Skip to content
Back to the timeline
AI News·

X Adjusts Recommendation Algorithm: Prioritizes Dwell Time and Completion Rate

X has adjusted its two-stage ranking mechanism for the "For You" feed, with the new model placing greater emphasis on user dwell time and post completion rates. The retrieval phase will also filter content based on completion rates, allowing posts from non-followed accounts to enter the recommendation pool after vector matching. X's total daily active user time increased from approximately 410 billion seconds in mid-August to around 500 billion seconds by mid-October.

X Adjusts Recommendation Algorithm: Prioritizing Dwell Time and Completion Rate

X has adjusted the two-stage ranking mechanism of its "For You" feed, with the new model placing greater emphasis on user dwell time and post completion rates. Musk stated that the new ranking model's "ability to predict effective time spent has improved significantly."

The second change occurred at an earlier retrieval stage. The system converts users and posts into vectors, pulls the most relevant content from accounts users don't follow, and filters it based on completion rates—posts that users typically abandon halfway through will have a harder time making it into the recommendation pool. X's total daily active user time increased from approximately 410 billion seconds in mid-August to about 500 billion seconds by mid-October.

Rohan Paul(@rohanpaul_ai)

Image