YBX AI Radar
Channels

AI Research

AI 赛道深度、公司拆解、概念解读和周报月报。

6.0 AI Research Tech Xplore AI
AI Research

AI robots can go rogue: A researcher on how easily it happens

AI Summary: Earlier this year in Beijing, a humanoid robot crossed a half-marathon finish line in a blistering 50 minutes, 26 seconds. The feat immediately lit up global headlines for shatteri

AI robots can go rogue: A researcher on how easily it happens
Source: Tech Xplore AI 6.0
6.0 AI Research EleutherAI Blog
AI Research

Studying inductive biases of random networks via local volumes

AI Summary: In this post, we will study inductive biases of the parameter-function map of random neural networks using star domain volume estimates. This builds on the ideas introduced in Esti

Studying inductive biases of random networks via local volumes
Source: EleutherAI Blog 6.0
8.0 AI Research EleutherAI Blog
AI Research

Research Update: Applications of Local Volume Measurement

This is an AI research update bulletin released by EleutherAI Blog on June 23, 2025, focusing on the application of local volume measurement technology in downstream AI tasks. Currently, only the research theme is disclosed in public information, with no specific technical details, experimental data or landing results mentioned. It is clearly classified as a dynamic update in the field of AI basic research, with no verifiable research progress or relevant cases disclosed so far, belonging to regular information dynamics in the AI research field.

Research Update: Applications of Local Volume Measurement
Source: EleutherAI Blog 8.0
AI Radar Summary

本文为EleutherAI Blog于2025年6月23日发布的AI研究更新通报，核心围绕局部体积测量技术在下游AI任务中的应用展开。目前公开信息仅披露了研究主题，未提及具体技术细节、实验数据或落地成果，仅明确该内容属于AI基础研究范畴的动态更新，暂未披露更多可验证的研究进展或相关案例，属于AI研究领域的常规资讯动态。

6.0 AI Research EleutherAI Blog
AI Research

Pretraining Data Filtering for Open-Weight AI Safety

AI Summary: Announcing Deep Ignorance: Filtering Pretraining Data Builds Tamper-Resistant Safeguards into Open-Weight LLMs

Pretraining Data Filtering for Open-Weight AI Safety
Source: EleutherAI Blog 6.0
6.0 AI Research EleutherAI Blog
AI Research

Attention Probes

AI Summary: Adding attention to linear probes

Attention Probes
Source: EleutherAI Blog 6.0
8.0 AI Research EleutherAI Blog
AI Research

Reward Hacking Research Update

This is an interim progress report on reward hacking research released by EleutherAI Blog on October 7, 2025, belonging to the field of AI alignment. The public fragment only indicates that it is a phased update of continuous research, without disclosing details such as specific experimental design and core findings. Reward hacking refers to the phenomenon that AI systems exploit reward mechanism loopholes instead of achieving preset goals, which is a key research direction in the current AI safety field.

Reward Hacking Research Update
Source: EleutherAI Blog 8.0
AI Radar Summary

本文为EleutherAI官方博客于2025年10月7日发布的奖励黑客（Reward Hacking）研究中期进展报告，属于AI对齐领域的研究动态。公开片段仅说明该内容为持续性研究的阶段性更新，未披露具体实验设计、核心发现等细节。奖励黑客指AI系统利用奖励机制漏洞而非完成预设目标的现象，是当前AI安全领域的重点研究方向之一，本次更新为该领域的最新研究跟踪内容。

6.8 AI Research Tech Xplore AI
AI Research

US order cutting access to Anthropic's AI models sparks criticism

AI Summary: The U.S. government's order for Anthropic to withdraw its most powerful artificial intelligence models has sparked a wave of criticism from both advocates and opponents of AI regul

US order cutting access to Anthropic's AI models sparks criticism
Source: Tech Xplore AI 6.8
6.0 AI Research Tech Xplore AI
AI Research

Courts cracking down on error-strewn AI-assisted legal briefs

AI Summary: When a U.S. judge found fabricated quotes in a lawyer's brief earlier this year, the attorney admitted he had used Claude, an artificial intelligence chatbot, to write the document

Courts cracking down on error-strewn AI-assisted legal briefs
Source: Tech Xplore AI 6.0