CatLIP: CLIP-level Visual Recognition Accuracy with 2.7x Faster Pre-training on Web-scale Image-Text Data
Paper
•
2404.15653
•
Published
•
27
Infrastructure
-
-
MoDE: CLIP Data Experts via Clustering
Paper • 2404.16030 • Published • 13 -
MoRA: High-Rank Updating for Parameter-Efficient Fine-Tuning
Paper • 2405.12130 • Published • 47 -
Reducing Transformer Key-Value Cache Size with Cross-Layer Attention
Paper • 2405.12981 • Published • 28 -
LiteVAE: Lightweight and Efficient Variational Autoencoders for Latent Diffusion Models
Paper • 2405.14477 • Published • 17 -
Thermodynamic Natural Gradient Descent
Paper • 2405.13817 • Published • 13 -
Tele-Aloha: A Low-budget and High-authenticity Telepresence System Using Sparse RGB Cameras
Paper • 2405.14866 • Published • 6 -
Transformers Can Do Arithmetic with the Right Embeddings
Paper • 2405.17399 • Published • 52 -
2BP: 2-Stage Backpropagation
Paper • 2405.18047 • Published • 23 -
VeLoRA: Memory Efficient Training using Rank-1 Sub-Token Projections
Paper • 2405.17991 • Published • 12 -
Jina CLIP: Your CLIP Model Is Also Your Text Retriever
Paper • 2405.20204 • Published • 35 -
Artificial Generational Intelligence: Cultural Accumulation in Reinforcement Learning
Paper • 2406.00392 • Published • 13 -
Block Transformer: Global-to-Local Language Modeling for Fast Inference
Paper • 2406.02657 • Published • 38 -
PLaD: Preference-based Large Language Model Distillation with Pseudo-Preference Pairs
Paper • 2406.02886 • Published • 9 -
Scaling Laws for Reward Model Overoptimization in Direct Alignment Algorithms
Paper • 2406.02900 • Published • 12 -
GenAI Arena: An Open Evaluation Platform for Generative Models
Paper • 2406.04485 • Published • 21 -
WildBench: Benchmarking LLMs with Challenging Tasks from Real Users in the Wild
Paper • 2406.04770 • Published • 28 -
Why Has Predicting Downstream Capabilities of Frontier AI Models with Scale Remained Elusive?
Paper • 2406.04391 • Published • 8 -
Boosting Large-scale Parallel Training Efficiency with C4: A Communication-Driven Approach
Paper • 2406.04594 • Published • 6 -
The Prompt Report: A Systematic Survey of Prompting Techniques
Paper • 2406.06608 • Published • 58 -
DiTFastAttn: Attention Compression for Diffusion Transformer Models
Paper • 2406.08552 • Published • 23 -
Interpreting the Weight Space of Customized Diffusion Models
Paper • 2406.09413 • Published • 18 -
Cognitively Inspired Energy-Based World Models
Paper • 2406.08862 • Published • 9 -
MLKV: Multi-Layer Key-Value Heads for Memory Efficient Transformer Decoding
Paper • 2406.09297 • Published • 4 -
A Simple and Effective L_2 Norm-Based Strategy for KV Cache Compression
Paper • 2406.11430 • Published • 23 -
MMBench-Video: A Long-Form Multi-Shot Benchmark for Holistic Video Understanding
Paper • 2406.14515 • Published • 33 -
Model Merging and Safety Alignment: One Bad Model Spoils the Bunch
Paper • 2406.14563 • Published • 30 -
Judging the Judges: Evaluating Alignment and Vulnerabilities in LLMs-as-Judges
Paper • 2406.12624 • Published • 37 -
Efficient World Models with Context-Aware Tokenization
Paper • 2406.19320 • Published • 8 -
We-Math: Does Your Large Multimodal Model Achieve Human-like Mathematical Reasoning?
Paper • 2407.01284 • Published • 75 -
MMEvalPro: Calibrating Multimodal Benchmarks Towards Trustworthy and Efficient Evaluation
Paper • 2407.00468 • Published • 34 -
μ-Bench: A Vision-Language Benchmark for Microscopy Understanding
Paper • 2407.01791 • Published • 5 -
No Training, No Problem: Rethinking Classifier-Free Guidance for Diffusion Models
Paper • 2407.02687 • Published • 22 -
Learning to (Learn at Test Time): RNNs with Expressive Hidden States
Paper • 2407.04620 • Published • 28 -
HEMM: Holistic Evaluation of Multimodal Foundation Models
Paper • 2407.03418 • Published • 9 -
MJ-Bench: Is Your Multimodal Reward Model Really a Good Judge for Text-to-Image Generation?
