diff --git a/cache/seen_urls.json b/cache/seen_urls.json
index 00f78c3..e6b3646 100644
--- a/cache/seen_urls.json
+++ b/cache/seen_urls.json
@@ -1,5 +1,5 @@
 {
-  "lastUpdate": "2026-03-04T15:01:07.708Z",
+  "lastUpdate": "2026-03-05T15:01:12.455Z",
   "urls": [
     "http://arxiv.org/abs/2602.23360v1",
     "http://arxiv.org/abs/2602.23359v1",
@@ -386,6 +386,95 @@
     "https://github.com/sponsors/fujiwarachoki",
     "https://github.com/login?return_to=%2fitzcrazykns%2fperplexica",
     "https://github.com/login?return_to=%2fnautechsystems%2fnautilus_trader",
-    "https://github.com/sponsors/flowiseai"
+    "https://github.com/sponsors/flowiseai",
+    "http://arxiv.org/abs/2603.04390v1",
+    "http://arxiv.org/abs/2603.04385v1",
+    "http://arxiv.org/abs/2603.04378v1",
+    "http://arxiv.org/abs/2603.04370v1",
+    "http://arxiv.org/abs/2603.04366v1",
+    "http://arxiv.org/abs/2603.04364v1",
+    "http://arxiv.org/abs/2603.04359v1",
+    "http://arxiv.org/abs/2603.04356v1",
+    "http://arxiv.org/abs/2603.04355v1",
+    "http://arxiv.org/abs/2603.04348v1",
+    "http://arxiv.org/abs/2603.04334v1",
+    "http://arxiv.org/abs/2603.04333v1",
+    "http://arxiv.org/abs/2603.04321v1",
+    "http://arxiv.org/abs/2603.04317v1",
+    "http://arxiv.org/abs/2603.04314v1",
+    "http://arxiv.org/abs/2603.04309v1",
+    "http://arxiv.org/abs/2603.04308v1",
+    "http://arxiv.org/abs/2603.04293v1",
+    "http://arxiv.org/abs/2603.04291v1",
+    "http://arxiv.org/abs/2603.04289v1",
+    "http://arxiv.org/abs/2603.04277v1",
+    "http://arxiv.org/abs/2603.04276v1",
+    "http://arxiv.org/abs/2603.04259v1",
+    "http://arxiv.org/abs/2603.04247v1",
+    "http://arxiv.org/abs/2603.04245v1",
+    "http://arxiv.org/abs/2603.04244v1",
+    "http://arxiv.org/abs/2603.04384v1",
+    "http://arxiv.org/abs/2603.04380v1",
+    "http://arxiv.org/abs/2603.04337v1",
+    "http://arxiv.org/abs/2603.04319v1",
+    "http://arxiv.org/abs/2603.04304v1",
+    "http://arxiv.org/abs/2603.04299v1",
+    "http://arxiv.org/abs/2603.04292v1",
+    "http://arxiv.org/abs/2603.04257v1",
+    "http://arxiv.org/abs/2603.04238v1",
+    "http://arxiv.org/abs/2603.04217v1",
+    "http://arxiv.org/abs/2603.04212v1",
+    "http://arxiv.org/abs/2603.04162v1",
+    "http://arxiv.org/abs/2603.04161v1",
+    "http://arxiv.org/abs/2603.04145v1",
+    "http://arxiv.org/abs/2603.04124v1",
+    "http://arxiv.org/abs/2603.04123v1",
+    "http://arxiv.org/abs/2603.04083v1",
+    "http://arxiv.org/abs/2603.04069v1",
+    "http://arxiv.org/abs/2603.04033v1",
+    "http://arxiv.org/abs/2603.03915v1",
+    "http://arxiv.org/abs/2603.04399v1",
+    "http://arxiv.org/abs/2603.04395v1",
+    "http://arxiv.org/abs/2603.04392v1",
+    "http://arxiv.org/abs/2603.04383v1",
+    "http://arxiv.org/abs/2603.04360v1",
+    "http://arxiv.org/abs/2603.04354v1",
+    "http://arxiv.org/abs/2603.04353v1",
+    "http://arxiv.org/abs/2603.04343v1",
+    "http://arxiv.org/abs/2603.04340v1",
