-recurrent-agent-failures-into-synthetic-rl-environment\/#article\",\"isPartOf\":{\"@id\":\"https:\/\/youzum.net\/stanford-researchers-introduce-trace-a-capability-targeted-agentic-training-system-that-turns-recurrent-agent-failures-into-synthetic-rl-environment\/\"},\"author\":{\"name\":\"admin NU\",\"@id\":\"https:\/\/yousum.gpucore.co\/#\/schema\/person\/97fa48242daf3908e4d9a5f26f4a059c\"},\"headline\":\"Stanford Researchers Introduce TRACE: A Capability-Targeted Agentic Training System That Turns Recurrent Agent Failures Into Synthetic RL Environment\",\"datePublished\":\"2026-07-13T19:15:40+00:00\",\"mainEntityOfPage\":{\"@id\":\"https:\/\/youzum.net\/stanford-researchers-introduce-trace-a-capability-targeted-agentic-training-system-that-turns-recurrent-agent-failures-into-synthetic-rl-environment\/\"},\"wordCount\":565,\"commentCount\":0,\"publisher\":{\"@id\":\"https:\/\/yousum.gpucore.co\/#organization\"},\"image\":{\"@id\":\"https:\/\/youzum.net\/stanford-researchers-introduce-trace-a-capability-targeted-agentic-training-system-that-turns-recurrent-agent-failures-into-synthetic-rl-environment\/#primaryimage\"},\"thumbnailUrl\":\"https:\/\/s.w.org\/images\/core\/emoji\/17.0.2\/72x72\/1f512.png\",\"articleSection\":[\"AI\",\"Committee\",\"News\",\"Uncategorized\"],\"inLanguage\":\"it-IT\",\"potentialAction\":[{\"@type\":\"CommentAction\",\"name\":\"Comment\",\"target\":[\"https:\/\/youzum.net\/stanford-researchers-introduce-trace-a-capability-targeted-agentic-training-system-that-turns-recurrent-agent-failures-into-synthetic-rl-environment\/#respond\"]}]},{\"@type\":\"WebPage\",\"@id\":\"https:\/\/youzum.net\/stanford-researchers-introduce-trace-a-capability-targeted-agentic-training-system-that-turns-recurrent-agent-failures-into-synthetic-rl-environment\/\",\"url\":\"https:\/\/youzum.net\/stanford-researchers-introduce-trace-a-capability-targeted-agentic-training-system-that-turns-recurrent-agent-failures-into-synthetic-rl-environment\/\",\"name\":\"Stanford Researchers Introduce TRACE: A Capability-Targeted Agentic Training System That Turns Recurrent Agent Failures Into Synthetic RL Environment - YouZum\",\"isPartOf\":{\"@id\":\"https:\/\/yousum.gpucore.co\/#website\"},\"primaryImageOfPage\":{\"@id\":\"https:\/\/youzum.net\/stanford-researchers-introduce-trace-a-capability-targeted-agentic-training-system-that-turns-recurrent-agent-failures-into-synthetic-rl-environment\/#primaryimage\"},\"image\":{\"@id\":\"https:\/\/youzum.net\/stanford-researchers-introduce-trace-a-capability-targeted-agentic-training-system-that-turns-recurrent-agent-failures-into-synthetic-rl-environment\/#primaryimage\"},\"thumbnailUrl\":\"https:\/\/s.w.org\/images\/core\/emoji\/17.0.2\/72x72\/1f512.png\",\"datePublished\":\"2026-07-13T19:15:40+00:00\",\"description\":\"\u0e01\u0e34\u0e08\u0e01\u0e23\u0e23\u0e21\u0e40\u0e01\u0e35\u0e48\u0e22\u0e27\u0e01\u0e31\u0e1a\u0e42\u0e14\u0e23\u0e19\",\"breadcrumb\":{\"@id\":\"https:\/\/youzum.net\/stanford-researchers-introduce-trace-a-capability-targeted-agentic-training-system-that-turns-recurrent-agent-failures-into-synthetic-rl-