Xing Fu

Current memory retrieval methods for emotional support agents fall far short of human-level empathy, struggling to proactively retrieve memories that address users' latent emotional needs.

TabEmbed leapfrogs existing text embedding models to achieve SOTA performance on tabular data by reformulating tasks as semantic matching problems and using contrastive learning.

Ditch the manual feature engineering: KMLP's hybrid KAN-gMLP architecture automatically learns complex feature transformations and interactions, outperforming GBDTs on web-scale tabular data.

LLMs can be adapted for scenario-specific user representation by conditioning on queries, achieving state-of-the-art performance on Alipay benchmarks and demonstrating practical effectiveness in online A/B testing.