牛哥精选 · 所有

1

🤖 AI·大模型 arXiv 机器学习 2026-05-25

AGZO: Activation-Guided Zeroth-Order Optimization for LLM Fine-Tuning

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…

agzo 大模型微调零阶优化激活指导效率提升

2

📝 深度技术 arXiv AI 2026-05-25

GILT: An LLM-Free, Tuning-Free Graph Foundational Model for In-Context Learning

无需大模型和微调，图基础模型GILT通过上下文学习实现高效推理，为图领域开辟新范式。

arXiv:2510.04567v2 Announce Type: replace-cross Abstract: Graph Neural Networks (GNNs) are powerful tools for processing relational data but often str…

图基础模型上下文学习无微调 llm-free icml 2026

3

📝 深度技术 arXiv 机器学习 2026-05-25

Convex Optimization for Alignment and Preference Learning on a Single GPU

单GPU实现凸优化方法，高效解决LLM偏好对齐难题，降低RLHF计算成本。

arXiv:2605.23244v1 Announce Type: new Abstract: Fine-tuning large language models (LLMs) to align with human preferences has driven the success of sys…

凸优化偏好学习单gpu llm对齐 direct pre

4

📝 深度技术 arXiv AI 2026-05-25

PoisonForge: Task-Level Targeted Poisoning Benchmark for Instruction-Tuned LLMs

指令微调LLM面临的任务级定向投毒威胁，首个系统性基准PoisonForge发布，助力模型安全评估。

arXiv:2605.23168v1 Announce Type: cross Abstract: When practitioners fine-tune LLMs on unvetted datasets, an adversary can exploit the data supply cha…

poisonforg 指令微调 llm安全任务级投毒基准测试

5

🤖 AI·大模型 arXiv 机器学习 2026-05-23

Provably Protecting Fine-Tuned LLMs from Training Data Extraction while Preserving Utility

新研究用可证明方式保护微调大模型免遭训练数据窃取，同时维持模型效能，隐私与实用兼得。

arXiv:2602.00688v2 Announce Type: replace Abstract: Fine-tuning large language models (LLMs) on sensitive datasets raises privacy concerns, as trainin…

大语言模型训练数据提取隐私保护可证明防御微调

6

📝 深度技术 arXiv 机器学习 2026-05-23

On Robustness and Chain-of-Thought Consistency of RL-Finetuned VLMs

首份系统研究RL微调VLM的鲁棒性与思维链一致性，揭示模型脆弱性根源

arXiv:2602.12506v3 Announce Type: replace Abstract: Reinforcement learning (RL) finetuning has become a key technique for enhancing large language mod…

rl微调视觉语言模型鲁棒性思维链一致性模型脆弱性

7

🤖 AI·大模型 arXiv 机器学习 2026-05-21

torchtune: PyTorch native post-training library

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, …

torchtune pytorch 后训练大模型微调库

8

📝 深度技术 arXiv 机器学习 2026-05-21

FT-Dojo: Towards Autonomous LLM Fine-Tuning with Language Agents

用语言代理实现大模型自主微调的创新框架，省去人工干预，让微调过程自动化

arXiv:2603.01712v2 Announce Type: replace-cross Abstract: Fine-tuning large language models for vertical domains remains labor-intensive, requiring pr…

llm微调语言代理自动化学术论文大模型定制

9

📝 深度技术 arXiv 机器学习 2026-05-21

Complementing reinforcement learning with SFT through logit averaging in the post training of LLMs

新方法通过logit averaging融合强化学习与监督微调，显著提升LLM后训练的性能和稳定性。

arXiv:2605.20555v1 Announce Type: new Abstract: We introduce a novel method that averages the logits of a frozen reference policy (e.g., SFT) and a tr…

大语言模型强化学习监督微调 logit aver 后训练

10

📝 深度技术 arXiv 机器学习 2026-05-21

Task-conditioned probing of instruction-tuned multimodal LLMs: Region-specific brain alignment patterns under naturalistic stimuli

探索指令微调多模态大模型在自然刺激下的脑区对齐模式，交叉验证AI与神经科学

arXiv:2506.08277v3 Announce Type: replace-cross Abstract: Recent voxel-wise multimodal brain encoding studies have shown that multimodal large languag…

多模态大模型指令微调脑对齐神经科学自然刺激

11

🤖 AI·大模型 OpenAI 官方博客 2026-05-20

Customizing GPT-3 for your application

OpenAI官方指南：一条命令即可微调GPT-3，快速定制专属AI模型。

Fine-tune with a single command.

