The ACM RecSys 2026 Tutorials Track is now open for submissions!

We welcome proposals on recommender systems topics, especially tutorials addressing evaluation, deployment, feedback loops, and real-world constraints.

#recsys #recommendersystems #RS_c

📣 Call for Papers 📣

Working on user-centered #news #recsys or their legal & ethical aspects?

👉 Submit to the 14th INRA workshop, co-located w/
@umapconf.bsky.social in Gotenburg!

🗓️ Paper deadline: ** April 9th, 2026 **

More info: research.idi.ntnu.no/NewsTech/INR...

#INRA2026 #UMAP2026

Data Scientist III - Remote - Indeed.com Indeed

I am biased but this is one of the coolest, high leverage jobs out there.

AI may cause a lot of volatility un the labor market, join my team and help make it more efficient so that we can keep up with all the change. Feel free to DM. #recsys #econsky

www.indeed.com/m/viewjob?jk...

Not All RecSys Problems Are Created Equal Explore why recommender systems challenges differ dramatically between transaction-oriented and content-oriented RecSys. Discover real-world strategies, LLM innovations, and key metrics shaping industrial recommendations today.

Not All RecSys Problems Are Created Equal Introduction: The Hidden Complexity of Recommender Systems Imagine scrolling through Netflix recommendations that perfectly match.... @cosmicmeta.ai #RecSys

https://u2m.io/ZNu8xu7c

TikTok could be forced to change app’s ‘addictive design’ by European Commission Preliminary EU ruling says app shifts brains of users into ‘autopilot mode’, with concerns for children and vulnerable adults

A case to follow.

This cuts to the very core of TikTok and other platforms’ business model.

The charge is to change. Or pay (up to 6% of turnover)

#addictiveDesign
#infiniteScrolling
#recsys

www.theguardian.com/technology/2...

Maria Teleki Personal website for Maria Teleki.

More on my work here: mariateleki.github.io

#AI #Speech #RecSys

WooPush: technology of the attention economy, for informed citizenship • David Graus TikTok knows what you want to see. Why doesn’t the government know what you need to know? Yesterday I gave a keynote at the Dutch Ministry of the Interior’s Open Data Conference, where I shared my lat...

Gave a keynote at the #OpenData Conferentie on WooPush, my latest brainwave on #opengov x #recsys to proactively inform citizens about government decisions that affect them. The attention economy's technology, but for informed citizenship instead of clicks.

Read more: graus.nu/blog/woopush...

es, as a #recsys person, I know why… they aren't editors, CBS was a reputable news source until recently, practical ways to avoid this are hard, etc.

But the end user experience is still (attempted) slander of an innocent victim of unaccountable state violence when I open a new tab in my browser.

This summary got the research design (somewhat) wrong; but the general idea is captured. Overall, it's just a wordy abstract.

Main caveats:
- No author consent / notification
- No control / opt-out
- No room for corrections

What would the effect be: Fewer article clicks or more?
#ACM #RecSys

Мои книги по Search & Recsys Друзья, я наконец опубликовал третью книгу по теме поиска (плюс еще одна по близкой т...

#search #recsys #ecommerce #solr

Origin | Interest | Match

PISA: Combining Transformers and ACT-R for Repeat-Aware Sequential Listening Session Recommendation | Deezer Research Happy to share that one of our #RecSys24 paper got an extended version accepted for publication in the Transactions On Recommender Systems Journal #TORS with the title: "PISA: Combining Transformers a...

New Journal paper accepted! #RecSys www.linkedin.com/posts/deezer...

TransAct V2: Lifelong User Action Sequence Modeling on Pinterest Recommendation Modeling user action sequences has become a popular focus in industrial recommendation system research, particularly for Click-Through Rate (CTR) prediction tasks. However, industry-scale CTR models o...

Read Xia et al., “TransAct V2.” earlier today, some notes:
- long and short term user histories (long term [10k items], short term [100], impressions [100])
- sampled soft max loss function seems useful in environments with very weak negatives

#recsys #ml
arxiv.org/abs/2506.02267

Quick test: swapped chronological feed for topic-weighted feed for 24h — engagement rose but time-on-site dropped. Tradeoff? #Recsys #Research

TORS: Vol 4, No 1

📜ACM Transactions on Recommender Systems introduces a special issue on deep reinforcement learning for recommender systems, collecting cutting-edge research that addresses pressing challenges and showcases innovations. #recsys #deeplearning #reinforcementlearning
dl.acm.org/toc/tors/202...

