Skip to main content

Paper Reviews by AI

2024

Fourier Position Embedding: Enhancing Attention's Periodic Extension for Length Generalization
·1717 words·9 mins· loading · loading
AI Generated πŸ€— Daily Papers Natural Language Processing Large Language Models 🏒 Tsinghua University
FoPE: 주파수 μ˜μ—­ νŠΉμ§• κ°œμ„ μœΌλ‘œ κΈ΄ λ¬Έλ§₯ 길이 μΌλ°˜ν™” 달성!
DRT-o1: Optimized Deep Reasoning Translation via Long Chain-of-Thought
·366 words·2 mins· loading · loading
AI Generated πŸ€— Daily Papers Natural Language Processing Machine Translation 🏒 Tencent AI Lab
DRT-01 λͺ¨λΈμ€ μž₯문의 사고 과정을 ν™œμš©ν•˜μ—¬ λ¬Έν•™ λ²ˆμ—­μ˜ 정확도와 μœ μ°½μ„±μ„ 크게 ν–₯μƒμ‹œμΌ°μŠ΅λ‹ˆλ‹€.
Diving into Self-Evolving Training for Multimodal Reasoning
·2584 words·13 mins· loading · loading
AI Generated πŸ€— Daily Papers Natural Language Processing Large Language Models 🏒 Hong Kong University of Science and Technology
M-STAR: λ‹€λͺ¨λ‹¬ 좔둠을 μœ„ν•œ 자기 진화 ν›ˆλ ¨μ˜ μƒˆλ‘œμš΄ ν”„λ ˆμž„μ›Œν¬λ₯Ό μ œμ‹œ!
Deliberation in Latent Space via Differentiable Cache Augmentation
·2751 words·13 mins· loading · loading
AI Generated πŸ€— Daily Papers Natural Language Processing Large Language Models 🏒 Google DeepMind
λŒ€κ·œλͺ¨ μ–Έμ–΄ λͺ¨λΈμ˜ μΆ”λ‘  μ„±λŠ₯을 ν–₯μƒμ‹œν‚€λŠ” μƒˆλ‘œμš΄ 방법인 β€˜μ°¨λ³„ κ°€λŠ₯ν•œ μΊμ‹œ 증강’ 기법 μ œμ‹œ!
B-STaR: Monitoring and Balancing Exploration and Exploitation in Self-Taught Reasoners
·1797 words·9 mins· loading · loading
AI Generated πŸ€— Daily Papers Natural Language Processing Large Language Models 🏒 Hong Kong University of Science and Technology
B-STAR: 자기 ν•™μŠ΅ μΆ”λ‘ μžμ—μ„œ 탐색과 ν™œμš©μ˜ κ· ν˜•μ„ λͺ¨λ‹ˆν„°λ§ν•˜κ³  μ‘°μ •ν•˜μ—¬ μ„±λŠ₯을 ν–₯μƒμ‹œν‚€λŠ” μƒˆλ‘œμš΄ ν”„λ ˆμž„μ›Œν¬
Revisiting In-Context Learning with Long Context Language Models
·3818 words·18 mins· loading · loading
AI Generated πŸ€— Daily Papers Natural Language Processing Large Language Models 🏒 Google DeepMind
μž₯λ¬Έ μ»¨ν…μŠ€νŠΈ μ–Έμ–΄ λͺ¨λΈμ—μ„œ μ •κ΅ν•œ μƒ˜ν”Œ 선택 μ „λž΅λ³΄λ‹€ λ¬΄μž‘μœ„ μƒ˜ν”Œλ§μ΄ ICL μ„±λŠ₯ ν–₯상에 더 효과적이며, 데이터 증강을 톡해 μ €μžμ› μž‘μ—… μ„±λŠ₯을 5% ν–₯μƒμ‹œμΌ°λ‹€λŠ” λ†€λΌμš΄ 연ꡬ κ²°κ³Όλ₯Ό λ°œν‘œ!
OpenRFT: Adapting Reasoning Foundation Model for Domain-specific Tasks with Reinforcement Fine-Tuning
·1880 words·9 mins· loading · loading
AI Generated πŸ€— Daily Papers Natural Language Processing Large Language Models 🏒 Beijing Jiaotong University
OpenRFTλŠ” μ œν•œλœ 도메인 νŠΉμ • 데이터λ₯Ό μ‚¬μš©ν•˜μ—¬ 일반적인 μΆ”λ‘  λͺ¨λΈμ„ λ―Έμ„Έ μ‘°μ •ν•˜λŠ” μƒˆλ‘œμš΄ 방법을 μ œμ‹œν•©λ‹ˆλ‹€.
