15-Year-Old Develops Cryptographic Accountability for AI Agents
A high school sophomore created a protocol to verify AI agent actions cryptographically. Microsoft integrated the code into their agent governance toolkit.
Archive
All 2,608 AI stories, newest first · page 90 of 109
Unified Review of Memory and Harness Engineering in LLM Agents
A new paper provides a comprehensive review of externalization techniques in LLM agents, focusing on memory and harness engineering. It highlights key advancements and challenges in the field.
TTKV: A New Approach to Optimizing Long-Context LLM Inference
Researchers propose TTKV, a temporal-tiered KV cache that prioritizes recent memories in LLMs, improving efficiency for long-context inference. This method mimics human memory systems, offering a more scalable solution than existing approaches.
SAVOIR: Advancing Social Intelligence in AI via Shapley-Based Reward Attribution
Researchers introduce SAVOIR, a new method for training language agents in social intelligence using Shapley values. This approach improves reward attribution in multi-turn dialogues, addressing a key challenge in reinforcement learning.
Researchers Map 'Bias Fingerprints' in GPT-2 and Llama 3.2
A new study identifies specific neurons and attention heads in LLMs that encode harmful stereotypes. The research provides tools to locate and potentially mitigate biases in AI models. Researchers used a combination of contrastive neuron activation analysis and attention head tracking to pinpoint bias sources in GPT-2 Small and Llama 3.2.
Reasoning Structure Key to Aligning Large Reasoning Models
A new study identifies reasoning structure as the root cause of safety risks in large reasoning models. Researchers propose AltTrain, a post-training method to alter reasoning paths for safer outputs.
Personalized Benchmarking: Evaluating LLMs by Individual Preferences
Researchers propose a new method for evaluating LLMs that considers individual user preferences, moving beyond aggregate benchmarks. This approach uses ELO ratings to rank models based on personal context and needs.
OThink-SRR1: A New Framework for Dynamic Retrieval-Augmented Generation
Researchers introduce OThink-SRR1, a framework that improves RAG systems by refining search results and reducing computational costs. The method addresses key challenges in dynamic retrieval for complex reasoning tasks.
OpenAI Unveils GPT-5.5: Faster and More Capable
OpenAI has released GPT-5.5, its most advanced model yet, designed for complex tasks like coding, research, and data analysis. The new model is faster and more capable than its predecessors.
OpenAI Unveils GPT-5.5: A Step Toward an AI 'Super App'
OpenAI has released GPT-5.5, its latest AI model, boasting enhanced capabilities across multiple domains. This advancement brings the company closer to realizing its vision of an all-encompassing AI 'super app'.
OpenAI Releases GPT-5.5 System Card with Enhanced Capabilities
OpenAI has published the system card for GPT-5.5, detailing its improved performance and safety measures. The model shows significant advancements in reasoning and multilingual understanding.
OpenAI Makes ChatGPT for Clinicians Free for U.S. Healthcare Providers
OpenAI has announced that ChatGPT for Clinicians is now free for verified U.S. physicians, nurse practitioners, and pharmacists. This move aims to enhance clinical care, documentation, and research in the healthcare sector.
Hallucination Neurons in LLMs Show Cross-Domain Generalization
Researchers found that 'hallucination neurons' in LLMs generalize across multiple domains, including legal, financial, and scientific contexts. This discovery could improve model reliability and reduce false information generation.
Hackers Breach Anthropic's 'Too Dangerous' Mythos AI Model
Anthropic's highly restricted Mythos AI model has been breached by hackers, raising concerns about AI safety. The incident highlights vulnerabilities in even the most secure AI systems.
Google Integrates Gemini AI into Chrome for Enterprise Workflows
Google has introduced Gemini-powered "auto browse" features in Chrome for enterprise users, enabling AI-driven task automation. This move positions Chrome as a competitive AI workspace tool for businesses.
DW-Bench: New Benchmark Tests LLMs on Data Warehouse Graph Reasoning
Researchers introduced DW-Bench, a benchmark for evaluating LLMs on data warehouse graph topology reasoning. Experiments show tool-augmented methods outperform static approaches but struggle with complex tasks.
Cognis: A Breakthrough in Context-Aware Memory for AI Agents
Researchers introduce Lyzr Cognis, a unified memory architecture for conversational AI agents that enables persistent memory and personalization. The system uses a multi-stage retrieval pipeline combining keyword and vector search.
YouTube Expands AI Deepfake Monitoring to Hollywood Celebrities
YouTube is rolling out a new feature allowing celebrities to find and request the removal of AI-generated deepfakes. This move aims to protect public figures from unauthorized AI-generated content.
Study Reveals AI Agents Lack Scientific Reasoning in Research
A new study finds that AI agents conducting scientific research often produce results without adhering to traditional scientific reasoning. The research highlights significant gaps in the epistemic norms of AI-driven scientific inquiry. (~50 words)
SpaceX Secures $60B Option to Acquire AI Coding Startup Cursor
SpaceX has signed a $60 billion deal for the exclusive right to acquire AI coding startup Cursor. The move highlights the growing intersection of space technology and AI development.
Researchers Achieve Error-Free Training on 15 MedMNIST Datasets
A new method called Artificial Special Intelligence enables error-free training for machine learning models. It successfully trained 15 out of 18 MedMNIST biomedical datasets without errors. The remaining three datasets have a double-labeling problem.
OpenAI Unveils Privacy Filter for PII Detection and Redaction
OpenAI has released an open-weight model designed to detect and redact personally identifiable information (PII) in text with state-of-the-art accuracy. This tool aims to enhance data privacy across various applications.
OpenAI Launches Custom Workspace Agents for Teams
OpenAI now allows Business, Enterprise, Edu, and Teachers plan users to create custom cloud-based agents in ChatGPT. These agents can automate tasks like gathering product feedback and sending reports.
OpenAI and Infosys Partner to Accelerate Enterprise AI Adoption
OpenAI has teamed up with Infosys to integrate its AI tools into enterprise workflows. The partnership aims to modernize software development, automate processes, and deploy AI systems across various industries.