SFA: model averaging beats replay in continual learning, no replay or auxiliary loss needed
jefrankle · x · 2026-11-20
Promo thread for the same paper "Forget About It" (arXiv:2501.05559): SFA outperforms replay-style methods in sequential fine-tuning with up to 36% less training time, using no replay and no auxiliary loss.
More from Research
- AI Proves Optimal Solutions to MIPLIB Open Problems Unsolved for 9 Years — IgorCarron · 2026-11-20
- Yao Shunyu's private talk notes leak: small-model hierarchies, RSI, small-experiment traps — yangyi · 2026-11-20
- Murmur Transcribes Long Audio 2.5x Faster Than Single-Pass, Beats WhisperX — bariskasikci · 2026-11-20
- Salesforce releases BLIP-Live: SOTA proactive streaming video understanding with PAUC-F metric — Salesforce · 2026-11-20
- RICoT paper: LLMs read page streams like humans for document segmentation, boundary F1 0.915 — Prashant Khatri · 2026-11-20
- One neuron per language hits 99.1% language-ID accuracy, beating dedicated models — julianweisser · 2026-11-20