Distilled Decoding 1: One-step Sampling of Image Auto-regressive Models with Flow Matching
·3113 words·15 mins· loading · loading
AI Generated πŸ€— Daily Papers Computer Vision Image Generation 🏒 Tsinghua University
단일 단계 μƒ˜ν”Œλ§μœΌλ‘œ 이미지 μžλ™ νšŒκ·€ λͺ¨λΈ 속도λ₯Ό 획기적으둜 ν–₯μƒμ‹œν‚¨ 증λ₯˜ λ””μ½”λ”©(DD) 기법 μ œμ•ˆ!
NILE: Internal Consistency Alignment in Large Language Models
·2709 words·13 mins· loading · loading
AI Generated πŸ€— Daily Papers Natural Language Processing Large Language Models 🏒 Chinese University of Hong Kong
NILE ν”„λ ˆμž„μ›Œν¬λŠ” LLM의 λ‚΄λΆ€ 지식과 IFT λ°μ΄ν„°μ…‹μ˜ 세계 지식 κ°„ 일관성을 λ†’μ—¬ LLM μ„±λŠ₯을 μ΅œλŒ€ 68.5%κΉŒμ§€ ν–₯μƒμ‹œν‚΅λ‹ˆλ‹€.
LearnLM: Improving Gemini for Learning
·3761 words·18 mins· loading · loading
AI Generated πŸ€— Daily Papers AI Applications Education 🏒 Google DeepMind
LearnLM은 ꡐ윑적 λ§₯λ½μ—μ„œ μƒμ„±ν˜• AI의 νŽ˜λ‹€κ³ μ§€(Pedagogy)λ₯Ό ν–₯μƒμ‹œν‚¨ λͺ¨λΈμž…λ‹ˆλ‹€. κ΅μ‚¬λ‚˜ κ°œλ°œμžκ°€ μ›ν•˜λŠ” νŽ˜λ‹€κ³ μ§€μ  νŠΉμ„±μ„ λͺ¨λΈμ— μ£Όμž…ν•˜λŠ” μƒˆλ‘œμš΄ ν”„λ ˆμž„μ›Œν¬λ₯Ό 톡해 κΈ°μ‘΄ λͺ¨λΈλ³΄λ‹€ ν•™μŠ΅ 효과λ₯Ό 31% ν–₯μƒμ‹œμΌ°μŠ΅λ‹ˆλ‹€.
Toward Robust Hyper-Detailed Image Captioning: A Multiagent Approach and Dual Evaluation Metrics for Factuality and Coverage
·2414 words·12 mins· loading · loading
AI Generated πŸ€— Daily Papers Computer Vision Visual Question Answering 🏒 Seoul National University
μ΄ˆμ •λ°€ 이미지 μΊ‘μ…˜ μƒμ„±μ˜ ν™˜κ° 문제 해결을 μœ„ν•΄, LLM-MLLM ν˜‘μ—… 기반의 닀쀑 μ—μ΄μ „νŠΈ μ‹œμŠ€ν…œ(CapMAS)을 μ œμ•ˆν•˜μ—¬ 사싀성과 포괄성을 λ†’μ˜€μŠ΅λ‹ˆλ‹€.
Multi-LLM Text Summarization
·2623 words·13 mins· loading · loading
AI Generated πŸ€— Daily Papers Natural Language Processing Text Summarization 🏒 UC Santa Cruz
λ‹€μˆ˜μ˜ κ±°λŒ€ μ–Έμ–΄ λͺ¨λΈ(LLM)을 ν™œμš©ν•œ ν˜μ‹ μ μΈ μž₯λ¬Έ μš”μ•½ ν”„λ ˆμž„μ›Œν¬κ°€ μ œμ‹œλ˜μ–΄ μš”μ•½ ν’ˆμ§ˆμ„ μ΅œλŒ€ 3λ°° ν–₯μƒμ‹œμΌ°μŠ΅λ‹ˆλ‹€!
MotiF: Making Text Count in Image Animation with Motion Focal Loss
·2819 words·14 mins· loading · loading
AI Generated πŸ€— Daily Papers Computer Vision Video Understanding 🏒 Brown University
MotiF: μ›€μ§μž„μ— μ΄ˆμ μ„ 맞좘 손싀 ν•¨μˆ˜λ‘œ ν…μŠ€νŠΈ 기반 이미지 μ• λ‹ˆλ©”μ΄μ…˜ κ°œμ„ 
Ensembling Large Language Models with Process Reward-Guided Tree Search for Better Complex Reasoning
·4085 words·20 mins· loading · loading
AI Generated πŸ€— Daily Papers Natural Language Processing Large Language Models 🏒 Microsoft Research
λŒ€κ·œλͺ¨ μ–Έμ–΄ λͺ¨λΈλ“€μ˜ 앙상블을 톡해 λ³΅μž‘ν•œ μΆ”λ‘  문제λ₯Ό λ”μš± 효과적으둜 ν•΄κ²°ν•˜λŠ” μƒˆλ‘œμš΄ ν”„λ ˆμž„μ›Œν¬, LE-MCTSλ₯Ό μ œμ•ˆν•©λ‹ˆλ‹€!