environment\/#breadcrumb\"},\"inLanguage\":\"it-IT\",\"potentialAction\":[{\"@type\":\"ReadAction\",\"target\":[\"https:\/\/youzum.net\/stanford-researchers-introduce-trace-a-capability-targeted-agentic-training-system-that-turns-recurrent-agent-failures-into-synthetic-rl-environment\/\"]}]},{\"@type\":\"ImageObject\",\"inLanguage\":\"it-IT\",\"@id\":\"https:\/\/youzum.net\/stanford-researchers-introduce-trace-a-capability-targeted-agentic-training-system-that-turns-recurrent-agent-failures-into-synthetic-rl-environment\/#primaryimage\",\"url\":\"https:\/\/s.w.org\/images\/core\/emoji\/17.0.2\/72x72\/1f512.png\",\"contentUrl\":\"https:\/\/s.w.org\/images\/core\/emoji\/17.0.2\/72x72\/1f512.png\"},{\"@type\":\"BreadcrumbList\",\"@id\":\"https:\/\/youzum.net\/stanford-researchers-introduce-trace-a-capability-targeted-agentic-training-system-that-turns-recurrent-agent-failures-into-synthetic-rl-environment\/#breadcrumb\",\"itemListElement\":[{\"@type\":\"ListItem\",\"position\":1,\"name\":\"Home\",\"item\":\"https:\/\/youzum.net\/\"},{\"@type\":\"ListItem\",\"position\":2,\"name\":\"Stanford Researchers Introduce TRACE: A Capability-Targeted Agentic Training System That Turns Recurrent Agent Failures Into Synthetic RL Environment\"}]},{\"@type\":\"WebSite\",\"@id\":\"https:\/\/yousum.gpucore.co\/#website\",\"url\":\"https:\/\/yousum.gpucore.co\/\",\"name\":\"YouSum\",\"description\":\"\",\"publisher\":{\"@id\":\"https:\/\/yousum.gpucore.co\/#organization\"},\"potentialAction\":[{\"@type\":\"SearchAction\",\"target\":{\"@type\":\"EntryPoint\",\"urlTemplate\":\"https:\/\/yousum.gpucore.co\/?s={search_term_string}\"},\"query-input\":{\"@type\":\"PropertyValueSpecification\",\"valueRequired\":true,\"valueName\":\"search_term_string\"}}],\"inLanguage\":\"it-IT\"},{\"@type\":\"Organization\",\"@id\":\"https:\/\/yousum.gpucore.co\/#organization\",\"name\":\"Drone Association Thailand\",\"url\":\"https:\/\/yousum.gpucore.co\/\",\"logo\":{\"@type\":\"ImageObject\",\"inLanguage\":\"it-IT\",\"@id\":\"https:\/\/yousum.gpucore.co\/#\/schema\/logo\/image\/\",\"url\":\"https:\/\/youzum.net\/wp-content\/uploads\/2024\/11\/tranparent-logo.png\",\"contentUrl\":\"https:\/\/youzum.net\/wp-content\/uploads\/2024\/11\/tranparent-logo.png\",\"width\":300,\"height\":300,\"caption\":\"Drone Association Thailand\"},\"image\":{\"@id\":\"https:\/\/yousum.gpucore.co\/#\/schema\/logo\/image\/\"},\"sameAs\":[\"https:\/\/www.facebook.com\/DroneAssociationTH\/\"]},{\"@type\":\"Person\",\"@id\":\"https:\/\/yousum.gpucore.co\/#\/schema\/person\/97fa48242daf3908e4d9a5f26f4a059c\",\"name\":\"admin NU\",\"image\":{\"@type\":\"ImageObject\",\"inLanguage\":\"it-IT\",\"@id\":\"https:\/\/yousum.gpucore.co\/#\/schema\/person\/image\/\",\"url\":\"https:\/\/youzum.net\/wp-content\/uploads\/avatars\/2\/1746849356-bpfull.png\",\"contentUrl\":\"https:\/\/youzum.net\/wp-content\/uploads\/avatars\/2\/1746849356-bpfull.png\",\"caption\":\"admin NU\"},\"url\":\"https:\/\/youzum.net\/it\/members\/adminnu\/\"}]}<\/script>\n<!