6 papers
Sequence-Aware Split Heuristic to Mitigate SM Underutilization in FlashAttention-3 Low-Head-Count Decoding
Martà Llopart Font, Javier Hernando, Cristina España-Bonet
The standard FlashAttention-3 heuristic exhibits a GPU occupancy bottleneck in low-head-count decoding configurations because it disables sequence splitting based on sequence lengt…
Revisiting Direct Speech-to-Text Translation with Speech LLMs: Better Scaling than CoT Prompting?
Oriol Pareras, Gerard I. Gállego, Gerard I. Gállego +4
Recent work on Speech-to-Text Translation (S2TT) has focused on LLM-based models, introducing the increasingly adopted Chain-of-Thought (CoT) prompting, where the model is guided t…
Listening or Reading? Evaluating Speech Awareness in Chain-of-Thought Speech-to-Text Translation
Jacobo Romero-DÃaz, Jacobo Romero-Díaz, Gerard I. Gállego +6
Speech-to-Text Translation (S2TT) systems built from Automatic Speech Recognition (ASR) and Text-to-Text Translation (T2TT) modules face two major limitations: error propagation an…
Speech-to-Text Translation with Phoneme-Augmented CoT: Enhancing Cross-Lingual Transfer in Low-Resource Scenarios
Gerard I. Gállego, Gerard I. Gállego, Oriol Pareras +4
We propose a Speech-to-Text Translation (S2TT) approach that integrates phoneme representations into a Chain-of-Thought (CoT) framework to improve translation in low-resource and z…
Breaking Language Barriers in Visual Language Models via Multilingual Textual Regularization
Iñigo Pikabea, Iñaki Lacunza, Oriol Pareras +4
Rapid advancements in Visual Language Models (VLMs) have transformed multimodal understanding but are often constrained by generating English responses regardless of the input lang…
Mass-Editing Memory with Attention in Transformers: A cross-lingual exploration of knowledge
Daniel Tamayo, Aitor Gonzalez-Agirre, Javier Hernando +1
Recent research has explored methods for updating and modifying factual knowledge in large language models, often focusing on specific multi-layer perceptron blocks. This study exp…