FiX: fine-grained forgetting in softmax attention
Yongduck Kim examines FiX’s feature-wise gates, numerical implementation and paged cache, distinguishing reported gains from unresolved limitations.
Read articleNEWS & INSIGHTS
Ideas, developments
and perspectives on IP.
Analysis of IP practice, AI technology and international protection, alongside news from IPLEX. Explore the articles in English.
336 articles and updates
AllPatentsAI & SoftwareOverseas TrademarksDesignsTrademarksIPLEX NewsOverseas DesignsOverseas PatentsCopyrightTechnology Listing
62 articles
MHAR: Reading earlier layers through different feature subspaces
Yongduck Kim examines Multi-Head Attention Residuals: depth routing, reported training results, implementation costs and the relationship between technical features and effects.
Read articleColumn: Claude Computer Use and the Data That Trains AI Agents
Writing for AI Times, IPLEX Managing Partner Yongduck Kim examines the training data behind computer-operating AI agents through U.S. Patent No. 12,585,862.
Read articleIP Strategy Begins with Understanding Technology: An IPLEX Interview
Explore how IPLEX combines AI and software expertise with a practical understanding of business strategy. The interview discusses education for founders, mentoring and communication with clients.
Read articleAI Patents: What Can Be Protected and How to Prepare
This guide explains the issues to consider and the materials to prepare when seeking patent protection for AI technology.
Read articleColumn: Google’s Computer Vision Reinforcement Learning Patent — Learning Rewards, Not Just Answers
AI Times, a Korean AI publication, has published Managing Partner Yongduck Kim's column on artificial intelligence patents.
Read articlePatent Expertise Grounded in AI and Software Technology
We advise on applications and patent strategy for AI, generative AI and LLMs, machine learning, deep learning and software.
Read articleTROLL Paper Analysis: Token-Level Trust Regions for More Stable LLM Reinforcement Learning Beyond PPO Clipping
Reward-based reinforcement learning is increasingly used to improve the reasoning capabilities of large language models. PPO-style policy optimization is a common approach. Methods such as GRPO, Dr.GRPO, GSPO and REINFORCE++ modify advantage estimation or normalization while seeking to avoid excessive policy changes in a single update...
Read articleUnderstanding the Paper “Agentic Reasoning for Large Language Models”
The term AI agent is used widely, but its meaning varies. It may describe a chatbot with search capabilities or a system that plans tasks and operates external tools. Recent research explores these distinctions...
Read articleWhy Do LLMs Lose Track as Conversations Get Longer?
Conversational AI can lose track of requirements as a discussion grows longer. A model may repeat a corrected mistake or continue relying on an earlier assumption even though the conversation remains in its context...
Read articleColumn: Google’s Generative AI Reasoning Patent — Teaching the Process of Thinking
AI Times, a Korean AI publication, has published Managing Partner Yongduck Kim's column on generative AI patents.
Read articleAgentGym-RL Paper Analysis: Training Long-Horizon AI Agents with Multi-Turn Reinforcement Learning
An AI agent asked to navigate a website, find information and recover from a failed action faces a different task from simply answering a question. It must coordinate a sequence of actions, observations and decisions, including correcting errors along the way...
Read article