-- \/ Yoast SEO plugin. -->","yoast_head_json":{"title":"Stanford Researchers Introduce TRACE: A Capability-Targeted Agentic Training System That Turns Recurrent Agent Failures Into Synthetic RL Environment - YouZum","description":"\u0e01\u0e34\u0e08\u0e01\u0e23\u0e23\u0e21\u0e40\u0e01\u0e35\u0e48\u0e22\u0e27\u0e01\u0e31\u0e1a\u0e42\u0e14\u0e23\u0e19","robots":{"index":"index","follow":"follow","max-snippet":"max-snippet:-1","max-image-preview":"max-image-preview:large","max-video-preview":"max-video-preview:-1"},"canonical":"https:\/\/youzum.net\/it\/stanford-researchers-introduce-trace-a-capability-targeted-agentic-training-system-that-turns-recurrent-agent-failures-into-synthetic-rl-environment\/","og_locale":"it_IT","og_type":"article","og_title":"Stanford Researchers Introduce TRACE: A Capability-Targeted Agentic Training System That Turns Recurrent Agent Failures Into Synthetic RL Environment - YouZum","og_description":"\u0e01\u0e34\u0e08\u0e01\u0e23\u0e23\u0e21\u0e40\u0e01\u0e35\u0e48\u0e22\u0e27\u0e01\u0e31\u0e1a\u0e42\u0e14\u0e23\u0e19","og_url":"https:\/\/youzum.net\/it\/stanford-researchers-introduce-trace-a-capability-targeted-agentic-training-system-that-turns-recurrent-agent-failures-into-synthetic-rl-environment\/","og_site_name":"YouZum","article_publisher":"https:\/\/www.facebook.com\/DroneAssociationTH\/","article_published_time":"2026-07-13T19:15:40+00:00","og_image":[{"url":"https:\/\/s.w.org\/images\/core\/emoji\/17.0.2\/72x72\/1f512.png","type":"","width":"","height":""}],"author":"admin NU","twitter_card":"summary_large_image","twitter_misc":{"Scritto da":"admin NU","Tempo di lettura stimato":"3 minuti"},"schema":{"@context":"https:\/\/schema.org","@graph":[{"@type":"Article","@id":"https:\/\/youzum.net\/stanford-researchers-introduce-trace-a-capability-targeted-agentic-training-system-that-turns-recurrent-agent-failures-into-synthetic-rl-environment\/#article","isPartOf":{"@id":"https:\/\/youzum.net\/stanford-researchers-introduce-trace-a-capability-targeted-agentic-training-system-that-turns-recurrent-agent-failures-into-synthetic-rl-environment\/"},"author":{"name":"admin NU","@id":"https:\/\/yousum.gpucore.co\/#\/schema\/person\/97fa48242daf3908e4d9a5f26f4a059c"},"headline":"Stanford Researchers Introduce TRACE: A Capability-Targeted Agentic Training System That Turns Recurrent Agent Failures Into Synthetic RL Environment","datePublished":"2026-07-13T19:15:40+00:00","mainEntityOfPage":{"@id":"https:\/\/youzum.net\/stanford-researchers-introduce-trace-a-capability-targeted-agentic-training-system-that-turns-recurrent-agent-failures-into-synthetic-rl-environment\/"},"wordCount":565,"commentCount":0,"publisher":{"@id":"https:\/\/yousum.gpucore.co\/#organization"},"image":{"@id":"https:\/\/youzum.net\/stanford-researchers-introduce-trace-a-capability-targeted-agentic-training-system-that-turns-recurrent-agent-failures-into-synthetic-rl-environment\/#primaryimage"},"thumbnailUrl":"https:\/\/s.w.org\/images\/core\/emoji\/17.0.2\/72x72\/1f512.png","articleSection":["AI","Committee","News","Uncategorized"],"inLanguage":"it-IT","potentialAction":[{"@type":"CommentAction","name":"Comment","target":["https:\/\/youzum.net\/stanford