HUM uses a compression prompt, a dedicated user token, input masking, and domain-reweighted training to improve LLM-based multi-domain recommendation.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.IR 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Heterogeneous User Modeling for LLM-based Recommendation
HUM uses a compression prompt, a dedicated user token, input masking, and domain-reweighted training to improve LLM-based multi-domain recommendation.