Benchmarking AnLLMs: Insights from OpenBookQA to BoolQ

10 Oct 2024

Discover how AnLLMs perform across various benchmarks, including OpenBookQA and BoolQ.

Pre-Training AnLLMs: Leveraging RedPajama Data for Enhanced Performance

10 Oct 2024

Discover the training procedure for AnLLMs using the RedPajama-Data-1TSample dataset, featuring 850,000 samples and 1 billion tokens.

Anchor-based Large Language Models: Experiments and Implementation

10 Oct 2024

Explore the implementation of AnLLM, featuring two strategies for anchor token selection: using punctuation marks and introducing a new token.

Improving Real-Time Inference with Anchor Tokens

10 Oct 2024

Discover how anchor tokens can optimize the inference process in LLMs by enhancing the keys/values caching mechanism.

The Role of Anchor Tokens in Self-Attention Networks

10 Oct 2024

Introducing anchor-based attention masks that optimize self-attention in LLMs by controlling how tokens access information across sequences.

Unlocking the Mechanics of Decoder-Only Transformers and Self-Attention

10 Oct 2024

Explore the inner workings of decoder-only transformers in LLMs, focusing on self-attention networks and the challenges of keys/values caching during inference.

How Anchor Tokens Transform Sequence Information Compression in LLMs

10 Oct 2024

Discover how our research extends in-context learning by introducing anchor tokens for sequence compression in LLMs.

Anchor-based Large Language Models

10 Oct 2024

Learn how anchor-based LLMs compress sequence information to reduce memory demand and boost inference speed.