Robust Fusion of Semantic and Behavioural Signals for LLM Reranking in Personalised Search
@asash et al. at Spotify show that behavioral stats boost an LLM reranker but become a shortcut on rare queries, and propose paired training to stay robust.
📝 https://t.co/sG1HG2BVL9
Do Sequential Recommendation Benchmarks Really Require Higher-Order Sequence Modelling?
@asash et al. at show simple pairwise transition models often beat Transformer recommenders, questioning what these benchmarks measure.
📝 https://t.co/4onRL6Cduk
👨🏽💻 https://t.co/bGqm6fQl14
Hypothesis-Driven Shelf Generation for Personalised Recommendation
@asash et al. at Spotify replace fixed shelf templates with LLM-generated natural-language hypotheses, then fulfill them via constrained generative retrieval.
📝 https://t.co/qcoqZIWmzs
For anyone worried their LLM might be making stuff up, we made a budget‐friendly truth serum (semantic entropy + Bayesian). See for yourself: https://t.co/gq8oFP5Eqr
Paper: https://t.co/s6SuQW4sAM
LLMs for estimating positional bias in logged interaction data
@asash et al. at Viator use LLMs to estimate position bias in logged user interaction data as an alternative to online experimentation.
📝https://t.co/hoLhBe49KO
Thrilled to join @asash on the Recsperts podcast! Thanks @MarcelKurovski for having us. We had a blast discussing our #RecSys research & transformer-based sequential recommendation. Tune in on your favorite podcast platform https://t.co/ROi9MhWCGK
Huge congratulations to @macavaney on receiving the prestigious ACM SIGIR Early Career Researcher Award in the research category! This well-deserved recognition highlights the excellence & impact of his work in the IR community 👏🎉#sigir2025
Cc @GlasgowCS@UofGlasgow@ACMSIGIR
2nd day at #SIGIR2025 ☀️
After the first keynote, a great talk by @asash on combining joint product quantization and dynamic pruning to accelerate top-k computation. No need to compute all scores anymore! 😎
Work with @craig_macdonald and @ntonellotto .
Just presented our work at #SIGIR2025. @craig_macdonald gave a great overview. If you have any questions, feel free to catch me during the coffee break!
.@Asash is talking about our #sigir2025 📄 applying dynamic pruning ideas in sequential recommended systems, using sub-id representations w/ myself and @ntonellotto
If you're at #SIGIR2025 and interested in large-scale RecSys, pop by my talk on Monday! I'll be presenting our paper (w/ @craig_macdonald and @ntonellotto): 'Efficient Recommendation with Millions of Items by Dynamic Pruning of Sub-Item Embeddings'.
🔗 https://t.co/gxJYGcJbTo
📢 We're off to #SIGIR2025 in Padova!
A large contingent of our students & staff will be at the main conference + tutorials, workshops & #ICTIR2025.
Let’s connect if you’re around! 🤝
Also: we're hiring in #FinTech — DM us if you're interested!
#IR#Hiring#Research#ACMSIGIR
Huge congratulations to our brilliant PhD graduates: @tjaenich, @mvlacho1, @asash - It’s been a joy having you @GlasgowCS. We’re so proud of all you’ve achieved and can’t wait to see what amazing things you’ll do next. Wishing you all the best! 🎉👏 #PhDGraduation#PhDSuccess
We’re excited to announce that WSDM 2026 will take place in Boise, Idaho, from February 22 to 26, 2026!
Stay tuned and visit 🔗https://t.co/vRdaZ9G7fV for updates!
#WSDM2026
Happy to share that the OARS@KDD2025 workshop accepts our work. In this work, we turn a sequential recommendation system into a semantic search using generative language models. Apparently, this works really well.
1/9 Happy to share that our paper GLoSS: Generative Language Models with Semantic Search for
Sequential Recommendation is accepted at the KDD OARS workshop! 🎉
Paper, code: https://t.co/TrgHgCnuPC
This is joint work with my wonderful collaborators @asash and Juba Ziani.
Delighted that @asash passed his 🎓 PhD defense this morning, without corrections. Thanks to @pcastells and Nicolas Pugeault for their thorough examination of the thesis, and @mobilelearnfeed for convening the defense!