What happens when you say:
“I want a horror -- comedy -- movie”? 🎥

That slip-of-the-tongue can confuse recommender systems.
Our INTERSPEECH 2025 paper shows some LLMs handle it better than others.

📄 mariateleki.github.io/pdf/HorrorCo...

#INTERSPEECH2025 #ConversationalAI #RecSys

Glasgow Information Retrieval Group | LinkedIn Glasgow Information Retrieval Group | 298 followers on LinkedIn. Founded in 1986, the Glasgow IR Group has been at the forefront of Research & Development in search and recommendation.

📢 We're also now on LinkedIn!

Follow the Glasgow Information Retrieval Group for updates on IR research, @irglasgow.bsky.social activities, events, and collaborations:

🔗 linkedin.com/company/glasgow-information-retrieval-group

#InformationRetrieval #IR #AI #recsys #RecommenderSystems #Glasgow

Dynamic Weighted Loss Improves Sparse Domain Recommendations

The Dynamic Weighted Loss adapts weights by interaction density to improve sparse-category recommendations. Tested on MovieLens and Amazon Electronics, it exceeded SIGMA. Read more: getnews.me/dynamic-weighted-loss-im... #dwloss #recsys

#SIGIR2025 had 1071 full submissions across all tracks. Considering that a large number of those are on search and IR, this track may very well be the largest outlet for #recsys research. (although, given recent numbers at @neuripsconf.bsky.social and @aaai.org, this may be very well be contested).

We (@md.ekstrandom.net and @ssahebi.bsky.social) received 706 abstracts in the User Modeling, Personalization, and Recommendation track at #TheWebConference2026. #usermodeling #personalization #recsys

"Leveraging Geometric Insights in Hyperbolic Triplet Loss for Improved Recommendations" poster

Let's start sharing a few posters/papers I found interesting at #RecSys this year.

1. Leveraging Geometric Insights in Hyperbolic Triplet Loss for Improved Recommendations

Very interesting. Some further work needed to convince me 100%, like a Figure 2 but for the original Euclidian Triplet Loss.

Very happy to see a full room at our #RecSys2025 tutorial "Hands on #QuantumComputing for Recommender Systems" presented by me and Paolo Cremonesi. We covered gate model QC and #quantumannealing with examples on community detection and feature selection.
Materials: qclef.dei.unipd.it

#recsys

And @drchole.bsky.social's presentation of our (with md.ekstrandom.net) call for a revised research agenda for #recsys

Particularly happy about Alex Tuzhilin's keynote on exploitation VS exploration in #recsys

#beyond2025 at @recsys.bsky.social is wrapping up. We've had a great program on the interdisciplinary roots of #RecSys.

#RecSys #RecSys2025 Very excited to welcome as many of you as can come to my city next year for #RecSys2026!

Recommending in the age of AI @xamat.bsky.social giving his keynote at #recsys2025 @recsys.bsky.social ("A very personal review of #RecSys")

Dynamic Temporal‑Aware LLM Simulator Boosts Recommender Feedback

DyTA4Rec adds a dynamic updater that continuously refines user profiles with new interaction data, and researchers report a marked improvement in alignment with real user logs. Read more: getnews.me/dynamic-temporal-aware-l... #dytarec #recsys

First post on Bluesky.
Any ML engineers working on recsys get tired of repeatedly building and (re)training N models for N different projects? What about a unified model for all N projects? Let me know your thoughts. #recsys #ml

MuRS 2025 - Program Program

The #MuRS2025 workshop on music recommenders is in full swing. We have a great program w/9 paper presentations+a keynote by Anna Huang.

Engaging participants in a packed room!

The happy organizers:@lorenzoporcaro.bsky.social, Andrés Ferraro,+me.
sites.google.com/view/murs-20...
#RecSys2025 #recsys

Learnable Fully‑Interacted Two‑Tower Model Boosts Pre‑Ranking Efficiency

FIT adds a Meta Query Module and a Lightweight Similarity Scorer, improving pre‑ranking recall and precision; it was presented at SIGIR 2025, according to the authors. Read more: getnews.me/learnable-fully-interact... #recsys #sigir2025