Soaring from 4K to 400K: Extending LLM's Context with Activation Beacon
Paper
•
2401.03462
•
Published
•
27
context
-
-
MEGABYTE: Predicting Million-byte Sequences with Multiscale Transformers
Paper • 2305.07185 • Published • 9 -
YaRN: Efficient Context Window Extension of Large Language Models
Paper • 2309.00071 • Published • 77 -
Infinite-LLM: Efficient LLM Service for Long Context with DistAttention and Distributed KVCache
Paper • 2401.02669 • Published • 16 -
LLM Maybe LongLM: Self-Extend LLM Context Window Without Tuning
Paper • 2401.01325 • Published • 27 -
Extending Context Window of Large Language Models via Semantic Compression
Paper • 2312.09571 • Published • 16 -
Zebra: Extending Context Window with Layerwise Grouped Local-Global Attention
Paper • 2312.08618 • Published • 15 -
Mamba: Linear-Time Sequence Modeling with Selective State Spaces
Paper • 2312.00752 • Published • 146 -
E^2-LLM: Efficient and Extreme Length Extension of Large Language Models
Paper • 2401.06951 • Published • 26 -
Extending LLMs' Context Window with 100 Samples
Paper • 2401.07004 • Published • 16 -
KVQuant: Towards 10 Million Context Length LLM Inference with KV Cache Quantization
Paper • 2401.18079 • Published • 8 -
KIVI: A Tuning-Free Asymmetric 2bit Quantization for KV Cache
Paper • 2402.02750 • Published • 4 -
A Human-Inspired Reading Agent with Gist Memory of Very Long Contexts
Paper • 2402.09727 • Published • 38 -
In Search of Needles in a 10M Haystack: Recurrent Memory Finds What LLMs Miss
Paper • 2402.10790 • Published • 42 -
LongRoPE: Extending LLM Context Window Beyond 2 Million Tokens
Paper • 2402.13753 • Published • 116 -
Data Engineering for Scaling Language Models to 128K Context
Paper • 2402.10171 • Published • 25 -
Striped Attention: Faster Ring Attention for Causal Transformers
Paper • 2311.09431 • Published • 4 -
Ring Attention with Blockwise Transformers for Near-Infinite Context
Paper • 2310.01889 • Published • 13 -
LLoCO: Learning Long Contexts Offline
Paper • 2404.07979 • Published • 22 -
Leave No Context Behind: Efficient Infinite Context Transformers with Infini-attention
Paper • 2404.07143 • Published • 111 -
InfLLM: Unveiling the Intrinsic Capacity of LLMs for Understanding Extremely Long Sequences with Training-Free Memory
Paper • 2402.04617 • Published • 4 -
Megalodon: Efficient LLM Pretraining and Inference with Unlimited Context Length
Paper • 2404.08801 • Published • 66 -
TransformerFAM: Feedback attention is working memory
Paper • 2404.09173 • Published • 43 -
Extending Llama-3's Context Ten-Fold Overnight
Paper • 2404.19553 • Published • 34 -
Make Your LLM Fully Utilize the Context
Paper • 2404.16811 • Published • 55 -
Long-context LLMs Struggle with Long In-context Learning
Paper • 2404.02060 • Published • 37 -
HyperAttention: Long-context Attention in Near-Linear Time
Paper • 2310.05869 • Published • 2 -
World Model on Million-Length Video And Language With RingAttention
Paper • 2402.08268 • Published • 40 -
LongNet: Scaling Transformers to 1,000,000,000 Tokens
Paper • 2307.02486 • Published • 81 -
MInference 1.0: Accelerating Pre-filling for Long-Context LLMs via Dynamic Sparse Attention
Paper • 2407.02490 • Published • 27 -
LongWriter: Unleashing 10,000+ Word Generation from Long Context LLMs
Paper • 2408.07055 • Published • 67