Search

Tag: #recommendation6 results

AgentSelect Benchmark for Agent Recommendation

AgentSelect Benchmark for Agent Recommendation

AgentSelect introduces a benchmark for recommending LLM agent configurations based on narrative queries, addressing the lack of query-conditioned supervision. It aggregates 111,179 queries, 107,721 agents, and 251,103 interactions from 40+ sources into unified data. Analyses highlight the shift to long-tail supervision and the need for content-aware capability matching.

ArXiv AIResearchMar 5#llm-agents#benchmark#recommendation
LinkedIn Unifies 5 Feeds with Single LLM

LinkedIn Unifies 5 Feeds with Single LLM

LinkedIn overhauled its feed for 1.3B users by replacing five separate retrieval pipelines with one LLM-based system, improving professional context understanding and cutting costs. The redesign spans retrieval, ranking via generative recommenders, and compute management. Engineers ran hundreds of tests to match user interests and behaviors more precisely.

VentureBeatMediaMar 16#recommendation#scaling#production
Ember Artline Ships April 22

Ember Artline Ships April 22

Amazon's Ember Artline TV, a budget alternative to Samsung's Frame, opens pre-orders today and ships April 22 in US/Canada. It features a matte 4K QLED panel with 2,000+ free artworks and an AI 'Match the room' tool that analyzes room photos for personalized suggestions. Additional perks include 10 frame colors, Amazon Photos integration, and Alexa+ support.

V-STAR: Value-Guided RecSys Sampling

V-STAR: Value-Guided RecSys Sampling

V-STAR addresses probability-reward mismatch in generative recsys via value-guided decoding and sibling-relative RL. VED efficiently explores high-potential prefixes; Sibling-GRPO focuses on decisive branches. Outperforms baselines in accuracy and diversity.

ArXiv AIResearchFeb 12#research#v-star#v1
Robust Policy Optimization for Recommendations

Robust Policy Optimization for Recommendations

DRPO tackles model collapse in off-policy generative recommendation via optimistic distributionally robust optimization. Proves hard filtering recovers high-quality data from noisy logs. Achieves SOTA on mixed-quality benchmarks.

ArXiv AIResearchFeb 12#research#drpo#v1