Training with Missed Targets in Generative Recommendation: Separating Supervision from Probability Competition

Key points

• A training strategy appends these missed targets to reranker training lists, although inference still ranks only original candidates.1
• An append/no-append comparison therefore cannot explain changes in returned-item rankings.1
• We construct three matched losses that hold retrieved-target weight fixed while introducing appended-target supervision and group competition separately.1
• Experiments with a released OneRec model and locally trained Amazon generators show that this competition can harm returned-item ranking.1

Sources (1)

• [1]Training with Missed Targets in Generative Recommendation: Separating Supervision from Probability Competition
  arXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 7, 02:07 PM

  Generative recommenders return a limited candidate set and may omit observed targets before reranking.

  A training strategy appends these missed targets to reranker training lists, although inference still ranks only original candidates.

Extractive summary: sentences quoted from the sources.