{"passport":{"slug":"ctfg","display_name":"CTFG","model_type":"embedding","parameter_count":null,"context_window":null,"claimed_developer":"Collaborative Temporal Feature Generation via Critic-Free Reinforcement Learning for Cross-User Sensor-Based Activity Recognition","confirmed_developer":null,"developer_confirmed_at":null,"developer_source":null,"first_appeared_at":"2023-03-16T00:00:00+00:00","first_appeared_on":"rss_pipeline","status":"confirmed","availability":null,"deployment_status":"not_deployed","availability_scope":null,"gate":1,"eu_ai_act":"GPAI","regulatory_status":{"eu_ai_act":"GPAI"},"open_source":null,"weights_available":null,"license":null,"provenance_chain":[],"parent_model":null,"model_family":null,"superseded_by":null,"supersedes":[],"recommended_replacement":null,"product_links":null},"events":[{"event_type":"first_appearance","event_date":"2023-03-16T00:00:00+00:00","title":"First recorded in AI news","detail":"A novel framework called CTFG (Collaborative Temporal Feature Generation) is proposed for cross-user sensor-based activity recognition. CTFG employs a Transformer-based autoregressive generator that incrementally constructs feature token sequences, optimized via a critic-free reinforcement learning algorithm called Group-Relative Policy Optimization. The framework aims to improve cross-user accuracy, reduce inter-task training variance, accelerate convergence, and achieve robust generalization.","source_url":"https://arxiv.org/abs/2603.16043","source_name":"ArXiv AI"}],"delta":{"days_unattributed":1212,"claimed_developer":"Collaborative Temporal Feature Generation via Critic-Free Reinforcement Learning for Cross-User Sensor-Based Activity Recognition","confirmed_developer":null,"match":false}}