New article explains why masking works in VAE-based recommender systems
burkov · x · 2026-10-18
Masking user interaction histories during training reliably improves VAE-based recommenders, but has been an unexplained practical trick. A new article closes the gap by explaining how user collaboration arises in these models, how information actually flows between users, and what side effects masking causes.
More from Research
- MIT lets a Claude-powered agent freely "play" on a simulated island for 30 hours — and finds play expands its capabilities — LuizaJarovsky · 2026-10-18
- New paper measures model capability via how models solve tasks, not success rates — maksym_andr · 2026-10-18
- Economy of Minds: local incentives may replace central managers in multi-agent AI — pliang279 · 2026-10-18
- Ropedia Super Week: world model W0, dexterous bimanual controller DONIC and more — liuziwei7 · 2026-10-18
- Google Research builds AI-agent-controlled wet lab with autonomous molecular design loop — Puzzleheaded-King584 · 2026-10-18
- Terence Tao asks: will AI bring a golden age or a dark age for mathematics and science? — smilelamp · 2026-10-18