+    "http://arxiv.org/abs/2603.04328v1",
+    "http://arxiv.org/abs/2603.04325v1",
+    "http://arxiv.org/abs/2603.04323v1",
+    "http://arxiv.org/abs/2603.04300v1",
+    "https://huggingface.co/papers/2603.01236",
+    "https://huggingface.co/papers/2603.02924",
+    "https://huggingface.co/papers/2603.03282",
+    "https://huggingface.co/papers/2603.03197",
+    "https://huggingface.co/papers/2603.04124",
+    "https://huggingface.co/papers/2603.03646",
+    "https://huggingface.co/papers/2603.01847",
+    "https://huggingface.co/papers/2603.04254",
+    "https://huggingface.co/papers/2603.03985",
+    "https://huggingface.co/papers/2603.02482",
+    "https://huggingface.co/papers/2603.03790",
+    "https://huggingface.co/papers/2603.04291",
+    "https://huggingface.co/papers/2603.03823",
+    "https://huggingface.co/papers/2603.02604",
+    "https://huggingface.co/papers/2603.03379",
+    "https://huggingface.co/papers/2603.03447",
+    "https://huggingface.co/papers/2603.04257",
+    "https://huggingface.co/papers/2603.04338",
+    "https://huggingface.co/papers/2603.03975",
+    "https://huggingface.co/papers/2603.04379",
+    "https://huggingface.co/papers/2603.01375",
+    "https://huggingface.co/papers/2603.01563",
+    "https://huggingface.co/papers/2603.02510",
+    "https://huggingface.co/papers/2602.20629",
+    "https://huggingface.co/papers/2602.21773",
+    "https://huggingface.co/papers/2602.23546",
+    "https://github.com/login?return_to=%2fthecraighewitt%2fseomachine",
+    "https://github.com/login?return_to=%2finclusionai%2fareal",
+    "https://github.com/login?return_to=%2fmicrosoft%2fmcp-for-beginners",
+    "https://github.com/login?return_to=%2fmicrosoft%2fhve-core"
   ]
 }
\ No newline at end of file
diff --git a/daily/2026-03-05_en.md b/daily/2026-03-05_en.md
new file mode 100644
index 0000000..85d3bf0
--- /dev/null
+++ b/daily/2026-03-05_en.md
@@ -0,0 +1,61 @@
+# AI Daily Brief - 2026-03-05
+
+> Collected at: 3/5/2026, 11:01:12 PM
+> Total items: 89
+
+## 🔥 Top 10 Highlights
+
+1. Heterogeneous Agent Collaborative Reinforcement Learning
+   https://huggingface.co/papers/2603.02604
+
+2. Proact-VL: A Proactive VideoLLM for Real-Time AI Companions
+   https://huggingface.co/papers/2603.03447
+
+3. Words & Weights: Streamlining Multi-Turn Interactions via Co-Adaptation
+   https://huggingface.co/papers/2603.01375
+
+4. LFPO: Likelihood-Free Policy Optimization for Masked Diffusion Models
+   https://huggingface.co/papers/2603.01563
+
+5. Humans and LLMs Diverge on Probabilistic Inferences
+   https://huggingface.co/papers/2602.23546
+
+6. login?return_to=%2FTheCraigHewitt%2Fseomachine
+   https://github.com/login?return_to=%2FTheCraigHewitt%2Fseomachine
+
+7. login?return_to=%2FinclusionAI%2FAReaL
+   https://github.com/login?return_to=%2FinclusionAI%2FAReaL
+
+8. login?return_to=%2Fmicrosoft%2Fmcp-for-beginners
+   https://github.com/login?return_to=%2Fmicrosoft%2Fmcp-for-beginners
