PGHIS-CPEG replaces user-item embeddings with hierarchical LLM-generated textual profiles and contrastively prompted high-quality ground truths, then fine-tunes Qwen2.5-7B to generate recommendation explanations.
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
-
Hierarchical Interaction Summarization and Contrastive Prompting for Explainable Recommendations
PGHIS-CPEG replaces user-item embeddings with hierarchical LLM-generated textual profiles and contrastively prompted high-quality ground truths, then fine-tunes Qwen2.5-7B to generate recommendation explanations.