From 3fb2a778d58457a78290a9c38ce754765ea583ba Mon Sep 17 00:00:00 2001
From: bojunc <316231662@qq.com>
Date: Mon, 2 Mar 2026 23:01:11 +0800
Subject: [PATCH] =?UTF-8?q?=F0=9F=93=84=20=E6=B7=BB=E5=8A=A0=202026-03-02?=
 =?UTF-8?q?=20AI=20=E6=96=B0=E9=97=BB=E7=AE=80=E6=8A=A5=20(=E4=B8=AD?=
 =?UTF-8?q?=E8=8B=B1=E6=96=87)?=
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

---
 daily/2026-03-02_en.md | 41 +++++++++++++++++++++++++++++++++++++++++
 daily/2026-03-02_zh.md | 41 +++++++++++++++++++++++++++++++++++++++++
 2 files changed, 82 insertions(+)
 create mode 100644 daily/2026-03-02_en.md
 create mode 100644 daily/2026-03-02_zh.md

diff --git a/daily/2026-03-02_en.md b/daily/2026-03-02_en.md
new file mode 100644
index 0000000..bcf55fd
--- /dev/null
+++ b/daily/2026-03-02_en.md
@@ -0,0 +1,41 @@
+# AI Daily Brief - 2026-03-02
+
+> Collected at: 3/2/2026, 11:00:55 PM
+> Total items: 20
+
+## 🔥 Top 10 Highlights
+
+1. LK Losses: Direct Acceptance Rate Optimization for Speculative Decoding
+   https://huggingface.co/papers/2602.23881
+
+2. CUDA Agent: Large-Scale Agentic RL for High-Performance CUDA Kernel Generation
+   https://huggingface.co/papers/2602.24286
+
+3. SenCache: Accelerating Diffusion Model Inference via Sensitivity-Aware Caching
+   https://huggingface.co/papers/2602.24208
+
+4. How to Take a Memorable Picture? Empowering Users with Actionable Feedback
+   https://huggingface.co/papers/2602.21877
+
+5. CL4SE: A Context Learning Benchmark For Software Engineering Tasks
+   https://huggingface.co/papers/2602.23047
+
+6. Compositional Generalization Requires Linear, Orthogonal Representations in Vision Embedding Models
+   https://huggingface.co/papers/2602.24264
+
+7. LongVideo-R1: Smart Navigation for Low-cost Long Video Understanding
+   https://huggingface.co/papers/2602.20913
+
+8. InfoNCE Induces Gaussian Distribution
+   https://huggingface.co/papers/2602.24012
+
+9. Recovered in Translation: Efficient Pipeline for Automated Translation of Benchmarks and Datasets
+   https://huggingface.co/papers/2602.22207
+
+10. Mode Seeking meets Mean Seeking for Fast Long Video Generation
+   https://huggingface.co/papers/2602.24289
+
+## 📂 Categories
+
+---
+*Generated by AINewsCollector*
diff --git a/daily/2026-03-02_zh.md b/daily/2026-03-02_zh.md
new file mode 100644
index 0000000..e7e0de9
--- /dev/null
+++ b/daily/2026-03-02_zh.md
@@ -0,0 +1,41 @@
+# AI Daily Brief - 2026-03-02
+
+> 采集时间: 2026/3/2 23:00:41
+> 总条目: 20
+
+## 🔥 Top 10 重要消息
+
+1. LK Losses：预测解码直接接受率优化
+   https://huggingface.co/papers/2602.23881
+
+2. CUDA Agent：大规模代理强化学习助力高性能CUDA内核生成
+   https://huggingface.co/papers/2602.24286
+
+3. SenCache：基于敏感性感知的缓存加速扩散模型推理
+   https://huggingface.co/papers/2602.24208
+
+4. 图像记忆度提升：赋能用户可操作反馈技术
+   https://huggingface.co/papers/2602.21877
+
+5. CL4SE：软件工程任务中的上下文学习基准
+   https://huggingface.co/papers/2602.23047
+
+6. 视觉嵌入模型需线性正交表示以实现组合泛化
+   https://huggingface.co/papers/2602.24264
+
+7. LongVideo-R1：低成本长视频智能导航理解
+   https://huggingface.co/papers/2602.20913
+
+8. InfoNCE诱导高斯分布对比学习
+   https://huggingface.co/papers/2602.24012
+
+9. 高效基准与数据集自动翻译，提升多语言LLM评估可靠性
+   https://huggingface.co/papers/2602.22207
+
+10. 视频生成新突破：模式寻优融合均值寻优，快速生成长视频
+   https://huggingface.co/papers/2602.24289
+
+## 📂 分类汇总
+
+---
+*Generated by AINewsCollector*