Ryan Sun

2 accepted papers

2024

SpecHub: Provable Acceleration to Multi-Draft Speculative Decoding

Ryan Sun, Tianyi Zhou, Xun Chen, Lichao Sun

EMNLP 2024main

Large Language Models (LLMs) have become essential in advancing natural language processing (NLP) tasks, but their sequential token generation limits inference speed. Multi-Draft Speculative Decoding (MDSD) offers a promising solution by using a smaller draft model to generate multiple token sequenc…

Cited by 2 Source ↗PDF ↗code ↗Scholar ↗

2024

Virtual Context Enhancing Jailbreak Attacks with Special Token Injection

Yuqi Zhou, Lin Lu, Ryan Sun, Pan Zhou, Lichao Sun

EMNLP 2024finding

Jailbreak attacks on large language models (LLMs) involve inducing these models to generate harmful content that violates ethics or laws, posing a significant threat to LLM security. Current jailbreak attacks face two main challenges: low success rates due to defensive measures and high resource req…

Cited by 8 Source ↗PDF ↗Scholar ↗