← Back to the feed
Energy & materialsResearch findings

NGA · Development of a Non-autoregressive Auction Framework Accounting for Global Externalities

Original published 2025.06.06 11:25 KSTAdded to Blackdeck 2026.10.23 16:19 KST

What has changed?

Explicitly models global externalities and interactions between ads and adjacent organic content

Key points

  1. Proposes NGA, a non-autoregressive auction framework that explicitly accounts for global externalities
  2. Jointly captures interactions between ads and adjacent organic content to improve efficiency
  3. Offline experiments and large-scale A/B testing show consistent superiority in effectiveness and efficiency

Connect it to your work

Uses non-autoregressive decoding and parallel multi-tower evaluator for real-time list-wise reward and payment computation

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. arxivNGA: Non-autoregressive Generative Auction with Global Externalities for Advertising Systems