[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-82504-en":3,"doc-seo-82504-105":29,"detail-sidebar-cat-0-en-105":90},{"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":4,"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},82504,687197100911,"Himbo","https://ap-avatar.wpscdn.com/avatar/a000239b6f1da00475?x-image-process=image/resize,m_fixed,w_180,h_180&k=1782698725881665579",8,"Research & Report","Learning Generalizable Skill Policy with Data-Efficient Unsupervised RL","Unsupervised Reinforcement Learning (URL) pre-trains scalable, skill-conditioned policies without extrinsic rewards to serve as a foundation for downstream control. Current off-policy URL is limited by two overlooked bottlenecks: non-stationary skill semantics and brittle generalization across distribution shifts. GenDa (Generalizable Data-efficient Agent) introduces skill relabeling to reduce semantic drift and improve data efficiency, and a Complementary Information Bottleneck (CIB) to emphasize ego-centric features, strengthening transfer robustness and scalability through experiments on multiple benchmarks.","Learning Generalizable Skill Policy with Data-Efficient Unsupervised RL  \nJongchan Park 1 Seungjun Oh 1 Seungho Baek 1 Yusung Kim 2 1  \narXiv :2607 .00392v 1 [ cs .LG] 1 Jul 2026  \nAbstract  \nUnsupervised Reinforcement Learning (URL) aims to pre-train scalable, skill-conditioned policies without extrinsic rewards, serving as a foundation for downstream control tasks. Despite recent progress, we argue that current off-policy URL methods are limited by two critical, overlooked bottlenecks: (1) non-stationary skill semantics and (2) brittle generalization. To address these challenges, we propose GenDa (Generalizable Data-efficient Agent), a unified framework for robust unsupervised reinforcement learning. First, we introduce a skill relabeling mechanism to mitigate non-stationarity and significantly improve data efficiency for pre-training.  \nSecond, we propose a Complementary Information Bottleneck (CIB), encouraging the learned skill policy to focus on ego-centric features and become robust to distribution shifts for downstream tasks. Through various experiments, we demonstrate that GenDa significantly enhances the scalability of URL with superior generalizability and data efficiency. Our code and videos are available at [https://ihatebroccoli](https://ihatebroccoli) .  \n[github.io/official-GenDa/](github.io/official-GenDa/) .  \n1. Introduction  \nRecent advances in unsupervised reinforcement learning (URL) have enabled the discovery of semantically distinct behaviors (“skills”) from state transitions, without access to external reward signals (Gregor et al., 2016 ; Kim et al., 2021 ; Kamienny et al., 2022 ; Park et al., 2023 ; Yang et al., 2023) . This paradigm aims to create a general-purpose foundation for control rather than just learning individual skills. It provides a pre-trained policy that readily adapts to diverse downstream tasks with minimal fine-tuning (Laskin et al.,  \n1Department of Artificial Intelligence, Sungkyunkwan University, Suwon, Republic of Korea 2Department of Computer Science and Engineering, Sungkyunkwan University, Suwon, Republic of Korea. Correspondence to: Yusung Kim \u003C[yskim525@skku.edu](yskim525@skku.edu) >.  \nProceedings of the 43 rd International Conference on Machine Learning, Seoul, South Korea. PMLR 306, 2026 . Copyright 2026 by the author(s) .  \n2021 ; Rajeswar et al., 2023) . Achieving scalability in URL requires both efficiency during pre-training and robustness during transfer, ensuring the learned policy generalizes to varied downstream tasks.  \nDespite recent progress in URL, we argue that current stateof-the-art methods face two critical bottlenecks that hinder this scalability: (1) sample inefficiency arising from the overlooked non-stationary skill semantics, and (2) brittle generalization caused by overfitting to global context.  \nFirst, real-world interactions are costly, making data efficiency important. To maximize data efficiency, modern URL methods rely on off-policy algorithms that reuse past experiences stored in a replay buffer. However, this introducesa fatal flaw: semantic drift. In off-policy settings, each trajectory is collected under a randomly sampled skill z and z-conditioned skill policy (Eysenbach et al., 2019 ; Sharma et al., 2020) . The resulting z–trajectory pair is stored ina replay buffer and reused throughout training (Haarnoja et al., 2018 ; Laskin et al., 2022 ; Park et al., 2024 ; Zheng et al., 2025) . As learning progresses, the behavioral trajectories induced by the same skill z can change, because skill policy evolves across off-policy learning. The time-varying semantics of z induce destabilization of the skill policy by providing stale samples.  \nSecond, for a skill policy to be a scalable foundation for various downstream tasks, it must be robust to distribution shifts. A skill such as “walking forward” should be executable regardless of the global contextual information, such as environmental factors. However, in many prior works (Eysenbach et al.,","cbCaivpNuoaX7PE1","https://ap.wps.com/l/cbCaivpNuoaX7PE1","pdf",2199597,1,16,"English","en",105,"# Abstract\n# Introduction\n# Preliminaries","[{\"question\":\"What are the two key bottlenecks in current off-policy unsupervised reinforcement learning identified by the paper?\",\"answer\":\"The paper points to (1) non-stationary skill semantics that cause semantic drift in replayed data, and (2) brittle generalization due to overfitting to global contextual information.\"},{\"question\":\"How does GenDa improve data efficiency in unsupervised skill learning?\",\"answer\":\"GenDa uses a skill relabeling mechanism to mitigate semantic drift, making pre-training more data-efficient in the off-policy URL setting.\"},{\"question\":\"What is the role of the Complementary Information Bottleneck (CIB)?\",\"answer\":\"CIB discourages the policy from relying on global contextual signals, promoting robustness to distribution shifts and improving skill execution consistency for downstream tasks.\"}]",1784180990,40,{"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":85,"head_meta":87,"extra_data":89,"updated_unix":27},"learning-generalizable-skill-policy-with-data-efficient-unsupervised-rl","",{"@graph":35,"@context":84},[36,53,67],{"@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/learning-generalizable-skill-policy-with-data-efficient-unsupervised-rl/82504/",4,{"url":51,"name":13,"@type":54,"author":55,"headline":13,"publisher":57,"fileFormat":60,"inLanguage":23,"description":14,"dateModified":61,"datePublished":61,"encodingFormat":60,"isAccessibleForFree":62,"interactionStatistic":63},"DigitalDocument",{"name":9,"@type":56},"Person",{"url":40,"name":58,"@type":59},"DocShare","Organization","application/pdf","2026-07-16",true,{"@type":64,"interactionType":65,"userInteractionCount":4},"InteractionCounter",{"@type":66},"ViewAction",{"@type":68,"mainEntity":69},"FAQPage",[70,76,80],{"name":71,"@type":72,"acceptedAnswer":73},"What are the two key bottlenecks in current off-policy unsupervised reinforcement learning identified by the paper?","Question",{"text":74,"@type":75},"The paper points to (1) non-stationary skill semantics that cause semantic drift in replayed data, and (2) brittle generalization due to overfitting to global contextual information.","Answer",{"name":77,"@type":72,"acceptedAnswer":78},"How does GenDa improve data efficiency in unsupervised skill learning?",{"text":79,"@type":75},"GenDa uses a skill relabeling mechanism to mitigate semantic drift, making pre-training more data-efficient in the off-policy URL setting.",{"name":81,"@type":72,"acceptedAnswer":82},"What is the role of the Complementary Information Bottleneck (CIB)?",{"text":83,"@type":75},"CIB discourages the policy from relying on global contextual signals, promoting robustness to distribution shifts and improving skill execution consistency for downstream tasks.","https://schema.org",{"og:url":51,"og:type":86,"og:title":13,"og:site_name":58,"og:description":14},"article",{"robots":88,"canonical":51},"index,follow",{"doc_id":7,"site_id":24},{"code":4,"msg":5,"data":91},[92,96,100,104,109,114,118,121,126,129,133],{"id":20,"doc_module":4,"doc_module_name":45,"category_name":93,"show_sort_weight":94,"slug":95},"Story & Novel",90,"story-novel",{"id":46,"doc_module":4,"doc_module_name":45,"category_name":97,"show_sort_weight":98,"slug":99},"Literature",80,"literature",{"id":52,"doc_module":4,"doc_module_name":45,"category_name":101,"show_sort_weight":102,"slug":103},"Exam",70,"exam",{"id":105,"doc_module":4,"doc_module_name":45,"category_name":106,"show_sort_weight":107,"slug":108},5,"Comic",60,"comic",{"id":110,"doc_module":4,"doc_module_name":45,"category_name":111,"show_sort_weight":112,"slug":113},6,"Technology",50,"technology",{"id":115,"doc_module":4,"doc_module_name":45,"category_name":116,"show_sort_weight":28,"slug":117},7,"Healthcare","healthcare",{"id":11,"doc_module":4,"doc_module_name":45,"category_name":12,"show_sort_weight":119,"slug":120},30,"research-report",{"id":122,"doc_module":4,"doc_module_name":45,"category_name":123,"show_sort_weight":124,"slug":125},9,"Religion & Spirituality",20,"religion-spirituality",{"id":124,"doc_module":4,"doc_module_name":45,"category_name":127,"show_sort_weight":124,"slug":128},"World Cup","world-cup",{"id":130,"doc_module":4,"doc_module_name":45,"category_name":131,"show_sort_weight":130,"slug":132},10,"Lifestyle","lifestyle",{"id":134,"doc_module":4,"doc_module_name":45,"category_name":135,"show_sort_weight":105,"slug":136},19,"General","general"]