--- title: Reward Hacking Resarch Update | daily.dev description: EleutherAI shares an interim research update on reward hacking in reinforcement learning systems. The team built a testbed called 'djinn' with ~750 coding... canonical: https://daily.dev/posts/reward-hacking-resarch-update-y6pqlkqnk twitter:card: summary_large_image twitter:site: @dailydotdev og:type: website og:site_name: daily.dev og:title: Reward Hacking Resarch Update | daily.dev og:description: EleutherAI shares an interim research update on reward hacking in reinforcement learning systems. The team built a testbed called 'djinn' with ~750 coding... og:url: https://daily.dev/posts/reward-hacking-resarch-update-y6pqlkqnk og:image: https://api.daily.dev/og/posts/y6pqlkqNK.png og:image:alt: Reward Hacking Resarch Update og:image:width: 1200 og:image:height: 630 og:locale: en --- > ## Documentation Index > Fetch the complete documentation index at: https://daily.dev/llms.txt > Use this file to discover all available pages before exploring further. # Reward Hacking Resarch Update **[EleutherAI](https://daily.dev/sources/eleutherai)** · 6 min read · 0 upvotes · 0 comments ## Summary EleutherAI shares an interim research update on reward hacking in reinforcement learning systems. The team built a testbed called 'djinn' with ~750 coding problems and 26 exploit types to study when and how models learn to game reward signals. RL experiments with Qwen 3 models (8B and 14B) failed to elicit reward hacking without explicit prompting. Switching to supervised fine-tuning revealed a key behavioral difference: GPT-OSS family models generalized hacking behavior to held-out exploit types even without being prompted to hack (~25% exploit rate), while Qwen 3 models dropped below 5% without explicit prompting. The team is now focusing on RL-tuning GPT-OSS 20B to reliably elicit hacking in a semi-realistic training environment, with the goal of testing monitoring and mitigation strategies including canary problems and interpretability-based probes. ## Full article daily.dev links to this article rather than hosting it. Read it at the original source: --- Tags: [#deep-learning](https://daily.dev/tags/deep-learning), [#reinforcement-learning](https://daily.dev/tags/reinforcement-learning), [#ai-safety](https://daily.dev/tags/ai-safety) [View this post on daily.dev](https://daily.dev/posts/reward-hacking-resarch-update-y6pqlkqnk) ```json {"@context":"https://schema.org","@graph":[{"@type":"Organization","@id":"https://daily.dev/#organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180},"sameAs":["https://twitter.com/dailydotdev","https://github.com/dailydotdev","https://www.linkedin.com/company/daily-dev-ltd"]},{"@type":"WebSite","@id":"https://daily.dev/#website","url":"https://daily.dev","name":"daily.dev","publisher":{"@id":"https://daily.dev/#organization"},"potentialAction":{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https://daily.dev/search?q={search_term_string}"},"query-input":"required name=search_term_string"}}]} {"@context":"https://schema.org","@type":"TechArticle","headline":"Reward Hacking Resarch Update","url":"https://daily.dev/posts/reward-hacking-resarch-update-y6pqlkqnk","mainEntityOfPage":{"@type":"WebPage","@id":"https://daily.dev/posts/reward-hacking-resarch-update-y6pqlkqnk"},"datePublished":"2026-08-02T07:04:31.251Z","dateModified":"2026-08-02T07:31:56.524Z","description":"EleutherAI shares an interim research update on reward hacking in reinforcement learning systems. The team built a testbed called 'djinn' with ~750 coding...","image":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/7fa3fbd6e99da7e05ab42115e7e68fde?_a=AQAEuop","thumbnailUrl":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/7fa3fbd6e99da7e05ab42115e7e68fde?_a=AQAEuop","isAccessibleForFree":true,"articleSection":"EleutherAI","inLanguage":"en","publisher":{"@type":"Organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180}},"author":{"@type":"Organization","name":"EleutherAI","logo":"https://media.daily.dev/image/upload/s--aAcRyWGF--/f_auto,q_auto/v1785654263/logos/eleutherai","url":"https://daily.dev/sources/eleutherai"},"commentCount":0,"discussionUrl":"https://daily.dev/posts/reward-hacking-resarch-update-y6pqlkqnk","interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":0},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":0}],"keywords":"deep-learning,reinforcement-learning,ai-safety","timeRequired":"PT6M"} {"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://daily.dev"},{"@type":"ListItem","position":2,"name":"EleutherAI","item":"https://daily.dev/sources/eleutherai"},{"@type":"ListItem","position":3,"name":"Reward Hacking Resarch Update"}]} ```