← Back to the feed
AI & softwareResearch findings

Ballad · Bandit-based Dynamic LLM Selection for Heuristic Discovery

Original published 2026.11.13 03:28 KSTAdded to Blackdeck 2026.11.15 06:45 KST

What has changed?

A dynamic routing strategy selects LLMs during search to maximize heuristic quality and efficiency

Key points

  1. Proposes a dynamic LLM selection method to optimize heuristic discovery
  2. Empirically demonstrates that larger models do not consistently outperform smaller ones
  3. Achieves superior heuristic quality and up to 65.6% API cost reduction compared to using the strongest model

Connect it to your work

Credit propagation via heuristic genealogy enables better model coordination

What remains unproven

This research note is based on an abstract; check the original for test conditions and application scope.

Originals used for this note

Links open the original publisher in a new tab.

  1. arxivBallad: Bandit-Based LLM Routing for Automated Heuristic Discovery