Blog(1)

• Learning to Align Temporal Sequences

  Temporal sequences (e.g., videos) are an appealing data source as they provide a rich source of information and additional constraints to leverage in learning. By far the main focus on temporal sequence analysis in computer vision has been on learning representations (i.e., compact abstractions of the input data) targeting high-level distinctions between signals (e.g., action classification, “What action is present in the video?”).

Research Areas(0)

Publications(30)

• Learning to Wait: Real Streaming Speech-to-Text Translation with an LLM

  AuthorTitouan Parcollet, Rogier C. van Dalen

  PublishedInterspeech 2026

  Date2026-09-28

• Listening Like a Judge: A Music-Aware Framework for Automatic Singing Performance Evaluation

  AuthorNeelam Saini, Sourav Ghosh

  PublishedInterspeech 2026 (Interspeech)

  Date2026-09-27

• Architecture-Agnostic Test-Time Adaptation via Backprop-Free Embedding Alignment

  AuthorYoung D. Kwon

  PublishedInternational Conference on Learning Representation (ICLR)

  Date2026-04-23

View More

News(7)

• [Blog] LittleBit-2: Maximizing the Spectral Energy Gain in Sub-1-Bit LLMs via Latent Geometry Alignment

  Large Language Models (LLMs) continue to deliver remarkable capabilities, but their memory footprint remains a rigid barrier to deployment.

• [Blog] LANTERN: Language Model Assessmenton Noisy and Transformed Tasks for Understanding Error and Robustness Nuances

  Large Language Models (LLMs) such as GPT‑3, LLaMA, and ChatGPT have made significant strides in natural language generation, reasoning, and instruction following. These advancements have been fuelled by scaling laws, massive pretraining, and alignment techniques such as supervised fine‑tuning and preference optimization.

• [Blog] Adversarial Deep Metric Learning for Cross-Modal Audio-Text Alignment in Open-Vocabulary Keyword Spotting

  Text-enrolled KWS with phoneme-level deep metric learning and adversarial training. We align acoustic and text embeddings across modalities and achieve SOTA on WSJ & LibriPhrase.

View More

Others(0)