牛哥精选 · 所有

📋 全部 ☁️ 云服务 🤖 AI 平台 🔗 API 中转 🔐 安全/认证 💳 支付 📧 通讯 📊 数据分析 🖼 媒体处理 🌐 域名/DNS

📝 深度技术 arXiv NLP 2026-06-04

Few Tokens, Big Leverage: Preserving Safety Alignment by Constraining Safety Tokens during Fine-tuning

通过约束微调中的安全token，在极少量参数变动下保住大模型的安全护栏，KDD 2026 新思路。

arXiv:2603.07445v2 Announce Type: replace Abstract: Large language models (LLMs) often require fine-tuning (FT) to perform well on downstream tasks, b…

安全对齐微调大模型安全 token约束 kdd 2026

📝 深度技术 arXiv AI 2026-05-29

Continuity and Ordinality Matter: Constraining Time Series Tokens for Effective Time Series Analysis with Large Language Models

全新方法约束时序token的连续性与序数性，大幅提升LLM处理时间序列任务的性能与理解力。

arXiv:2605.28866v1 Announce Type: cross Abstract: Token-based time series large language models (TS-LLMs) have emerged as a promising direction for ti…

大语言模型时间序列分析 token约束连续性与序数性机器学习

📅 日期

2026-05-20 2026-05-19

🐂 牛哥精选

Few Tokens, Big Leverage: Preserving Safety Alignment by Constraining Safety Tokens during Fine-tuning

Continuity and Ordinality Matter: Constraining Time Series Tokens for Effective Time Series Analysis with Large Language Models

📅 日期