将LoRA重新定义为知识记忆的实证研究，揭示其存储与检索机制，为微调提供新视角。
arXiv:2603.01097v3 Announce Type: replace Abstract: Continuous knowledge updating for pre-trained large language models (LLMs) is increasingly necessa…
专为大模型设备端微调设计的内存高效结构化反向传播，突破移动设备6-12GB内存限制，实现隐私保护与性能平衡。
arXiv:2602.13069v2 Announce Type: replace Abstract: On-device fine-tuning enables privacy-preserving personalization of large language models, but mob…
提出可学习的零阶优化器，无需梯度即可高效微调大模型，大幅降低内存开销。
arXiv:2510.00419v2 Announce Type: replace Abstract: Zeroth-order optimizers have recently emerged as an attractive approach for fine-tuning large lang…
最新研究揭示数据选择策略如何影响大模型微调的长期表现与鲁棒性。
arXiv:2605.30537v1 Announce Type: new Abstract: Data selection is increasingly used to reduce the cost of large language model (LLM) fine-tuning, with…
用语言模型引导贝叶斯优化，高效搜索LoRA超参数，ICML 2026论文带来微调新思路。
arXiv:2602.11171v2 Announce Type: replace-cross Abstract: Fine-tuning Large Language Models (LLMs) with Low-Rank Adaptation (LoRA) offers a resource-e…
动态调整数据与模型匹配度，让知识蒸馏更贴合学生能力，提升推理模型微调效率。
arXiv:2605.29229v1 Announce Type: new Abstract: Reasoning distillation transfers complex reasoning abilities from large language models (LLMs) to smal…
揭秘LoRA微调大模型时如何“记住”信息，提出参数化记忆定律，深入理论分析注意力机制。
arXiv:2605.30260v1 Announce Type: cross Abstract: Large Language Models (LLMs) must continuously learn and update knowledge to remain effective in dyn…
提出On-Policy Replay方法，解决持续监督微调中的灾难性遗忘问题，为LLM高效增量学习提供新思路。
arXiv:2605.29495v1 Announce Type: new Abstract: Continual supervised fine-tuning (SFT) is the de facto recipe for adapting large language models (LLMs…
零阶微调本质是推理负载，颠覆大模型微调的计算认知
arXiv:2605.28760v1 Announce Type: new Abstract: Zeroth-order (ZO) fine-tuning is attractive for large language models because it replaces backpropagat…
联邦学习遇上LoRA新框架，专治大模型在异构设备上协作微调的痛点，方案灵活高效。
arXiv:2501.19389v4 Announce Type: replace Abstract: Fine-tuning large language models (LLMs) on resource-constrained clients remains a challenging pro…
创新激活指导的零阶优化方法，大幅提升大模型微调效率。
arXiv:2601.17261v4 Announce Type: replace Abstract: Zeroth-Order (ZO) optimization has emerged as a promising solution for fine-tuning LLMs under stri…
PyTorch官方推出的后训练微调库torchtune，原生集成LoRA、QLoRA等高效技术，简化大模型适配流程。
arXiv:2605.21442v1 Announce Type: new Abstract: Modern LLMs typically require multistage training pipelines to achieve strong downstream performance, …
通过调整学习率，简单LoRA即可媲美复杂微调方法，揭示被忽视的关键因素。
arXiv:2602.04998v2 Announce Type: replace Abstract: Low-Rank Adaptation (LoRA) is the prevailing approach for efficient large language model (LLM) fin…
一种新的后训练微调技术DFT，专门修复大语言模型的写作水平，值得关注
Article URL: https://twitter.com/rosmine/status/2056406211369541947 Comments URL: https://news.ycombinator.com/item?id=48184164 Points: 1 # Comments: …