arXiv Artificial Intelligence

Learning to Retrieve via Reinforcement Learning in Embedding Space

1 hour ago 07.10.2026 07:00 👁 1 views
Learning to Retrieve via Reinforcement Learning in Embedding Space

Quick summary

arXiv:2610.07731v1 Announce Type: cross Abstract: Dense retrieval models are typically trained with contrastive objectives that learn effective representations but do not directly optimize retrieval metrics or downstream task performance. To address this problem, we introduce RELER (REinforcement LEarning for Retrieval), a reinforcement learning framework that enables existing embedding models to learn to retrieve directly in embedding space and align to task-specific rewards. We train RELER by sampling unit-length query and document embedding actions from von Mises-Fisher (vMF) distributions

Key takeaways

  • arXiv:2610.07731v1 Announce Type: cross Abstract: Dense retrieval models are typically trained with contrastive objectives that learn effective representations but do not directly optimize retrieval metrics or downstream task performance.
  • To address this problem, we introduce RELER (REinforcement LEarning for Retrieval), a reinforcement learning framework that enables existing embedding models to learn to retrieve directly in embedding space and align to task-specific rewards.
  • We train RELER by sampling unit-length query and document embedding actions from von Mises-Fisher (vMF) distributions

Why it matters

The importance of “Learning to Retrieve via Reinforcement Learning in Embedding Space” will be measured by what changes in practice. User behavior, access conditions, verifiable performance and responsible-use outcomes are the signals worth following.

Kaynak sitede devamını oku: arXiv Artificial Intelligence ↗