Topic
Qwen2 5
2 episodes
-
MemTrain: Self Supervised Context Memory Training
Self-supervised framework MemTrain improves LLM context memory by training on unlabeled Wikipedia with coupled proxy tasks—masked reconstruction and memory recall—using GRPO. Achieves up to 17.67-point gains on long-horizon reasoning without task-specific labels.
-
AI memory framework MeMo skips LLM retraining
MIT's MeMo framework encodes new knowledge into a small dedicated memory model so teams can swap in a better LLM without retraining — and the performance gains are real. Justy and Cody break down how it actually works, what the benchmarks mean, and where the trade-offs bite.