Search

From Open-ended Exploration to Decision Support: Proposing "QD-Rec" to Enhance Recommendation Quality and Diversity

Translated title. Original: オープンエンドな探索を意思決定支援へ：推薦の質と多様性を高める「QD-Rec」を提案

QD-Rec is a decision-support driven recommendation method that treats each item as a candidate paired with a hypothesis. It uses past good hypotheses to generate new, diverse proposals, aiming to build large, high-quality archives. Experiments across three domains show QD-Rec often outperforms zero-scratch baselines by promoting diverse, actionable Hypotheses and evolving recommendations over iterative cycles.

Sakana AI proposes "QD-Rec" to enhance recommendation quality and diversity through open-ended exploration | Forest of AI

Translated title. Original: オープンエンド探索で推薦の質と多様性を高める「QD-Rec」をSakana AIが提案 | AIの杜

Sakana AI presents QD-Rec, a recommendation method that combines open-ended hypothesis generation with quality-diversity search. It alternates between selecting promising candidates, generating new hypotheses, and evaluating them, showing improved average quality and larger result sets across M&A, marketing, and academic paper recommendations.

Recommender (experimental) | TypeType Docs

The document explains an experimental recommender feature for TypeType, detailing its optional nature, resource needs, stack changes, and three installation paths with steps to enable, install, or update the recommender while preserving data.

[2610.30128] Think Harder Where It Matters: Item-Level Adaptive Recursion for Sequential Recommendation

LoopRec introduces per-item adaptive recursion for sequential recommendation, replacing fixed-depth computation with learned, item-specific exit decisions. It achieves better performance across multiple backbone architectures and shows interpretable patterns where rare and recent items receive deeper processing; code is available online.

The RuStore neural network learned to guess what you want to play — game installations have doubled.

Translated title. Original: Нейросеть RuStore научилась угадывать, во что вы хотите поиграть, — установки игр выросли вдвое

RuStore has upgraded its AI-driven recommendation and search systems to boost discovery and combat manipulation, reporting a doubling of game installations after the update and highlighting AI-generated review summaries, with a vast catalog and large regional audience.

AI-Powered Shopping Research: Consumers Want Recommendations for Items That Suit Them - AI Watch

Translated title. Original: AI活用の買い物調査、期待するのは自分に似合うものの提案 - AI Watch

A Rakuten Insight survey found that people want AI shopping tools to recommend products that suit them, not just save time. Respondents generally kept their own judgment above AI advice and also reported concerns about poor purchases and less enjoyable, more uniform choices.

How does AI guess your entertainment desires? - intelligence-artificielle.com

Translated title. Original: Comment l'IA devine vos envies de divertissement ? - intelligence-artificielle.com

Recommender systems power entertainment and online gaming, using AI to learn user behavior and tailor suggestions. The article discusses algorithmic personalization, transparency concerns, and how blockchain could enhance trust and fairness in digital entertainment.

Netflix launches new AI recommendation system for users - Estadão

Translated title. Original: Netflix lança novo sistema de recomendação por IA para usuários - Estadão

Netflix has rolled out a limited AI-driven recommendation feature in the US app, presenting themed picks in a dedicated Discover tab. The system learns in real time, aims to improve engagement, is not chatbot-like, will exclude under-age profiles, and is described as an evolving, test-run with gradual rollout.

Paper page - RSI-Anything: Unlocking Recursive Self-Improvement with World Models

An automated message suggests papers related to a target work, invites user feedback on helpful comments, and points to Hugging Face spaces for recommendations via Semantic Scholar APIs.

Facebook offers you to customize its algorithm with AI - KultureGeek

Translated title. Original: Facebook vous propose de personnaliser son algorithme avec l'IA - KultureGeek

Meta unveils 'Your Algorithm' on Facebook, allowing users to steer recommendations by describing preferred topics in natural language, supplemented by 'Interested' and 'Not interested' signals. The system fuses explicit user input with existing behavioral signals and may use large language models to interpret intent in real time, though deployment timelines remain unspecified.