研究人员开发了改进语言模型适应性的新方法。一篇论文介绍了一种“迁移图”，用于预测源任务如何影响目标任务，表明帮助性可能是非对称的，并且仔细的任务选择可以显著提高 Qwen3 和 Mistral 等模型的性能。另一种方法 ReCAP 专注于多模态持续指令调优，通过使用检索引导框架来利用外部知识进行能力重用，旨在增强新技能的获取同时保留 LLM 中的现有知识。 AI
影响 这些研究为更有效和高效的语言模型训练和适应提供了新技术，有望在专业任务和多模态应用中提高性能。
排序理由 该集群包含两篇学术论文，详细介绍了语言模型适应和调优的新颖方法。
在 Hugging Face Daily Papers 阅读 →
AI 生成摘要 · Google Gemini · 来自 3 个来源。 我们如何撰写摘要 →
完整方法见我们的编辑标准。
arXiv:2609.39702v1 Announce Type: new Abstract: Adapting a language model to a specialized corpus means choosing which instruction-tuning tasks to train on under a fixed budget, and testing one choice costs a fine-tuning run. Common heuristics add more source tasks or pick source…
arXiv:2609.37889v1 Announce Type: cross Abstract: Multimodal continual instruction tuning (MCIT) aims to enable multimodal large language models to acquire new capabilities from sequential tasks while preserving previously learned knowledge. Existing methods primarily mitigate ca…
Multimodal continual instruction tuning (MCIT) aims to enable multimodal large language models to acquire new capabilities from sequential tasks while preserving previously learned knowledge. Existing methods primarily mitigate catastrophic forgetting by constraining parameter up…