Peeters, Rafii, Fuentes, Duan, Benetos, Nam, Mitsufuji: Twenty-Five Years of MIR Research: Achievements, Practices, Evaluations, and Future Challenges https://arxiv.org/abs/2511.07205 https://arxiv.org/pdf/2511.07205 https://arxiv.org/html/2511.07205

Termeh Taheri, Yinghao Ma, Emmanouil Benetos: SAR-LM: Symbolic Audio Reasoning with Large Language Models https://arxiv.org/abs/2511.06483 https://arxiv.org/pdf/2511.06483 https://arxiv.org/html/2511.06483

📢 Join our Conversational AI Reading Group! 📅 Thursday, Nov 6th | 11 AM - 12 PM EST 🎙 Speaker: Emmanouil Benetos ( @emmanouilb.bsky.social ) - Queen Mary University of London 📖 Topic: "Machine learning paradigms for music and audio understanding" 🔗 Details: (poonehmousavi.github.io/rg)

🎶 [C4DM Seminar] Advancing Music Experience Through Music Information Research 🕒 3–4 PM, 5th November 📍 TBC We’re excited to welcome Dr. Masataka Goto, Senior Principal Researcher at the National Institute of Advanced Industrial Science and Technology (AIST), Japan.

our inventor @antorrisi.bsky.social with poster on #VocalEcho: Closed-loop system for vocal interactions #BioDCASE at 12:21 CET (11.21 UK time) 🐥 🔀 🤖🎈 Streaming dcase.community/workshop2025... @c4dm.bsky.social @emmanouilb.bsky.social @preparedmindslab.bsky.social @elisabettaversace.bsky.social

Aditya Bhattacharjee, Marco Pasini, Emmanouil Benetos: Scalable Evaluation for Audio Identification via Synthetic Latent Fingerprint Generation https://arxiv.org/abs/2509.18620 https://arxiv.org/pdf/2509.18620 https://arxiv.org/html/2509.18620

New preprint on "Computational Hermeneutics," co-authored by too many people to list in one post. TL;DR: GenAI is a cultural technology, and needs to be evaluated in ways that recognize situatedness, plurality, and ambiguity as the conditions of meaning — not noise to be minimized.

Sungkyun Chang, Simon Dixon, Emmanouil Benetos: RUMAA: Repeat-Aware Unified Music Audio Analysis for Score-Performance Alignment, Transcription, and Mistake Detection https://arxiv.org/abs/2507.12175 https://arxiv.org/pdf/2507.12175 https://arxiv.org/html/2507.12175

Ludovic Tuncay, Etienne Labb\'e, Emmanouil Benetos, Thomas Pellegrini: Audio-JEPA: Joint-Embedding Predictive Architecture for Audio Representation Learning https://arxiv.org/abs/2507.02915 https://arxiv.org/pdf/2507.02915 https://arxiv.org/html/2507.02915

Our paper on automatic sample identification (arxiv.org/abs/2506.14684) was accepted at @ismir_conf 2025! 🎵🎶 We propose an architecture that can detect music samples that have been reused in new compositions, even after pitch-shifting, time-stretching, and other transformations!🧵

CMI-Bench, a comprehensive music instruction following benchmark, evaluates audio-text LLMs on diverse MIR tasks, revealing performance gaps compared to supervised models and biases.

🎉 Congrats to Sungkyun Chang, Simon Dixon, and Emmanouil Benetos — their submission earned 2nd place at the 2025 Automatic Music Transcription Challenge! 🥈 ai4musicians.org/transcriptio... #C4DM #MusicAI #AMTChallenge2025.

**Lyrics Transcription:** Consistency loss aligns vocal and mixture encoder representations, improving transcription on music mixtures.

Excited to share our new paper: arxiv.org/abs/2505.06224! We introduce a unified framework for evaluating model representations beyond downstream tasks, and use it to uncover some interesting insights about the structure of representations that challenge conventional wisdom 🔍🧵

how much music data do we actually need to pretrain effective music representation learning models? we decided to systematically investigate this in our latest #ICASSP2025 paper with @emmanouilb.bsky.social and Johan Pauwels

We are very thrilled to share our contributions to this year's ICASSP - 14 papers! For a list of the articles, please refer to: www.c4dm.eecs.qmul.ac.uk/news/2025-03...

Exciting research update! EECS PhD students have developed a novel approach that enables large language models (LLMs) to “hear” and “understand” sound - a breakthrough in multimodal generative #AI: www.qmul.ac.uk/eecs/news-an...

Introduced Audio-FLAN, a large-scale instruction-tuning dataset for unified audio-language models; dataset covers 80 tasks and over 100 million instances, available on HuggingFace and GitHub.

Next Tuesday, 3/12 at 2pm, we will host a seminar by Manvi Agarwal on 'Fast Structure-informed Positional Encoding for Music Generation'. More info at: www.c4dm.eecs.qmul.ac.uk/news/2024-03...