BLUE from Google internship lands NeurIPS: LLM-written user profiles boost recommendations
shangbinfeng · x · 2026-09-29
BLUE, a Google internship work accepted to NeurIPS 2026, uses reinforcement learning to align LLM-generated textual user profiles with embedding-based recommendation objectives: an embedding model provides reward signals pulling profiles toward positive items, plus next-item-prediction supervision in text space. On Amazon Reviews 2023 and Google Local Reviews in zero-shot sequential recommendation, BLUE consistently beats strong baselines under both frozen and trainable embeddings, with clear gains in cross-domain transfer.
More from Research
- 1MB replay script beats frontier models: NeurIPS oral paper exposes flaws in CUA benchmarks — proceduralia · 2026-09-29
- Replay agents hit SOTA on CUA benchmarks: NeurIPS oral paper exposes eval flaws — proceduralia · 2026-09-29
- GPT-6 Sol scores 89.6% on ARC-AGI-2 but only 23% on ARC-AGI-3, ARC Prize reports — fchollet · 2026-09-29
- Terence Tao-backed Caltech Mathathon redesigns around AI-assisted math understanding — AnimaAnandkumar · 2026-09-29
- 30+ labs fail to replicate Marcus et al's 1999 Science paper on infants vs RNNs, built on just 16 babies — tallinzen · 2026-09-29
- Chris Manning on why LMs learn verb categories first, and why he thinks LeCun is wrong about language — ziv_ravid · 2026-09-29