AI news for Monday, 26 October 2026
Stories 61 to 101 of 101, most important first. Back to the first page →
-
SK Hynix reportedly gained ~$15B from Kioxia stake sale and plans ~$12B dividend
SK Hynix gained approximately $15 billion earlier this year by reducing its stake in Kioxia, according to a Bloomberg report. The memory chipmaker plans to return about $12 billion of those proceeds to shareholders…
1 sourcebloomberg.com -
Self-Direction lets agents learn from mistakes, boosting success rates
Self‑Direction is a new mechanism that lets agents learn from their own mistakes without expensive fine‑tuning. When an agent fails, it reflects on the trajectory, extracts a lesson, and distills it into a lightweight…
2 sources primary sourcearXiv cs.AI -
Researchers warn automatic scorers may mislead LLM evaluation studies
A new paper on arXiv argues that many LLM studies overlook how automatic scorers can distort results when comparing model responses to different prompts. The authors note that scorers often use a single overall…
1 source primary sourcearXiv cs.CL -
Researchers propose agentic framework for emotion reasoning with evidence checks
A team of researchers introduced an agentic framework for multimodal emotion reasoning that uses evidence-grounded verification to improve accuracy without retraining models. Their approach splits tasks into…
1 source primary sourcearXiv cs.CL -
Google releases ML Drift GPU inference engine as open-source
google has released ml drift gpu inference engine as open-source under apache 2.0, infoq reported on october 25 2026. it replaces the tflite gpu delegate, which will no longer receive new features. ml drift offers a…
1 sourcetechtimes.com -
Kevin Roose’s book details AI labs’ shift toward AGI
Kevin Roose’s The AGI Chronicles traces how AI labs like OpenAI, Anthropic, and Google moved from dismissing artificial general intelligence (AGI) as a pipe dream to treating it as an urgent, existential goal. Roose, a…
1 sourceThe Verge AI -
Jane Street invests in AI chip startups and builds data centers
High‑frequency trading firm Jane Street is expanding its AI capabilities by funding emerging AI and chip startups and by constructing its own data‑center facilities. The move reflects a broader trend among quantitative…
2 sourcesThe Decoderbloomberg.com -
OpenAI may release GPT bell with security upgrades
OpenAI’s rumored GPT Bell, possibly GPT‑6.5 or GPT‑7, is said to bring major security fixes after a DNS misconfiguration exploit that halted training. The update is expected to improve natural‑language understanding…
1 sourcegeeky-gadgets.com -
Resolution autoformalizes math to speed AI alignment research
Resolution, an AI safety startup, published work on autoformalizing mathematical statements related to statistical learning theory. The team used agents to develop plans—one high-level and one detailed—and an…
1 sourceImport AI -
OpenAI’s internal model solves Navier-Stokes using a swarm of 10,000 AI agents
OpenAI’s unnamed internal model—claimed to surpass GPT-6 Astra—solved the long-unsolved Navier-Stokes equations by coordinating a colony of 10,000 AI agents. Unlike traditional AI approaches, the model did not work…
1 sourcetime.com -
Researchers propose method to safely compress AI agent prompts in production
A team of researchers has published a paper on arXiv outlining a new approach to evaluating and compressing prompts for AI agents in real-world production environments. The method addresses the challenge of maintaining…
1 source primary sourcearXiv cs.AI -
Claude skill model-cache-cleaner finds and safely removes local AI tool caches
A new Claude skill called model-cache-cleaner scans a Mac for storage used by local AI tools, shows what it finds, and moves only user-approved items to the Trash. The skill checks known locations such as…
1 source primary sourcegithub.com -
AI agents navigate software via clicks and typing without APIs
Computer-using agents are AI systems that perceive and interact with software interfaces—clicking, typing, scrolling—using visual cues and reasoning instead of relying on dedicated APIs. This allows them to operate…
1 sourceUnite.AI -
Foundation models match but sometimes lag task-specific models in prostate pathology
The study evaluated more than 100,000 prostate biopsy images from 6,268 patients across 15 sites, comparing self‑supervised foundation models (FMs) with task‑specific (TS) end‑to‑end models. Both model families reached…