+
+9. login?return_to=%2Fmicrosoft%2Fhve-core
+   https://github.com/login?return_to=%2Fmicrosoft%2Fhve-core
+
+10. Robustness of Agentic AI Systems via Adversarially-Aligned Jacobian Regularization
+   http://arxiv.org/abs/2603.04378v1
+
+## 📂 Categories
+
+### AI Coding Tools / Code Agent
+
+- Code Fingerprints: Disentangled Attribution of LLM-Generated Code
+  http://arxiv.org/abs/2603.04212v1
+
+### AI Infrastructure / Inference Optimization
+
+- Low-Resource Guidance for Controllable Latent Audio Diffusion
+  http://arxiv.org/abs/2603.04366v1
+- Dissecting Quantization Error: A Concentration-Alignment Perspective
+  http://arxiv.org/abs/2603.04359v1
+- Activation Outliers in Transformer Quantization: Reproduction, Statistical Analysis, and Deployment Tradeoffs
+  http://arxiv.org/abs/2603.04308v1
+- Online Learning for Multi-Layer Hierarchical Inference under Partial and Policy-Dependent Feedback
+  http://arxiv.org/abs/2603.04247v1
+- $V_1$: Unifying Generation and Self-Verification for Parallel Reasoners
+  http://arxiv.org/abs/2603.04304v1
+- Bielik-Q2-Sharp: A Comparative Study of Extreme 2-bit Quantization Methods for a Polish 11B Language Model
+  http://arxiv.org/abs/2603.04162v1
+
+---
+*Generated by AINewsCollector*
diff --git a/daily/2026-03-05_zh.md b/daily/2026-03-05_zh.md
new file mode 100644
index 0000000..0ea53f1
--- /dev/null
+++ b/daily/2026-03-05_zh.md
@@ -0,0 +1,61 @@
+# AI Daily Brief - 2026-03-05
+
+> 采集时间: 2026/3/5 23:00:22
+> 总条目: 89
+
+## 🔥 Top 10 重要消息
+
+1. 异构智能体协同强化学习：突破孤立策略优化效率瓶颈
+   https://huggingface.co/papers/2603.02604
+
+2. “Proact-VL：实时低延迟视频LLM主动交互伴侣技术”
+   https://huggingface.co/papers/2603.03447
+
+3. 《基于协同适配的多轮交互流式策略调整》
+   https://huggingface.co/papers/2603.01375
+
+4. LFPO：基于似然无模型掩码扩散模型的强化学习
+   https://huggingface.co/papers/2603.01563
+
+5. 人类与LLM在概率推理上存在差异
+   https://huggingface.co/papers/2602.23546
+
+6. “构建长篇SEO优化博客内容的Claude Code工作空间：助力企业内容排名提升”
+   https://github.com/login?return_to=%2FTheCraigHewitt%2Fseomachine
+
+7. 闪电速RL助力LLM推理与智能体，简易灵活
+   https://github.com/login?return_to=%2FinclusionAI%2FAReaL
+
+8. 《跨语言MCP协议开源教程：多语言实践入门》
+   https://github.com/login?return_to=%2Fmicrosoft%2Fmcp-for-beginners
+
+9. 超高速工程组件集：助力Copilots项目高效启动与升级
+   https://github.com/login?return_to=%2Fmicrosoft%2Fhve-core
+
+10. 基于对抗对齐雅可比正则化的代理型AI系统鲁棒性研究
+   http://arxiv.org/abs/2603.04378v1
+
+## 📂 分类汇总
+
+### AI 编程工具 / Code Agent
+
+- LLM生成代码指纹：解耦大型语言模型代码归因
+  http://arxiv.org/abs/2603.04212v1
+
+### AI 基础设施 / 推理优化
+
+- 低资源引导实现可控音频扩散生成
+  http://arxiv.org/abs/2603.04366v1
+- 量化误差剖析：基于集中对齐的函数保持变换
+  http://arxiv.org/abs/2603.04359v1
+- Transformer量化中激活异常的再现、统计分析与部署权衡
+  http://arxiv.org/abs/2603.04308v1
+- 在线学习实现分层推理的在线学习
+  http://arxiv.org/abs/2603.04247v1
+- 统一生成与自验证：并行推理器测试时扩展新方法
+  http://arxiv.org/abs/2603.04304v1
+- 《Bielik-Q2-Sharp：极低比特量化在波兰11B语言模型中的比较研究》
+  http://arxiv.org/abs/2603.04162v1
+
+---
+*Generated by AINewsCollector*