PulseExploreJournal ClubDebatesTrendingResearchersJournals
Instagram
HomeExploreJournal ClubTrending
Synapse
⌘+K
Synapse
January 22, 20260 citationsOpen Access

Phoenix: Candidate-Isolated Transformer Ranking in X's Recommendation System

View Full Paper
DADaniel An

Key Points

  • The aim is to analyze the Phoenix recommendation system and its candidate isolation attention pattern.
  • Conducted a technical analysis of the Phoenix recommendation system architecture.
  • Compared the 2023 Heavy Ranker and the 2026 Phoenix systems.
  • Evaluated the implications of candidate isolation for caching and fairness.
  • Enumerated 19 engagement signals predicted by the Phoenix system.
  • Discussed transparency limitations in the recommendation system.
  • Identified key differences between the Heavy Ranker and Phoenix systems.
  • Demonstrated how candidate isolation affects caching and consistency.
  • Enumerated 19 unique engagement signals associated with Phoenix.
  • Outlined transparency issues present in the current recommendation system.

Abstract

Technical analysis of X's Phoenix recommendation system, open-sourced January 19, 2026. This paper examines the novel candidate isolation attention pattern where post candidates cannot attend to each other during transformer inference. Key contributions include comparison of 2023 Heavy Ranker vs 2026 Phoenix architectures, analysis of candidate isolation design and its implications for caching, consistency, and fairness, enumeration of 19 engagement signals predicted by Phoenix, comparison to other production systems (YouTube, TikTok/Monolith, Instagram), and discussion of transparency limitations.

Ask AI
Helpful
Bookmark
Share
View Full Paper

Cite This Study

Daniel An (2026) studied this question.

synapsesocial.com/papers/6971bdec642b1836717e2961https://doi.org/10.5281/zenodo.18318222
Ask AI
Helpful
Bookmark
Share
View Full Paper