Zhihu contributor 周国睿 has spent the past year exploring a big Q:
👉 Can we take recommender systems to the next level?
Based on hands-on work with OneRec, they found E2E recsys might be the future. Here are his key thoughts:
🔍How big should a rec model be?
⚙️How to scale compute under high-concurrency, low-latency settings?
📈Can better recs come with more FLOPs?
His team designed OneRec, achieving 20–30% MFU — rare for recsys infra.
They solved codebook gen & compression 🧩 and now eye:
🎯Reward systems – crucial for rec
📏Clearer & solid scaling laws for model size vs. effect
📐Aligning rec behavior modality with other modalities on LLMs
🧠 Full Article on Zhihu: zhuanlan.zhihu.com/p/1918350…
📄 Tech Report: arxiv.org/abs/2506.13695v1#RecommenderSystem#LLM#AI#RL#EndToEnd#ScalingLaws
A #Recommendersystem that does 1000000 recommendations per second.
That's proper #scale. You should come and see how we did it with @hopsworks. The team told me that _generating_ the load was harder than _handling_ it :) ...
Tomorrow, i will give a webinar on scaling recommender systems to Tiktok scale (1m ops/sec). Part of it will be how to design streaming features in Flink - what type of window for what type of feature.
The webinar is free
hopsworks.ai/events/tiktok-p…
🌸2022 High Cited Articles🌸
"NLP-Based Bi-Directional Recommendation System: Towards Recommending Jobs to Job Seekers and Resumes to Recruiters" by Suleiman Ali Alsaif et al.
Views: 3996
Citations: 11
#information#recommendersystem#NLPmdpi.com/2504-2289/6/4/147
Recommender Update:
Started working on the frontend with Streamlit… still sketchy.
Plus, as I’m discovering, Streamlit is actually more powerful than meets the eye.
#RecommenderSystem#NLP#AI#DataScience
Use HeatWave's built-in #ML capabilities to deliver personalized recommendations to your customers. Try our NEW hands-on lab; build an app to predict which movies a user will like, who to target for promotion, and more: social.ora.cl/6015R89Wv#OCI#RecommenderSystem
I #recommendersystem sono i tanto discussi #algoritmi che analizzano il comportamento degli utenti e le loro preferenze per suggerire in modo personalizzato prodotti e servizi. Quelli esistenti però trascurano gli aspetti di sostenibilità ed etica degli oggetti proposti.
Come scegli i prodotti che compri?
Una nuova generazione di #recommendersystem potrebbe aiutarci a fare acquisti più responsabili!
/ segui il thread per saperne di più /