Instant AI that sees, hears, and talks + 9 papers & 10 AI trends | ArXiv June 25

Описание: 🔬 Today's ArXiv CS digest covers 10 hand-picked papers — starting with "Wan-Streamer v0.1: End-to-end Real-time Interactive". Each paper gets a full narrated walkthrough with methodology diagrams.

📅 Daily ArXiv CS Digest — June 25, 2026

📋 Full paper list in the pinned comment below.


━━━━ Timestamps ━━━━
0:00 Wan-Streamer v0.1: End-to-end Real-time Interactive Foundation Models
2:03 Recommendation as Generation: Unifying Personalized Video Generation and Recommendation at Industrial Scale
3:37 The Unfireable Safety Kernel: Execution-Time AI Alignment for AI Agents and Other Escapable AI Systems
5:29 Real-Time Voice AI Hears but Does Not Listen
7:12 Small edits, large models: How Wikipedia advocacy shapes LLM values
8:57 Concept Removal for Frontier Image Generative Models
10:25 Same Evidence, Different Answer: Auditing Order Sensitivity in Multimodal Large Language Models
12:14 In-Context World Modeling for Robotic Control
14:03 How Reliable Is Your Jailbreak Judge? Calibration and Adversarial Robustness of Automated ASR Scoring
15:49 Towards Scalable Multi-Task Reinforcement Learning with Large Decision Models
17:27 Today's Research Overview
17:39 All 10 Research Trends
17:52 Geometric Foundations of Learning Dynamics
18:24 Enhancing Speech and Audio Analysis
18:54 Advanced Reasoning and Knowledge Systems
19:33 Building and Securing AI Agents
20:10 Unlocking Fundamental Computational Principles
20:42 AI and Knowledge Ecosystems
21:16 Trustworthy and Interpretable Machine Learning
21:46 RL for Complex System Control
22:19 Securing and Optimizing AI Hardware
22:47 Robust AI for Complex Data

🔔 Subscribe for daily AI and science research digests.

#ArXiv #AI #machinelearning #deeplearning #NLP #llm #paperexplained #research #artificialintelligence #datascience #computervision #reinforcementlearning