Verified teacher conversations, the hard subset, and the traces of the automated heal cycle
AI & ML interests
Agent hallucination, self-healing agents, tool use, RL fine-tuning (GRPO/BCO), AI safety
Recent Activity
View all activity
models 11
reagent-lexsi/heal-trajectory-cycle-c1
Updated • 17
reagent-lexsi/grpo-decision-h3-aya32b-v2
Updated • 40
reagent-lexsi/heal-first-cycle-c1-rejected
Updated • 31
reagent-lexsi/grpo-heal-aya32b-v2
Updated • 28
reagent-lexsi/grpo-decision-aya32b-v2
Updated • 27
reagent-lexsi/grpo-aya32b-v2
Updated • 28
reagent-lexsi/sft-aya32b-r16-v4
Updated • 28
reagent-lexsi/sft-aya32b-r16-v3
Updated • 27
reagent-lexsi/sft-aya32b-r16-long
Updated • 25
reagent-lexsi/sft-aya32b-r16
Updated • 32