Picture for Hung-yi Lee

Hung-yi Lee

Transferring Textual Preferences to Vision-Language Understanding through Model Merging

Add code
Feb 19, 2025
Viaarxiv icon

Speech-FT: A Fine-tuning Strategy for Enhancing Speech Representation Models Without Compromising Generalization Ability

Add code
Feb 18, 2025
Viaarxiv icon

Gender Bias in Instruction-Guided Speech Synthesis Models

Add code
Feb 08, 2025
Viaarxiv icon

BreezyVoice: Adapting TTS for Taiwanese Mandarin with Enhanced Polyphone Disambiguation -- Challenges and Insights

Add code
Jan 29, 2025
Viaarxiv icon

Clear Minds Think Alike: What Makes LLM Fine-tuning Robust? A Study of Token Perplexity

Add code
Jan 24, 2025
Viaarxiv icon

CodecFake-Omni: A Large-Scale Codec-based Deepfake Speech Dataset

Add code
Jan 14, 2025
Figure 1 for CodecFake-Omni: A Large-Scale Codec-based Deepfake Speech Dataset
Figure 2 for CodecFake-Omni: A Large-Scale Codec-based Deepfake Speech Dataset
Figure 3 for CodecFake-Omni: A Large-Scale Codec-based Deepfake Speech Dataset
Figure 4 for CodecFake-Omni: A Large-Scale Codec-based Deepfake Speech Dataset
Viaarxiv icon

Spectral-Aware Low-Rank Adaptation for Speaker Verification

Add code
Jan 07, 2025
Figure 1 for Spectral-Aware Low-Rank Adaptation for Speaker Verification
Figure 2 for Spectral-Aware Low-Rank Adaptation for Speaker Verification
Figure 3 for Spectral-Aware Low-Rank Adaptation for Speaker Verification
Figure 4 for Spectral-Aware Low-Rank Adaptation for Speaker Verification
Viaarxiv icon

Detecting the Undetectable: Assessing the Efficacy of Current Spoof Detection Methods Against Seamless Speech Edits

Add code
Jan 07, 2025
Figure 1 for Detecting the Undetectable: Assessing the Efficacy of Current Spoof Detection Methods Against Seamless Speech Edits
Figure 2 for Detecting the Undetectable: Assessing the Efficacy of Current Spoof Detection Methods Against Seamless Speech Edits
Figure 3 for Detecting the Undetectable: Assessing the Efficacy of Current Spoof Detection Methods Against Seamless Speech Edits
Figure 4 for Detecting the Undetectable: Assessing the Efficacy of Current Spoof Detection Methods Against Seamless Speech Edits
Viaarxiv icon

Safeguard Fine-Tuned LLMs Through Pre- and Post-Tuning Model Merging

Add code
Dec 27, 2024
Viaarxiv icon

Enhancing Multilingual ASR for Unseen Languages via Language Embedding Modeling

Add code
Dec 21, 2024
Viaarxiv icon