Search papers, labs, and topics across Lattice.
University of Science and Technology of China & State Key Laboratory of Cognitive Intelligence
4
0
6
DNNs in recommendation models don't just learn feature interactions, they fundamentally reshape embedding spaces by preventing dimensional collapse.
Forget complex model architectures for cross-domain recommendation: Taesar shows that cleverly transforming your data can unlock better performance with standard models.
CoT reasoning can hurt recommender performance by drowning out important ID signals – unless you compress reasoning chains and use bias-subtracted contrastive decoding to realign the inference subspace.
Recommender systems can bootstrap their performance without external data via a recursive self-improvement loop that generates, filters, and learns from its own plausible user interaction sequences.