Paper • 2407.04842 • Published • 53 -
ANOLE: An Open, Autoregressive, Native Large Multimodal Models for Interleaved Image-Text Generation
Paper • 2407.06135 • Published • 21 -
An accurate detection is not all you need to combat label noise in web-noisy datasets
Paper • 2407.05528 • Published • 3 -
CosmoCLIP: Generalizing Large Vision-Language Models for Astronomical Imaging
Paper • 2407.07315 • Published • 6 -
SHERL: Synthesizing High Accuracy and Efficient Memory for Resource-Limited Transfer Learning
Paper • 2407.07523 • Published • 5 -
NeedleBench: Can LLMs Do Retrieval and Reasoning in 1 Million Context Window?
Paper • 2407.11963 • Published • 44 -
Efficient Training with Denoised Neural Weights
Paper • 2407.11966 • Published • 9 -
From GaLore to WeLore: How Low-Rank Weights Non-uniformly Emerge from Low-Rank Gradients
Paper • 2407.11239 • Published • 8 -
EfficientQAT: Efficient Quantization-Aware Training for Large Language Models
Paper • 2407.11062 • Published • 8 -
NNsight and NDIF: Democratizing Access to Foundation Model Internals
Paper • 2407.14561 • Published • 34 -
LongVideoBench: A Benchmark for Long-context Interleaved Video-Language Understanding
Paper • 2407.15754 • Published • 20 -
KAN or MLP: A Fairer Comparison
Paper • 2407.16674 • Published • 42 -
SIGMA: Sinkhorn-Guided Masked Video Modeling
Paper • 2407.15447 • Published • 8 -
PERSONA: A Reproducible Testbed for Pluralistic Alignment
Paper • 2407.17387 • Published • 19 -
Longhorn: State Space Models are Amortized Online Learners
Paper • 2407.14207 • Published • 18 -
VSSD: Vision Mamba with Non-Casual State Space Duality
Paper • 2407.18559 • Published • 19 -
Diffusion Feedback Helps CLIP See Better
Paper • 2407.20171 • Published • 36 -
Finch: Prompt-guided Key-Value Cache Compression
Paper • 2408.00167 • Published • 14 -
POA: Pre-training Once for Models of All Sizes
Paper • 2408.01031 • Published • 27 -
The Impact of Hyperparameters on Large Language Model Inference Performance: An Evaluation of vLLM and HuggingFace Pipelines
Paper • 2408.01050 • Published • 8 -
UniBench: Visual Reasoning Requires Rethinking Vision-Language Beyond Scaling
Paper • 2408.04810 • Published • 23 -
VisualAgentBench: Towards Large Multimodal Models as Visual Foundation Agents
Paper • 2408.06327 • Published • 16 -
Heavy Labels Out! Dataset Distillation with Label Space Lightening
Paper • 2408.08201 • Published • 19 -
Towards flexible perception with visual memory
Paper • 2408.08172 • Published • 21 -
MME-RealWorld: Could Your Multimodal LLM Challenge High-Resolution Real-World Scenarios that are Difficult for Humans?
