Star ...

The Next Recipe

Scaling Laws

The Second Curve of Scaling Law

Scaling Factors
BitNet

BitNet a4.8: 4-bit Activations for 1-bit LLMs
BitNet

Differential Transformer
BitNet

The Era of 1-bit LLMs: All Large Language Models are in 1.58 Bits
Fully Sparsely-Activated LLMs

1-bit AI Infra / bitnet.cpp: Running LLMs on CPUs
Fully Sparsely-Activated LLMs

Q-Sparse / Block Q-Sparse: Fully Sparsely-Activated LLMs
YOCO

You Only Cache Once: Decoder-Decoder Architectures for Large Language Models
// Gated RetNet (RetNet-3)
BitNet

The Era of 1-bit LLMs: Training Tips, Code and FAQ
MoE

MELLE: Autoregressive Speech Synthesis without Vector Quantization
MoE

VALL-E 2: Human Parity Zero-Shot Text to Speech Synthesis
MoE

Multi-Head Mixture-of-Experts
Learning Law

The Learning Law: Towards Optimal Learning of Language Models

The Mind's Eye of (M)LLMs: Visualization-of-Thought Elicits Spatial Reasoning in Large Language Models

Synthetic Data (Almost) from Scratch: Generalized Instruction Tuning for Language Models

Multilingual E5 Text Embeddings

BitNet: 1-bit Transformers and LLMs
RetNet

Retentive Network: Revolutionizing Transformers for Large Language Models
LongViT

LongViT (LongNet for Vision): When an Image is Worth 1,024 × 1,024 Words
Kosmos

Kosmos-G: Generating Images in Context with Multimodal Large Language Models
Kosmos 2.5

Kosmos-2.5: A Multimodal Literate Model
LLM4Science

Large Language Model for Science: A Study on P vs. NP
LongNet

LongNet: Scaling Transformers to 1,000,000,000 Tokens
KOSMOS-2

Kosmos-2: Grounding Multimodal Large Language Models (MLLMs) to the World
KOSMOS-1

Kosmos-1: A Multimodal Large Language Model (MLLM)
VALL-E

WavMark: Watermarking for Audio Generation
VALL-E

VALL-E (X): Neural Codec Language Models are Zero-Shot Text to Speech Synthesizers
LLMA

PoSE: Efficient Context Window Extension of LLMs via Positional Skip-wise Training
LLMA

AdaLLM: Adapting Large Language Models via Reading Comprehension
MiniLLM

MiniLLM: Knowledge Distillation of Large Language Models
LongMem

Large Language Models with Long-Term Memory
LLMA

LLM Accelerator: Lossless Acceleration of Large Language Models
XPos

A Length-Extrapolatable Transformer
ICL

Why Can GPT Learn In-Context? Language Models Secretly Perform Gradient Descent as Meta Optimizers
Promptist

Promptist: Optimizing Prompts for Text-to-Image Generation
Structured Prompting

Structured Prompting: Scaling In-Context Learning to 1,000 Examples

TorchScale

TorchScale: Transformers at (Any) Scale

Magneto: A Foundation Transformer
BEiT-3

BEiT-3: A General-Purpose Multimodal Foundation Model

Language Models are General-Purpose Interfaces

DeepNet: Scaling Transformers to 1,000 Layers

BEiT: BERT Pre-Training of Image Transformers

MiniLM: Multi-Head Self-Attention Relation Distillation for Compressing Pretrained Transformers

XLM-E: Efficient Multilingual Language Model Pre-training

UniLM: Unified Language Model Pre-training