新研究发现，大型语言模型因注意力关闭而失去对话线索

报道来源 [3]

1. arXiv cs.CL TIER_1 English(EN) · Dilek Hakkani-Tür · 2026-05-13 02:58

   When Attention Closes: How LLMs Lose the Thread in Multi-Turn Interaction

   Large language models can follow complex instructions in a single turn, yet over long multi-turn interactions they often lose the thread of instructions, persona, and rules. This degradation has been measured behaviorally but not mechanistically explained. We propose a channel-tr…

2. Mastodon — mastodon.social TIER_1 English(EN) · aihaberleri · 2026-05-14 04:24

   📰 Attention Closure in LLMs: Why Multi-Turn Conversations Lose the Thread (2026 Study) New research reveals that large language models lose track of instruction

   📰 Attention Closure in LLMs: Why Multi-Turn Conversations Lose the Thread (2026 Study) New research reveals that large language models lose track of instructions in long conversations due to a measurable 'attention closure' mechanism. The study introduces the Goal Accessibility R…

   链接 aihaberleri.org/…/attention-closure-in-ll…
3. Mastodon — mastodon.social TIER_1 Türkçe(TR) · aihaberleri · 2026-05-14 04:24

   📰 Distraction in Multi-Turn Conversations: Why is AI Losing the Thread in 2026? Why do AI models lose the thread in long conversations? Mistral AI's

   📰 Çok Aşamalı Sohbetlerde Dikkat Dağınıklığı: YZ 2026’da Neden Konuyu Kaybediyor? Yapay zeka modelleri uzun sohbetlerde neden konuyu kaybediyor? Mistral AI’nın yeni araştırması, dikkat mekanizmasının sınırlarını ve çözüm önerilerini ortaya koyuyor.... # BilimveAraştırma # AI # Te…

   链接 aihaberleri.org/…/cok-asamali-sohbetlerde…