Open-weight & owned stack · Training tuning

Lacuna Presents Reinforcement Unlearning Using GRPO

Sep 11, 2026

Lacuna presented a GRPO-based method for removing targeted personal or copyrighted data from language models.

Read the original at lacuna.tiptreesystems.comOpens the publisher's site in a new tab

Also covering this

Reinforcement Unlearning via Group Relative Policy Optimization | ML Anthologymlanthology.org, Sep 12

More in Open-weight & owned stack

Z.ai publishes GLM-5.3 open weightsSep 8 MiniMax H3 highlighted as open-weight multimodal modelSep 7 InternLM releases 397B multimodal Intern-S2 modelSep 13 DeepSeek Model Nearly Matches GPT-6 Astra DesignSep 10 DeepSeek presents V4 FlashSep 8