Paper Review
1.[Paper Review] SpeechGPT
SpeechGPT: Empowering Large Language Models with Intrinsic Cross-Modal Conversational Abilities
2.[Paper Review] SVD-LLM
SVD-LLM: Truncation-aware Singular Value Decomposition for Large Language Model Compression
3.[Paper Review] Mitigating the Impact of Outlier Channels
Mitigating the Impact of Outlier Channels for Language Model Quantization with Activation Regularization
4.[Paper Review] SpeechGPT-Gen
SpeechGPT-Gen: Scaling Chain-of-Information Speech Generation
5.[Paper Review] Slimming Large Language Models by Cutting Off the Layers
Layer Pruning 논문
6.[Paper Review] LaCo: Large Language Model Pruning via Layer Collapse
Layer Pruning
7.[Paper Review] Shortened LLaMA: A Simple Depth Pruning for Large Language Models
Layer Pruning
8.[Paper Review] ShortGPT: Layers in Large Language Models are More Redundant Than You Expect
Pruning - Depth Pruning (Layer Pruning)
9.[Paper Review] GPTQ: ACCURATE POST-TRAINING QUANTIZATION FOR GENERATIVE PRE-TRAINED TRANSFORMERS
Quantization - PTQ
10.[Paper Review] ROFORMER: ENHANCED TRANSFORMER WITH ROTARY POSITION EMBEDDING
NLP Position Embedding Method - RoPE
11.[Paper Review] KAN: Kolmogorov–Arnold Networks
KAN : MLP를 대체할 수 있을까? 다변수 함수를 단변수 함수들로 분해하고, Activation을 학습하게 함으로써, MLP의 차원의 저주 문제를 해결해보자!
12.[Paper Review] Robust Speech Recognition via Large-Scale Weak Supervision
ASR Sota: Whisper
13.[Paper Review] Diff-E: Diffusion-based Learning for Decoding Imagined Speech EEG
EEG Representation을 잘 뽑아보자! Imagined Text 에 대한 13 class Classification Task
14.[Paper Review] WavLM: Large-Scale Self-Supervised Pre-Training for Full Stack Speech Processing
Speech Representation
15.[Paper Review] Towards Good Practices for Missing Modality Robust Action Recognition
빈 모달리티 예측
16.[Paper Review] Reshape Dimensions Network for Speaker Recognition
Speaker Recognition Paper
17.[Paper Review] Segment Any Text: A Universal Approach for Robust, Efficient and Adaptable Sentence Segmentation
문장분리 Task
18.[Paper Review] Diffusion-LM Improves Controllable Text Generation
Diffusion 으로 LM 만들기, Novelty는 스타일 조절 가능!
19.[Paper Review] DINOv2: Learning Robust Visual Features without Supervision
Self Supervised Learning
20.[Paper Review] SimCSE: Simple Contrastive Learning of Sentence Embeddings
contrastive learning을 접목하여 문장 임베딩을 잘 추출하는 모델 개발