1 source primary sourceNature Machine Learning -
Researchers release ProBEiT, a 400M-parameter protein model
Researchers have introduced ProBEiT, a multimodal protein language model designed to integrate structural information with amino-acid sequences. The architecture uses an implicit mixture-of-experts scheme with…
1 source primary sourceNature Machine Learning -
Researchers propose new framework for training LLM agents with mixed On-Policy and Off-Policy data
A new paper on arXiv argues that training interactive LLM agents requires reconsidering how on-policy and off-policy distillation are applied. The authors introduce a hierarchical distillation framework that separates…
1 source primary sourcearXiv cs.AI -
Researchers use Claude Opus 4.8 to auto-design sub-THz transmitter with EDA tools
Researchers at an unnamed lab introduced TACS-RF, a framework where a Claude Opus 4.8 LLM agent automates RF and mmWave module design by directly controlling industry-standard tools like Cadence Virtuoso/Spectre and…
1 source primary sourcearXiv cs.AI -
Study finds no performance difference between FISER and CEMIR on HandMeThat v2
Pragmatic intent recognition is essential for human‑robot interaction, allowing agents to resolve ambiguous natural‑language instructions by inferring the speaker’s underlying goals. Recent research has focused on two…
1 source primary sourcearXiv cs.AI -
Coupang and OpenAI co-host AI grand prix singapore 2026 hackathon
The overnight AI Grand Prix Singapore 2026 was held at Carousell Campus from 24 to 25 October. Sixty participants, chosen from 150 applicants, tackled three Codex‑based challenges: Game Hack, Spot the Mismatch, and…
1 sourcemanilatimes.net -
Claude Code vs Cursor field test shows token usage differs
A field test compared token usage of Claude Code and Cursor, two code generation tools. The article cites two benchmarks that report conflicting token counts. The first benchmark indicates that Claude Code used 33,000…
1 sourceTowards Data Science -
Marissa Mayer's Dazzle builds AI assistant for non‑tech users
Marissa Mayer, former Google vice president and former Yahoo chief executive, has launched a new startup called Dazzle. The company employs a team of ten engineers who are developing an artificial‑intelligence‑driven…
1 sourcenytimes.com -
EBRD selects Cognizant for five-year AI-led technology transformation partnership
The European Bank for Reconstruction and Development (EBRD) has chosen Cognizant as a strategic partner under a five-year framework to support its Tech Blueprint technology transformation programme. The partnership,…
1 sourceUnite.AI -
Arya.ai studies recovery in AI agent workflows
The authors present a study on durable AI agents that can resume actions when authority changes and outcomes remain uncertain. The research isolates recovery decisions from language‑model behavior, examining checkpoint…
1 source primary sourcearXiv cs.AI -
Researchers release LAMB, a Latin language model trained on 74B tokens
It uses the ModernBERT architecture and was pretrained in two phases on over 74 billion Latin tokens, with a novel masking method informed by morphology. The model was developed alongside two new Latin datasets:…
1 source primary sourcearXiv cs.CL -
Marissa Mayer’s Dazzle app sparks backlash for AI-generated daughter traits
Marissa Mayer, former Google executive and founder of the AI app Dazzle, revealed in a New York Times interview that she used the tool to generate three qualities for her 10-year-old daughter’s school assignment. The…
1 sourceyahoo.com -
Verosoft releases Verovia to automate industrial AI workflows with Microsoft tools
Verosoft, a Microsoft-focused industrial tech firm, launched Verovia, an AI orchestration platform on October 26. The tool connects industrial equipment, enterprise software, and AI agents to automate operations…
1 sourcefinance.yahoo.com -
Researchers propose 24 quality dimensions for knowledge graphs used by LLMs
A new paper on arXiv outlines how knowledge graphs (KGs) interact with large language models (LLMs) and introduces a framework for assessing their quality. The authors argue that KGs—structured data used to ground LLM…
1 source primary sourcearXiv cs.CL -
Google Gemini predicts hypothetical Hyperliquid surge by Q4 2026