-researchers-introduce-trace-a-capability-targeted-agentic-training-system-that-turns-recurrent-agent-failures-into-synthetic-rl-environment\/#respond"]}]},{"@type":"WebPage","@id":"https:\/\/youzum.net\/stanford-researchers-introduce-trace-a-capability-targeted-agentic-training-system-that-turns-recurrent-agent-failures-into-synthetic-rl-environment\/","url":"https:\/\/youzum.net\/stanford-researchers-introduce-trace-a-capability-targeted-agentic-training-system-that-turns-recurrent-agent-failures-into-synthetic-rl-environment\/","name":"Stanford Researchers Introduce TRACE: A Capability-Targeted Agentic Training System That Turns Recurrent Agent Failures Into Synthetic RL Environment - YouZum","isPartOf":{"@id":"https:\/\/yousum.gpucore.co\/#website"},"primaryImageOfPage":{"@id":"https:\/\/youzum.net\/stanford-researchers-introduce-trace-a-capability-targeted-agentic-training-system-that-turns-recurrent-agent-failures-into-synthetic-rl-environment\/#primaryimage"},"image":{"@id":"https:\/\/youzum.net\/stanford-researchers-introduce-trace-a-capability-targeted-agentic-training-system-that-turns-recurrent-agent-failures-into-synthetic-rl-environment\/#primaryimage"},"thumbnailUrl":"https:\/\/s.w.org\/images\/core\/emoji\/17.0.2\/72x72\/1f512.png","datePublished":"2026-07-13T19:15:40+00:00","description":"\u0e01\u0e34\u0e08\u0e01\u0e23\u0e23\u0e21\u0e40\u0e01\u0e35\u0e48\u0e22\u0e27\u0e01\u0e31\u0e1a\u0e42\u0e14\u0e23\u0e19","breadcrumb":{"@id":"https:\/\/youzum.net\/stanford-researchers-introduce-trace-a-capability-targeted-agentic-training-system-that-turns-recurrent-agent-failures-into-synthetic-rl-environment\/#breadcrumb"},"inLanguage":"it-IT","potentialAction":[{"@type":"ReadAction","target":["https:\/\/youzum.net\/stanford-researchers-introduce-trace-a-capability-targeted-agentic-training-system-that-turns-recurrent-agent-failures-into-synthetic-rl-environment\/"]}]},{"@type":"ImageObject","inLanguage":"it-IT","@id":"https:\/\/youzum.net\/stanford-researchers-introduce-trace-a-capability-targeted-agentic-training-system-that-turns-recurrent-agent-failures-into-synthetic-rl-environment\/#primaryimage","url":"https:\/\/s.w.org\/images\/core\/emoji\/17.0.2\/72x72\/1f512.png","contentUrl":"https:\/\/s.w.org\/images\/core\/emoji\/17.0.2\/72x72\/1f512.png"},{"@type":"BreadcrumbList","@id":"https:\/\/youzum.net\/stanford-researchers-introduce-trace-a-capability-targeted-agentic-training-system-that-turns-recurrent-agent-failures-into-synthetic-rl-environment\/#breadcrumb","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https:\/\/youzum.net\/"},{"@type":"ListItem","position":2,"name":"Stanford Researchers Introduce TRACE: A Capability-Targeted Agentic Training System That Turns Recurrent Agent Failures Into Synthetic RL Environment"}]},{"@type":"WebSite","@id":"https:\/\/yousum.gpucore.co\/#website","url":"https:\/\/yousum.gpucore.co\/","name":"YouSum","description":"","publisher":{"@id":"https:\/\/yousum.gpucore.co\/#organization"},"potentialAction":[{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https:\/\/yousum.gpucore.co\/?s={search_term_string}"},"query-input":{"@type":"PropertyValueSpecification","valueRequired":true,"valueName":"search_term_string"}}],"inLanguage":"it-IT"},{"@type":"Organization","@id":"https:\/\/yousum.gpucore.co\/#organization","name":"Drone