Fine-Tuning LLMs for Translation: General Forgetting Mitigation Does Not Preserve MT-Specific Instruction Following
Paper • 2609.28395 • Published • 1
AI & ML interests
Machine Learning and Human Language Technology Group, RWTH Aachen University. Topics: Automatic speech recognition (ASR), language modeling (LM), machine translation (MT), text-to-speech (TTS), natural language processing (NLP), language technologies
Recent Activity
dthulke
authored a
paper 26 days ago
dthulke
authored 2
papers about 1 month ago
dthulke
authored a
paper about 1 year ago
dthulke
updated a
collection about 1 year ago
dthulke
authored 2
papers over 1 year ago
albertzeyer
updated a
model over 1 year ago
albertzeyer
published a
model over 1 year ago
albertzeyer
authored 10
papers over 1 year ago
Language Modeling with Deep Transformers
Paper • 1905.04226 • Published • 1
Monotonic segmental attention for automatic speech recognition
Paper • 2210.14742 • Published
Librispeech Transducer Model with Internal Language Model Prior Correction
Paper • 2104.03006 • Published
Why does CTC result in peaky behavior?
Paper • 2105.14849 • Published
A systematic comparison of grapheme-based vs. phoneme-based label units for encoder-decoder-attention models
Paper • 2005.09336 • Published
A study of latent monotonic attention variants
Paper • 2103.16710 • Published
A New Training Pipeline for an Improved Neural Transducer
Paper • 2005.09319 • Published
RETURNN as a Generic Flexible Neural Toolkit with Application to Translation and Speech Recognition
Paper • 1805.05225 • Published
Improved training of end-to-end attention models for speech recognition
Paper • 1805.03294 • Published
A Comprehensive Study of Deep Bidirectional LSTM RNNs for Acoustic Modeling in Speech Recognition
Paper • 1606.06871 • Published
dthulke
authored a
paper almost 3 years ago