Google’s Gemini AI generated a speculative scenario suggesting Hyperliquid’s token, HYPE, could rise sharply if selected as the primary liquidity engine for a new generation of regulated on-chain prediction markets.…
1 sourcefinance.yahoo.com -
Narmi launches Microsoft 365 Copilot connector for secure banking data access
Narmi announced a new connector that brings customer financial data into Microsoft 365 Copilot for banks and credit unions. The integration uses Narmi's Model Context Protocol to let staff query customer information,…
1 sourcefinance.yahoo.com -
Matt Wood says paid software may not be over
Matt Wood, Senior Director of Cyber Security Architecture at ABBYY, argues that the rise of AI‑driven software cloning does not spell the end of paid software. Wood compares the current AI landscape to music sampling,…
1 sourceUnite.AI -
OpenAI Agents SDK lets Python developers build web-searching, tool-using agents
The OpenAI Agents SDK simplifies creating AI agents in Python that can search the web, call custom Python functions, remember past conversations, and trace their steps. Released in a tutorial by KDnuggets, the SDK…
1 sourceKDnuggets -
Researchers explore how aligned LLMs generalize to new tasks
A new paper on arXiv examines how large language models can align their output distributions with human references and whether those models retain performance on unseen questions or unrelated topics. The study uses two…
1 source primary sourcearXiv cs.CL -
Researchers predict conversation success using multimodal data in autistic and non-autistic dyads
A new study published on arXiv explores how to predict perceived conversation success in spontaneous interactions between autistic and non-autistic individuals. The research uses a combination of articulatory, facial,…
1 source primary sourcearXiv cs.CL -
OpenAI reportedly bans user for reverse-engineering TV
A user known as @log4j lost his OpenAI account after attempting to reverse-engineer a Samsung television he owned. According to reports, the activity was classified as 'Cyber Abuse' by OpenAI and resulted in a $500…
1 sourcetech.yahoo.com -
Opinion: AI safety focus ignores everyday harms like teen dependence and cognitive decline
An opinion piece argues that AI safety debates overlook the daily harms caused by AI products. While companies meet basic safety requirements—like preventing suicide prompts or fentanyl access—they often design…
1 sourceMIT Technology Review AI -
Researchers propose BERT-based system to detect Patronizing language
A new paper from arXiv introduces a system for automatically detecting patronizing and condescending language (PCL) in English and Mandarin. The authors note that PCL is a form of harmful speech that belittles…
1 source primary sourcearXiv cs.AI -
Researchers find trust-aware AI agents fail to improve negotiation outcomes
A new study on arXiv examines how AI agents negotiate in repeated multi-party settings where one participant tries to sabotage deals. The authors designed a framework where agents assess trust after each round and…
1 source primary sourcearXiv cs.AI -
Ben Thompson argues public internet is dead, AI security risks rising
Ben Thompson, writing for Stratechery, argues that the traditional public internet model has effectively collapsed due to content saturation, algorithmic aggregation, and escalating security threats. He notes that…
1 sourcestratechery.com -
KDnuggets outlines four-stage roadmap for computer Vision engineers
The guide breaks down skills into four stages, starting with spatial mathematics and foundational vision primitives like matrix transformations and projective geometry. It emphasizes hands-on practice with PyTorch for…
1 sourceKDnuggets -
Monolithic Power Systems and Analog Devices linked to AI data center power demand
Markets are being pulled by elections in Brazil and Serbia, oil worries in the Middle East, and a busy earnings week for Alphabet, Microsoft and Amazon. The article uses this backdrop to highlight three AI‑focused…
1 sourcefinance.yahoo.com -
Machine learning explores quantum superpositions but cannot solve Schrödinger’s cat paradox
This opinion piece argues that while machine learning cannot resolve the philosophical measurement problem in quantum mechanics—like Schrödinger’s cat paradox—it is reshaping how researchers observe, predict, and…
1 sourceTowards Data Science