Paper • 2408.13257 • Published • 26 -
K-Sort Arena: Efficient and Reliable Benchmarking for Generative Models via K-wise Human Preferences
Paper • 2408.14468 • Published • 36 -
Dolphin: Long Context as a New Modality for Energy-Efficient On-Device Language Models
Paper • 2408.15518 • Published • 43 -
SaRA: High-Efficient Diffusion Model Fine-tuning with Progressive Sparse Low-Rank Adaptation
Paper • 2409.06633 • Published • 15 -
RetrievalAttention: Accelerating Long-Context LLM Inference via Vector Retrieval
Paper • 2409.10516 • Published • 41 -
InfiMM-WebMath-40B: Advancing Multimodal Pre-Training for Enhanced Mathematical Reasoning
Paper • 2409.12568 • Published • 48 -
Oryx MLLM: On-Demand Spatial-Temporal Understanding at Arbitrary Resolution
Paper • 2409.12961 • Published • 25 -
Prithvi WxC: Foundation Model for Weather and Climate
Paper • 2409.13598 • Published • 40 -
Making Text Embedders Few-Shot Learners
Paper • 2409.15700 • Published • 30 -
Time-MoE: Billion-Scale Time Series Foundation Models with Mixture of Experts
Paper • 2409.16040 • Published • 13 -
VPTQ: Extreme Low-bit Vector Post-Training Quantization for Large Language Models
Paper • 2409.17066 • Published • 28 -
Differential Transformer
Paper • 2410.05258 • Published • 169 -
Addition is All You Need for Energy-efficient Language Models
Paper • 2410.00907 • Published • 145 -
Selective Attention Improves Transformer
Paper • 2410.02703 • Published • 24 -
A Comprehensive Survey of Mamba Architectures for Medical Image Analysis: Classification, Segmentation, Restoration and Beyond
Paper • 2410.02362 • Published • 18 -
AuroraCap: Efficient, Performant Video Detailed Captioning and a New Benchmark
Paper • 2410.03051 • Published • 5 -
MLP-KAN: Unifying Deep Representation and Function Learning
Paper • 2410.03027 • Published • 29 -
Towards World Simulator: Crafting Physical Commonsense-Based Benchmark for Video Generation
Paper • 2410.05363 • Published • 45 -
Unveiling the Backbone-Optimizer Coupling Bias in Visual Representation Learning
Paper • 2410.06373 • Published • 36 -
One Initialization to Rule them All: Fine-tuning via Explained Variance Adaptation
Paper • 2410.07170 • Published • 15 -
SageAttention: Accurate 8-Bit Attention for Plug-and-play Inference Acceleration
Paper • 2410.02367 • Published • 47 -
Contrastive Localized Language-Image Pre-Training
Paper • 2410.02746 • Published • 33 -
CLIP-MoE: Towards Building Mixture of Experts for CLIP with Diversified Multiplet Upcycling
Paper • 2409.19291 • Published • 19 -
PrefixQuant: Static Quantization Beats Dynamic through Prefixed Outliers in LLMs
Paper • 2410.05265 • Published • 30 -
LOKI: A Comprehensive Synthetic Data Detection Benchmark using Large Multimodal Models
Paper • 2410.09732 • Published • 54 -
Your Mixture-of-Experts LLM Is Secretly an Embedding Model For Free
Paper • 2410.10814 • Published • 49 -
MixEval-X: Any-to-Any Evaluations from Real-World Data Mixtures
Paper • 2410.13754 • Published • 75 -
Harnessing Webpage UIs for Text-Rich Visual Understanding
Paper • 2410.13824 • Published • 30 -
FlatQuant: Flatness Matters for LLM Quantization
Paper • 2410.09426 • Published • 13 -
LoLDU: Low-Rank Adaptation via Lower-Diag-Upper Decomposition for Parameter-Efficient Fine-Tuning
Paper • 2410.13618 • Published • 6 -
SeerAttention: Learning Intrinsic Sparse Attention in Your LLMs
Paper • 2410.13276 • Published • 26 -
AutoTrain: No-code training for state-of-the-art models
Paper • 2410.15735 • Published • 59 -
PyramidDrop: Accelerating Your Large Vision-Language Models via Pyramid Visual Redundancy Reduction
Paper • 2410.17247 • Published • 45 -
Breaking the Memory Barrier: Near Infinite Batch Size Scaling for Contrastive Loss
Paper • 2410.17243 • Published • 89 -
MMAU: A Massive Multi-Task Audio Understanding and Reasoning Benchmark
Paper • 2410.19168 • Published • 19 -
COAT: Compressing Optimizer states and Activation for Memory-Efficient FP8 Training
Paper • 2410.19313 • Published • 19 -
LARP: Tokenizing Videos with a Learned Autoregressive Generative Prior
Paper • 2410.21264 • Published • 9 -
Dialog2Flow: Pre-training Soft-Contrastive Action-Driven Sentence Embeddings for Automatic Dialog Flow Extraction
Paper • 2410.18481 • Published • 5 -
Relaxed Recursive Transformers: Effective Parameter Sharing with Layer-wise LoRA
Paper • 2410.20672 • Published • 6 -
CLEAR: Character Unlearning in Textual and Visual Modalities
Paper • 2410.18057 • Published • 200 -
ShadowKV: KV Cache in Shadows for High-Throughput Long-Context LLM Inference
Paper • 2410.21465 • Published • 11 -
Accelerating Direct Preference Optimization with Prefix Sharing
Paper • 2410.20305 • Published • 6 -
ReferEverything: Towards Segmenting Everything We Can Speak of in Videos
Paper • 2410.23287 • Published • 19 -
A Pointer Network-based Approach for Joint Extraction and Detection of Multi-Label Multi-Class Intents
Paper • 2410.22476 • Published • 25 -
NeuZip: Memory-Efficient Training and Inference with Dynamic Compression of Neural Networks
Paper • 2410.20650 • Published • 16 -
M2rc-Eval: Massively Multilingual Repository-level Code Completion Evaluation
Paper • 2410.21157 • Published • 6 -
"Give Me BF16 or Give Me Death"? Accuracy-Performance Trade-Offs in LLM Quantization
Paper • 2411.02355 • Published • 47 -
How Far is Video Generation from World Model: A Physical Law Perspective
Paper • 2411.02385 • Published • 33 -
Sparsing Law: Towards Large Language Models with Greater Activation Sparsity
Paper • 2411.02335 • Published • 11 -
Controlling Language and Diffusion Models by Transporting Activations
Paper • 2410.23054 • Published • 16 -
LLM2CLIP: Powerful Language Model Unlock Richer Visual Representation
Paper • 2411.04997 • Published • 37 -
Can sparse autoencoders be used to decompose and interpret steering vectors?
