研究人员开发了一种名为潜在记忆的新方法，以改进面向资源受限环境的问答系统。该方法将文本和图像等多模态证据压缩成单个潜在令牌。通过在统一的潜在空间中运行，潜在记忆显著减少了令牌消耗，与传统的基于检索的系统相比，使用的令牌数量减少了 3 倍至 10 倍，同时在各种问答基准测试中保持了有竞争力的性能。 AI
影响 减少了问答系统中的令牌消耗，使先进的多模态 AI 在资源受限的应用中更易于访问。
排序理由 该集群包含一篇详细介绍多模态问答新方法的论文。
在 Hugging Face Daily Papers 阅读 →
AI 生成摘要 · Google Gemini · 来自 4 个来源。 我们如何撰写摘要 →
arXiv:2606.10572v1 Announce Type: new Abstract: External memory effectively grounds large language models (LLMs) and vision-language models (VLMs)-based question answering (QA) in relevant multimodal evidence. However, existing memory paradigms represent each memory item in raw t…
External memory effectively grounds large language models (LLMs) and vision-language models (VLMs)-based question answering (QA) in relevant multimodal evidence. However, existing memory paradigms represent each memory item in raw text and image forms, so retrieval-based systems …
Latent Memory introduces a compressed representation approach for external memory in question answering, reducing token consumption and storage requirements while maintaining competitive performance across text-only and multimodal benchmarks.
External memory effectively grounds large language models (LLMs) and vision-language models (VLMs)-based question answering (QA) in relevant multimodal evidence. However, existing memory paradigms represent each memory item in raw text and image forms, so retrieval-based systems …