TSPORec: Optimizing Token Selection for LLM-Based Sequential Recommendation

_reachsumit · x · 2026-08-11

The paper TSPORec proposes a learned policy for token selection via preference optimization. This method selects informative tokens from item text, replacing the traditional first-k truncation approach used in LLM-based recommender systems to improve performance.

Original post →

More from Research

Research channel →