CLAIMar 13, 2024

StreamingDialogue: Prolonged Dialogue Learning via Long Context Compression with Minimal Losses

arXiv:2403.08312v310 citationsh-index: 10NIPS
Originality Incremental advance
AI Analysis

This addresses efficiency and consistency issues in prolonged dialogues for LLM applications, though it is incremental as it builds on existing long-context capabilities.

The paper tackles the problem of LLMs struggling with long dialogue contexts by compressing dialogue history into conversational attention sinks, achieving a 4x speedup and 18x memory reduction compared to dense attention recomputation.

Standard Large Language Models (LLMs) struggle with handling dialogues with long contexts due to efficiency and consistency issues. According to our observation, dialogue contexts are highly structured, and the special token of \textit{End-of-Utterance} (EoU) in dialogues has the potential to aggregate information. We refer to the EoU tokens as ``conversational attention sinks'' (conv-attn sinks). Accordingly, we introduce StreamingDialogue, which compresses long dialogue history into conv-attn sinks with minimal losses, and thus reduces computational complexity quadratically with the number of sinks (i.e., the number of utterances). Current LLMs already demonstrate the ability to handle long context window, e.g., a window size of 200K or more. To this end, by compressing utterances into EoUs, our method has the potential to handle more than 200K of utterances, resulting in a prolonged dialogue learning. In order to minimize information losses from reconstruction after compression, we design two learning strategies of short-memory reconstruction (SMR) and long-memory reactivation (LMR). Our method outperforms strong baselines in dialogue tasks and achieves a 4 $\times$ speedup while reducing memory usage by 18 $\times$ compared to dense attention recomputation.

Code Implementations1 repo
Foundations

The foundational work for this paper's niche, ranked by how specifically the neighbourhood builds on it — not by global fame.

Your Notes