← Volver a los artículos

LLMS GET LOST IN MULTI-TURN CONVERSATION

★ ★ ★ ★ ☆

Resumen del artículo

Título de Paperzilla
LLMs: Great at One-Liners, Lost in Conversation

Large Language Models (LLMs) exhibit significantly lower performance in multi-turn conversations compared to single-turn interactions, primarily due to a substantial increase in unreliability rather than a loss in aptitude. This "lost in conversation" phenomenon stems from LLMs making early assumptions, prematurely proposing solutions, and struggling to incorporate new information effectively. The study employed simulated conversations across six diverse generation tasks, revealing consistent performance degradation across various LLMs, regardless of size or reasoning capabilities.

Explícamelo como si tuviera cinco años

Scientists found that smart computer friends, even super smart ones, sometimes get lost when you talk back and forth with them a lot. It's like they guess the answer too quickly or forget what you said before.

Posibles conflictos de intereses

Authors are employed by Microsoft Research and Salesforce Research, organizations with vested interest in LLM development and performance.

Limitaciones identificadas

Reliance on Simulated Conversations
The reliance on simulated conversations, while enabling scalability, limits the generalizability of findings to real-world human-AI interactions, as the simulated conversations lack the nuances and complexities of natural human communication.
Focus on Analytical Tasks
The focus on analytical tasks restricts the scope of the findings, as it is unclear whether similar performance degradation occurs in open-ended or creative tasks.
Focus on English Language, Text-Only Tasks
The concentration on English-language, text-only tasks limits the applicability of the findings to other languages and modalities, such as speech or images.

Explicación de la calificación

This paper presents a comprehensive and large-scale study highlighting a critical issue in current LLM performance: unreliability in multi-turn conversations. The methodology is sound, and the findings are significant and potentially impactful for future LLM development. The limitations regarding simulated conversations and task scope are acknowledged.

Conviene saber

Este es el análisis de Starter. Paperzilla Pro verifica cada cita, investiga los antecedentes de los autores y las fuentes de financiación, y utiliza razonamiento avanzado con IA para ofrecer información más exhaustiva.

Explorar Pro →

Jerarquía temática

Información del archivo

Título original: LLMS GET LOST IN MULTI-TURN CONVERSATION
Subido: 8 jul 2025, 12:06:18
Privacidad: Público