Deadline extended! The IR-RAG Workshop @ SIGIR 2025 has postponed its submission deadline to May 5. Don’t miss the chance to contribute to the future of Retrieval-Augmented Generation! Submit: easychair.org/conferences/?c… Info: coda.io/@rstless-group… #RAG #IR #SIGIR #AI #NLP #LLM

🚨 Deadline Extended! The submission deadline for the IR-RAG Workshop SIGIR 2025 has been extended to May 5, 2025! 🗓️✍️ If you're working on Retrieval-Augmented Generation, feel free to check our website lnkd.in/dx93i-6h #SIGIR2025 #LLMs #RAG

GREEN: an inference-time method using our new dataset EcoTaskSet to recommend Pareto-optimal models balancing accuracy and emissions. Our UI lets users choose dataset & accuracy-emissions trade-off and recommends the best model 📚arxiv.org/pdf/2505.01468 💻 github.com/antoniopurific…

"The Distracting Effect: Understanding Irrelevant Passages in RAG" co-authored by Florin Cuconasu was accepted at #ACL2025NLP main conference! This work analyzes how different types of distracting passages affect LLMs and achieves 7.5% accuracy gains via targeted fine-tuning.

1/4 Excited to share that our paper "Beyond Position: the emergence of wavelet-like properties in Transformers" has been accepted to #ACL2025NLP! 🎉 Fabrizio Silvestri Valeria Ruscio

1/n While positional bias affects LLMs, it's actually not a real problem for practical RAG systems. In our new paper, we tackle the question “Do RAG Systems Suffer From Positional Bias?”

In essence, lost-in-the-middle is not a problem for practical RAG applications; the real challenge lies in retrieval quality and LLM robustness to distractors. Big thanks to my incredible co-authors Simone Filice, Guy Horowitz, Yoelle Maarek, and Fabrizio Silvestri for their amazing work!

🆕 Il presidente Antonio Naddeo in una nuova puntata di Pazzesco ma vero, la serie che combatte gli stereotipi sulla pubblica amministrazione. In questo episodio il tema è l'intelligenza artificiale (AI) e il suo rapporto con la PA. L'ospite intervistato è il Professor Fabrizio

Excited to share that our paper “Joint Factual and Counterfactual Explanations for Top-k GNN-based Recommendations”has been accepted to ACM Transactions on Recommender Systems (ACM TORS) 🎉🎉🎉. Many thanks to Gabriele Tolomei Fabrizio Silvestri Yongfeng Zhang. Here is the accepted version of our paper: dl.acm.org/doi/10.1145/37…

Next Monday will start the #IJCNN2025 conference! And our group will present lots of papers! On July 5th at 2.30 PM CET in Room Loyola francesco caso will present “Renormalized Graph Representations for Node Classification”. Preprint: arxiv.org/pdf/2306.00707 #Classication #ML

On July 4th from 11 to 12.30, room Belli Giulia Di Teodoro will present the work “A theoretical analysis of recommendation loss functions under negative sampling”. On July 2nd will present the Competition on Learning with Noisy Graph Labels from 10.30 to 12.30, room Manzoni. #IJCNN25

Our Federico Siciliano yesterday was interviewed by Hospodárske noviny, a Slovak newspaper, because of his expertise in AI. They discussed about the current status of AI, with a focus on the European Union vision of the topic. Bringing Italian style to Slovak 🇸🇰🇮🇹 #AI #Research #Recommender

Andrea Giuseppe Di Francesco, Giulia Di Teodoro and Antonio Purificato presenting their work in the special session for Neural IR and RecSys at #IJCNN25, where Giulia Di Teodoro also served as session chair! Lots of insights on Recommender systems. Link to the papers in the comments! #IR #Research

Last #IJCNN25 paper from our group, presented by francesco caso during the topological deep learning special session. Renormalized Graph Representations for Node Classification, co-authored with Andrea Bacciu , Giovanni Trappolini, Pietro Lio' and Fabrizio Silvestri.

📢 Our TIM-Rec dataset has been accepted at the ACM RecSys 2025 Resource Track! TIM-Rec tackles multi-item upselling in a Telco setting with sparse feedback, a real-world challenge for recommender systems. Congrats to the team Alessandro Sbandi Federico Siciliano Fabrizio Silvestri! #RS

Today Giulio D'Erasmo presented two works in the poster session of #ICTIR25 Eclipse: Contrastive Dimension Importance Estimation with Pseudo-Irrelevance Feedback for Dense Retrieval arxiv.org/pdf/2412.14967 and as a proxy presenter: QPP-RA: Aggregating Large Language Model Rankings

RSTLess people at #ACL2025 presenting 2 papers + meeting with friends! Florin Cuconasu presenting “The Distracting Effect: Understanding Irrelevant Passages in RAG” and Valeria Ruscio presenting “Beyond Position: the emergence of wavelet-like properties in Transformers”. #NLP

🚀 Breaking News: The 3rd run of #AutoBench is out! Ranking 33 LLMs with 300k+ ranks, unprecedent correlation with leading benchmarks (92% on Artificial Analysis Index). Our most epic benchmark yet. And that's not all: announcing the official site autobench.org 🎉 1/14

Exciting news!!🤟 “What are you sinking? A geometric approach on attention sink" got accepted to NeurIPS Conference spotlight! Valeria Ruscio Fabrizio Silvestri Link to the paper: arxiv.org/abs/2508.02546

🚨 ACM The Web Conference 2026 deadlines are just around the corner! Don’t miss your chance to submit your work! 📅 Research/Industry: Abs Sept 30, Papers Oct 7 📅 Short Papers: Abs Nov 10, Papers Nov 17 📍 Dubai · Apr 13–17, 2026 👉 www2026.thewebconf.org/important-date… #TheWebConf #CFP