arXiv Artificial Intelligence

TRACE: Trajectory Selection for Parallel Scaling of Search Agents

1 hour ago 01.10.2026 07:00 👁 1 views
TRACE: Trajectory Selection for Parallel Scaling of Search Agents

Quick summary

arXiv:2609.39912v1 Announce Type: cross Abstract: Parallel search may generate a correct answer that final-answer voting fails to select. We formulate this consolidation stage as trajectory selection and introduce TRACE (Trajectory Ranking with Aggregated Cross-Rollout Evidence), a lightweight learned selector that ranks completed trajectories using the search evidence behind their answers. TRACE preserves individual query and evidence occurrences, connects rollouts through shared content or document identity, and propagates information across these relations. Each candidate answer then reads

Key takeaways

  • arXiv:2609.39912v1 Announce Type: cross Abstract: Parallel search may generate a correct answer that final-answer voting fails to select.
  • We formulate this consolidation stage as trajectory selection and introduce TRACE (Trajectory Ranking with Aggregated Cross-Rollout Evidence), a lightweight learned selector that ranks completed trajectories using the search evidence behind their answers.
  • TRACE preserves individual query and evidence occurrences, connects rollouts through shared content or document identity, and propagates information across these relations.

Why it matters

The importance of “TRACE: Trajectory Selection for Parallel Scaling of Search Agents” 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 ↗