Tag
Netflix replaced its long-standing recommendation algorithm with an LLM-backed system called GenRec, which outperforms the previous method using significantly less training data, indicating a major shift in AI-driven product engineering.
DMRL is a Document-Mediated Reinforcement Learning framework that optimizes skill documents for advertising recommendation using structured editing actions, policy optimization, and long-term reward prediction, demonstrating improved performance over baselines in large-scale deployment.
A tweet highlights Audible's book suggestion feature, which recommends books based on a user's listening history in warfare and conquest.
The author details building a personal AI agent for book recommendations, sharing lessons on tool compliance, cost management, and feedback loops using a Go-based implementation.
The article discusses how AI's growing capabilities may shift interaction paradigms, with AI prompting users based on their desires to help achieve goals, similar to recommender systems.
The post describes how GStack, used with Claude Code, allows users to handle one-way-door questions by accepting all recommendations for convenience.
X (Twitter) open sourced its 'For You' timeline ranking algorithm under the Apache v2 license, revealing how it weights predicted user behaviors like likes, replies, and dwell time, and applies separate safety and diversity filters.
A new study finds that while TikTok's 'not interested' button effectively reduces unwanted content by 84%, the algorithm can slowly reintroduce such content if users do not maintain consistent feedback. Researchers cloned accounts to study personalization, highlighting limits on user agency.
Introducing the AI Agent operation backend generated by Mimo, designed for self-media to automatically mine trending topics and recommend topics.
This paper proposes a Federated Hash Projected Latent Factor (FHPLF) model that integrates hash learning into federated learning to reduce communication costs and enhance privacy, using binary gradient-like matrices and projected Hamming distance to improve accuracy and efficiency.
A detailed breakdown of Netflix's hybrid weighted recommendation system design, covering scale estimation, cold start strategies for new users, behavioral signal capture, and the balance between recall and precision.
This article proposes a community governance and algorithm recommendation philosophy called 'Archipelago Ecological Algorithmic Isolation', which aims to solve the problem of community tone dilution caused by user growth through semantic vector immune blocking, local sandbox isolation of controversial topics, and private island non-retrieval mechanisms, achieving both scale and tone.
SinkRec introduces a hybrid memory-transition architecture to mitigate semantic state sink in long sequence recommendation, using memory-conditioned gated delta networks to decouple pattern storage from dynamic modeling, achieving linear-time efficiency.
Musk has open-sourced X's (formerly Twitter) For You recommendation algorithm on GitHub. The algorithm uses a Grok-based transformer to predict user interaction preferences.
Cognition created comprehensive documentation of the latest X algorithm via DeepWiki, noting that engagement parameter weights remain private but maximizing dwell time helps boost posts.
Elon Musk announced that the latest 𝕏 algorithm, including the For You feed recommendation system powered by a Grok-based transformer, has been published on GitHub. The release includes an end-to-end inference pipeline, pre-trained model artifacts, and new components for content understanding and ads.
Elon Musk's xAI open-sourced the X 'For You' recommendation algorithm on GitHub, revealing the code that determines virality on the platform. The article analyzes the code and outlines six rules for going viral.
Elon Musk states that X will open-source its recommendation algorithms to demonstrate transparency and build user trust.