CLEAR: Conv-Like Linearization Revs Pre-Trained Diffusion Transformers Up
·3581 words·17 mins· loading · loading
AI Generated πŸ€— Daily Papers Computer Vision Image Generation 🏒 National University of Singapore
CLEAR: μ„ ν˜•ν™”λœ μ–΄ν…μ…˜μœΌλ‘œ 고해상도 이미지 생성 속도λ₯Ό 획기적으둜 높이닀!
UIP2P: Unsupervised Instruction-based Image Editing via Cycle Edit Consistency
·2616 words·13 mins· loading · loading
AI Generated πŸ€— Daily Papers Computer Vision Image Generation 🏒 ETH Zurich
비지도 ν•™μŠ΅ 기반 μˆœν™˜ νŽΈμ§‘ 일관성(CEC) ν™œμš©, μ§€μ‹œμ–΄ 기반 이미지 νŽΈμ§‘μ˜ μƒˆλ‘œμš΄ 지평을 μ—΄λ‹€!
TOMG-Bench: Evaluating LLMs on Text-based Open Molecule Generation
·3930 words·19 mins· loading · loading
AI Generated πŸ€— Daily Papers Natural Language Processing Large Language Models 🏒 Hong Kong Polytechnic University
TOMG-Bench: LLM 기반 μ˜€ν”ˆ λΆ„μž 생성 벀치마크 μ œμ‹œ! 25개 LLM 평가 및 μƒˆλ‘œμš΄ instruction tuning 데이터셋 OpenMolIns 곡개둜, μ˜€ν”ˆμ†ŒμŠ€ LLM의 μ„±λŠ₯ ν–₯상 및 λΆ„μž 발견의 μƒˆλ‘œμš΄ κ°€λŠ₯μ„± μ œμ‹œ!
Taming Multimodal Joint Training for High-Quality Video-to-Audio Synthesis
·1340 words·7 mins· loading · loading
AI Generated πŸ€— Daily Papers Multimodal Learning Multimodal Generation 🏒 University of Illinois Urbana-Champaign
κ³ ν’ˆμ§ˆ λΉ„λ””μ˜€-μ˜€λ””μ˜€ 합성을 μœ„ν•œ ν˜μ‹ μ μΈ 닀쀑 λͺ¨λ“œ 쑰인트 ν•™μŠ΅ ν”„λ ˆμž„μ›Œν¬ MMAudio μ œμ•ˆ!
RobustFT: Robust Supervised Fine-tuning for Large Language Models under Noisy Response
·2295 words·11 mins· loading · loading
AI Generated πŸ€— Daily Papers Natural Language Processing Large Language Models 🏒 Peking University
ROBUSTFTλŠ” 작음이 ν¬ν•¨λœ 응닡 μ•„λž˜μ—μ„œ λŒ€κ·œλͺ¨ μ–Έμ–΄ λͺ¨λΈμ˜ κ°•κ±΄ν•œ 지도 ν•™μŠ΅ λ―Έμ„Έ 쑰정을 μœ„ν•œ ν”„λ ˆμž„μ›Œν¬λ‘œ, 작음 감지 및 μž¬λΌλ²¨λ§μ„ 톡해 ν•˜λ₯˜ μž‘μ—… μ„±λŠ₯을 ν–₯μƒμ‹œν‚΅λ‹ˆλ‹€.
ReMoE: Fully Differentiable Mixture-of-Experts with ReLU Routing
·4863 words·23 mins· loading · loading
AI Generated πŸ€— Daily Papers Natural Language Processing Large Language Models 🏒 Tsinghua University
ReLU λΌμš°νŒ…μ„ μ‚¬μš©ν•˜λŠ” μ™„μ „ λ―ΈλΆ„ κ°€λŠ₯ν•œ MoE μ•„ν‚€ν…μ²˜ ReMoEλ₯Ό 톡해 λŒ€κ·œλͺ¨ μ–Έμ–΄ λͺ¨λΈμ˜ ν™•μž₯μ„±κ³Ό νš¨μœ¨μ„±μ„ 획기적으둜 κ°œμ„ ν–ˆμŠ΅λ‹ˆλ‹€!