微调 gpt-3 openai 定制化命令行

12

📝 深度技术 arXiv 计算机视觉 2026-05-20

StrLoRA: Towards Streaming Continual Visual Instruction Tuning for MLLMs

StrLoRA提出流式持续视觉指令微调新方法，有效缓解多模态大模型在序列任务中的灾难性遗忘。

arXiv:2605.16353v1 Announce Type: new Abstract: Continual Visual Instruction Tuning (CVIT) enables Multimodal Large Language Models to incrementally a…

strlora 流式学习持续学习视觉指令微调多模态大模型

13

📝 深度技术 arXiv 机器学习 2026-05-20

SMART Fine-tuning Factor Augmented Neural Lasso

提出SMART框架，将预训练模型融入高维非参数变量选择，为微调提供理论基础。

arXiv:2604.12288v2 Announce Type: replace-cross Abstract: Fine-tuning is a widely used strategy for adapting pre-trained models to new tasks, yet its …

微调非参数高维统计变量选择 smart框架

14

🤖 AI·大模型 arXiv 机器学习 2026-05-20

Hybrid-LoRA: Bridging Full Fine-Tuning and Low-Rank Adaptation for Post-Training

混合全微调与低秩适应的新方法，专为后训练场景优化，效率与性能兼得

arXiv:2605.18822v1 Announce Type: new Abstract: Post-training has become essential for adapting large language models (LLMs) to complex downstream beh…

hybrid-lor 低秩适应全微调后训练模型压缩

15

📝 深度技术 arXiv 机器学习 2026-05-20

Learning Rate Matters: Vanilla LoRA May Suffice for LLM Fine-tuning

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…

lora 学习率大模型微调参数高效微调深度学习优化

16

🤖 AI·大模型 arXiv 机器学习 2026-05-20

Fine-tuning Large Language Model for Automated Algorithm Design

微调大模型实现算法自动设计，探索AI驱动的研究新范式

arXiv:2507.10614v2 Announce Type: replace Abstract: The integration of large language models (LLMs) into automated algorithm design has shown promisin…

大语言模型微调算法自动设计自动化 arxiv论文

17

📝 深度技术 arXiv NLP 2026-05-20

Ancient Greek to Modern Greek Machine Translation: A Novel Benchmark and Fine-Tuning Experiments on LLMs and NMT Models

面向古籍翻译难题，提出古希腊语到现代希腊语的专用基准，对比LLM与NMT模型的微调效果。

arXiv:2605.18504v1 Announce Type: new Abstract: Machine Translation (MT) for Ancient Greek (AG) to Modern Greek (MG) is a low-resource task, constrain…

古希腊语现代希腊语机器翻译基准 llm

18

📝 深度技术 arXiv NLP 2026-05-20

From BERT to T5: A Study of Named Entity Recognition

从BERT到T5，一篇扎实的NER微调实战对比，技术细节丰富。

arXiv:2605.18462v1 Announce Type: new Abstract: Named entity recognition (NER) has been one of the essential preliminary steps in modern NLP applicati…

ner bert t5 预训练模型微调

19

📝 深度技术 arXiv 机器学习 2026-05-20

Discrete Tilt Matching

介绍一种针对掩码扩散大语言模型的离散倾斜匹配方法，解决RL微调中边际似然难解问题。

arXiv:2604.18739v2 Announce Type: replace Abstract: Masked diffusion large language models (dLLMs) are a promising alternative to autoregressive gener…

扩散模型大语言模型强化学习微调掩码

20

📝 深度技术 arXiv 机器学习 2026-05-20

FediLoRA: Practical Federated Fine-Tuning of Foundation Models Under Missing-Modality Constraints

提出FediLoRA方法，在联邦微调中解决模态缺失难题，兼顾通信效率与模型性能。

arXiv:2509.06984v3 Announce Type: replace Abstract: Federated Learning with LoRA fine-tuning offers an efficient and privacy-aware solution for instit…

联邦学习 lora 多模态微调基础模型模态缺失

🐂 牛哥精选

AGZO: Activation-Guided Zeroth-Order Optimization for LLM Fine-Tuning

GILT: An LLM-Free, Tuning-Free Graph Foundational Model for In-Context Learning

Convex Optimization for Alignment and Preference Learning on a Single GPU

PoisonForge: Task-Level Targeted Poisoning Benchmark for Instruction-Tuned LLMs

Provably Protecting Fine-Tuned LLMs from Training Data Extraction while Preserving Utility

On Robustness and Chain-of-Thought Consistency of RL-Finetuned VLMs

torchtune: PyTorch native post-training library

FT-Dojo: Towards Autonomous LLM Fine-Tuning with Language Agents

Complementing reinforcement learning with SFT through logit averaging in the post training of LLMs

Task-conditioned probing of instruction-tuned multimodal LLMs: Region-specific brain alignment patterns under naturalistic stimuli

Customizing GPT-3 for your application

StrLoRA: Towards Streaming Continual Visual Instruction Tuning for MLLMs

SMART Fine-tuning Factor Augmented Neural Lasso

Hybrid-LoRA: Bridging Full Fine-Tuning and Low-Rank Adaptation for Post-Training

Learning Rate Matters: Vanilla LoRA May Suffice for LLM Fine-tuning

Fine-tuning Large Language Model for Automated Algorithm Design

Ancient Greek to Modern Greek Machine Translation: A Novel Benchmark and Fine-Tuning Experiments on LLMs and NMT Models

From BERT to T5: A Study of Named Entity Recognition

Discrete Tilt Matching

FediLoRA: Practical Federated Fine-Tuning of Foundation Models Under Missing-Modality Constraints

📅 日期