Anchored SFT: Combining DFT With KL Trust Regions
Anchored SFT (ASFT) pairs DFT confidence-weighted token loss with a forward KL divergence trust region on frozen LoRA base weights to prevent policy drift.

Anchored SFT (ASFT) pairs DFT confidence-weighted token loss with a forward KL divergence trust region on frozen LoRA base weights to prevent policy drift.

Afterlife helps a continuous learning AI agent develop a unique personality and skills at weight level with LoRA adapters, not just a memory.md file.

When an agent's actions alter its environment for good, cumulative replay causes catastrophic remembering; a sliding window sustains continual learning.

Continual learning LoRA for researchers: train a small adapter nightly on the day's successful runs with modified SFT, so the agent improves at its work.

AI agent memory carries facts, not skill: why an agent that solved a task yesterday is no better today, and how memory, RL and LoRA fine-tuning compare.