AI & ML
Safe Personal RSI: Evaluating and Repairing Personal-Agent Updates
arXiv cs.SE0 views
arXiv:2611.23651v1 Announce Type: new Abstract: Kicker is a personal health assistant. Changes to an assistant's reusable instructions can improve task handling while changing what it discloses, which requests it follows, and how it reports actions. We study Safe Personal RSI, an update-evaluation component developed for Kicker. We restrict recursive self-improvement (RSI) to prompt updates, with model weights and runtime code fixed. The procedure tests proposed and current instructions, repeats failing cases, temporarily undoes changes to guide repairs, and checks safety and task performance before adoption. If no version passes, the current instructions stay in use. Our executable benchmark contains 29 email, calendar, health, and marketplace cases, with separate probes for authorization reuse, memory, and tool-output injection. Across three models and two deliberately risky edits, the procedure rejects 20 of 21 update attempts. In repair confirmation, an admitted GPT-5.6 Luna repair reduces attack violations from 10.7% to 0% and increases safe task completion from 89.7% to 100%; a broad handwritten authority rule matches these outcomes. Under stronger attacks, a repair transferred from Qwen2.5-7B-Instruct to Qwen3-14B reduces violations from 69.0% to 17.9% and increases safe task completion from 19.5% to 64.4%. Separate runtime probes examine memory-entry deletion and identify missing local OAuth scope checks in the evaluated snapshot.
Source:https://arxiv.org/abs/2611.23651PreviousOption-GRPO: Hierarchical RL Against AI Slop in Code Generation for Unseen SDKsNextCratesBump: A Version-Bump Dataset for the Rust Ecosystem
Related posts
The Songless Archive Is a Better Listening Log Than a Playlist
0dev.to
AI makes it easier to understand the whole project
0dev.to
ChatGPT Has No Secret Codes. These Features Are What Help
0dev.to
How to Check a Freelancer's AI Work Before You Pay Them
0dev.to
I scanned 277 public MCP servers. Here's what 90% of them miss.
0dev.to

The JSON Parsed. Is It Right? How to Score LLM Data Extraction
0dev.to