Association Thailand","url":"https:\/\/yousum.gpucore.co\/","logo":{"@type":"ImageObject","inLanguage":"it-IT","@id":"https:\/\/yousum.gpucore.co\/#\/schema\/logo\/image\/","url":"https:\/\/youzum.net\/wp-content\/uploads\/2024\/11\/tranparent-logo.png","contentUrl":"https:\/\/youzum.net\/wp-content\/uploads\/2024\/11\/tranparent-logo.png","width":300,"height":300,"caption":"Drone Association Thailand"},"image":{"@id":"https:\/\/yousum.gpucore.co\/#\/schema\/logo\/image\/"},"sameAs":["https:\/\/www.facebook.com\/DroneAssociationTH\/"]},{"@type":"Person","@id":"https:\/\/yousum.gpucore.co\/#\/schema\/person\/97fa48242daf3908e4d9a5f26f4a059c","name":"admin NU","image":{"@type":"ImageObject","inLanguage":"it-IT","@id":"https:\/\/yousum.gpucore.co\/#\/schema\/person\/image\/","url":"https:\/\/youzum.net\/wp-content\/uploads\/avatars\/2\/1746849356-bpfull.png","contentUrl":"https:\/\/youzum.net\/wp-content\/uploads\/avatars\/2\/1746849356-bpfull.png","caption":"admin NU"},"url":"https:\/\/youzum.net\/it\/members\/adminnu\/"}]}},"rttpg_featured_image_url":null,"rttpg_author":{"display_name":"admin NU","author_link":"https:\/\/youzum.net\/it\/members\/adminnu\/"},"rttpg_comment":0,"rttpg_category":"<a href=\"https:\/\/youzum.net\/it\/category\/ai-club\/\" rel=\"category tag\">AI<\/a> <a href=\"https:\/\/youzum.net\/it\/category\/committee\/\" rel=\"category tag\">Committee<\/a> <a href=\"https:\/\/youzum.net\/it\/category\/news\/\" rel=\"category tag\">News<\/a> <a href=\"https:\/\/youzum.net\/it\/category\/uncategorized\/\" rel=\"category tag\">Uncategorized<\/a>","rttpg_excerpt":"Agentic LLMs often fail the same way, again and again. A Stanford research team traced this to missing, reusable capabilities. Their system, TRACE, diagnoses those gaps and trains for them directly. TRACE stands for Turning Recurrent Agent failures into Capability-targeted training Environments. It was released open-source under an MIT license. What problem does TRACE solve?&hellip;","_links":{"self":[{"href":"https:\/\/youzum.net\/it\/wp-json\/wp\/v2\/posts\/103947","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/youzum.net\/it\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/youzum.net\/it\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/youzum.net\/it\/wp-json\/wp\/v2\/users\/2"}],"replies":[{"embeddable":true,"href":"https:\/\/youzum.net\/it\/wp-json\/wp\/v2\/comments?post=103947"}],"version-history":[{"count":0,"href":"https:\/\/youzum.net\/it\/wp-json\/wp\/v2\/posts\/103947\/revisions"}],"wp:attachment":[{"href":"https:\/\/youzum.net\/it\/wp-json\/wp\/v2\/media?parent=103947"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/youzum.net\/it\/wp-json\/wp\/v2\/categories?post=103947"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/youzum.net\/it\/wp-json\/wp\/v2\/tags?post=103947"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}