Paper • 2411.08790 • Published • 8 -
Cut Your Losses in Large-Vocabulary Language Models
Paper • 2411.09009 • Published • 43 -
Search, Verify and Feedback: Towards Next Generation Post-training Paradigm of Foundation Models via Verifier Engineering
Paper • 2411.11504 • Published • 20 -
SmoothCache: A Universal Inference Acceleration Technique for Diffusion Transformers
Paper • 2411.10510 • Published • 8 -
SageAttention2 Technical Report: Accurate 4 Bit Attention for Plug-and-play Inference Acceleration
Paper • 2411.10958 • Published • 52 -
When Precision Meets Position: BFloat16 Breaks Down RoPE in Long-Context Training
Paper • 2411.13476 • Published • 15 -
Multimodal Autoregressive Pre-training of Large Vision Encoders
Paper • 2411.14402 • Published • 43 -
Marco-o1: Towards Open Reasoning Models for Open-Ended Solutions
Paper • 2411.14405 • Published • 58 -
Natural Language Reinforcement Learning
Paper • 2411.14251 • Published • 28 -
Ultra-Sparse Memory Network
Paper • 2411.12364 • Published • 19 -
Factorized Visual Tokenization and Generation
Paper • 2411.16681 • Published • 17 -
Cautious Optimizers: Improving Training with One Line of Code
Paper • 2411.16085 • Published • 15 -
Star Attention: Efficient LLM Inference over Long Sequences
Paper • 2411.17116 • Published • 49 -
Training Noise Token Pruning
Paper • 2411.18092 • Published • 1 -
GATE OpenING: A Comprehensive Benchmark for Judging Open-ended Interleaved Image-Text Generation
Paper • 2411.18499 • Published • 18 -
The Well: a Large-Scale Collection of Diverse Physics Simulations for Machine Learning
Paper • 2412.00568 • Published • 14 -
WF-VAE: Enhancing Video VAE by Wavelet-Driven Energy Flow for Latent Video Diffusion Model
Paper • 2411.17459 • Published • 10 -
KV Shifting Attention Enhances Language Modeling
Paper • 2411.19574 • Published • 8 -
APOLLO: SGD-like Memory, AdamW-level Performance
Paper • 2412.05270 • Published • 38 -
OmniDocBench: Benchmarking Diverse PDF Document Parsing with Comprehensive Annotations
Paper • 2412.07626 • Published • 22 -
ACDiT: Interpolating Autoregressive Conditional Modeling and Diffusion Transformer
Paper • 2412.07720 • Published • 30 -
KaSA: Knowledge-Aware Singular-Value Adaptation of Large Language Models
Paper • 2412.06071 • Published • 7 -
Efficient Generative Modeling with Residual Vector Quantization-Based Tokens
Paper • 2412.10208 • Published • 19 -
Byte Latent Transformer: Patches Scale Better Than Tokens
Paper • 2412.09871 • Published • 88 -
No More Adam: Learning Rate Scaling at Initialization is All You Need
Paper • 2412.11768 • Published • 41 -
Distilled Decoding 1: One-step Sampling of Image Auto-regressive Models with Flow Matching
Paper • 2412.17153 • Published • 34 -
Fourier Position Embedding: Enhancing Attention's Periodic Extension for Length Generalization
Paper • 2412.17739 • Published • 39 -
Tensor Product Attention Is All You Need
Paper • 2501.06425 • Published • 45