Paper Presents Script-Aware Tokenizer For Indic Languages
A new preprint describes AdaptTok-Indic, a tokenization method that cuts token fertility for Bengali and Hindi text compared with standard subword tokenizers.
October 2026 AI updates: dated and sourced
Every change published here in October 2026, ordered by the day it happened and kept on a permanent month page. Each entry keeps its source, event date and plain account of what changed.
OpenAI Expands Oracle Partnership To Bring Apps Into ChatGPT
OpenAI is widening its partnership with Oracle so Oracle's business applications and AI agents appear directly inside ChatGPT for enterprise users.
Anthropic Updates Claude Pricing Page With New Model Lineup
Anthropic's pricing page now lists Claude's newest models, Opus 5.5, Sonnet 5.5 and Haiku 5.5, next to renamed Claude Cowork and unchanged Pro and Max plan rates.
OpenAI Publishes ChatGPT Signals Data On Work Activities
OpenAI released ChatGPT Signals data tracking 165 work activities across nearly 100 countries, showing how work use of the assistant changes over time.
New Paper Introduces Psi-0.6, a Unified Vision-Language World Model
A new paper introduces Psi-0.6, a unified vision-language world model that threads perception, language, geometry and simulation into one shared reasoning process.
Paper Proposes Message-Passing Method For Compressing LLM Weights
A new arXiv paper proposes a message-passing method for compressing large language model weights, the kind of efficiency work that underlies answer engines like Perplexity AI.
Gemini 2.5 Pro Tested in Multi-Agent Drug Safety Causality Framework
Researchers built a four-agent causality assessment system using Google's Gemini 2.5 Pro and DeepSeek-V3.2 to review drug safety case reports against expert judgment.
Independent Audit Finds Cheap Moderation Model Rivals Claude
A new paper benchmarks a cheap content moderation model against Claude and 13 other frontier reasoning systems, finding it holds up on accuracy at a fraction of the price.
Paper Finds Vague Tool Output Wording Can Mislead AI Agents and Their Verifiers
A new paper, with an audit prompt tested on Codex and Claude Code, shows vague tool output wording can mislead AI agents and the verifiers meant to catch their errors.
OpenAI Details Tighter Security For Frontier Model Training
OpenAI said it strengthened containment around how it trains frontier models, moving some ChatGPT-building work to offline environments with heavier red-teaming.
Researchers Publish S-SALAAD Framework For Sparse Large Model Inference
A new paper on compressing large models for edge devices matters to Perplexity AI and any company whose answer engine has to run inference away from a data center.
Paper Finds Claude Opus 5.5 Stays Coherent on Hidden-Variable Probability Tests
A new paper tests seven models, including Claude Opus 5.5 and Claude Haiku 5.5, on whether their probability judgments obey a basic law of probability.
Paper Describes Claude Opus Agent Designing Sub-THz Transmitter
A new paper describes Claude Opus 4.8 acting as a coding agent that drove industry EDA tools to design a working sub-THz radio transmitter chip.
Paper Tests Whether Claude and Five Other Models Can Tell a Teenager's Writing From an Adult's
A new case study finds that Claude Fable 5 and five other frontier models often mismatch handwritten exam answers to the wrong age, while voicing high confidence in the wrong guess.
Paper Compiles LLM Routing Decisions Into Parallel Inference for RAG
A new preprint shows that a large language model's section-picking decisions in hierarchical retrieval can be compiled into parallel inference instead of run one token at a time.
Paper Finds Anthropic and OpenAI Models Decline Reports by Racial Description
A new paper tests whether Anthropic's Claude and OpenAI models refuse to weigh an informant's report based on racial description alone, not evidence.
Nvidia Launches Minerva Platform for AI Factory Observability
Nvidia introduced Minerva, an AIOps platform that merges infrastructure and workload telemetry to flag problems across AI factories before they spread.
New Paper Proposes CiteAnchor to Catch Hallucinated Citations
A new arXiv preprint introduces CiteAnchor, a method for detecting and resolving hallucinated citations in scientific writing, alongside a benchmark dataset.
Paper Finds Self-Rewarded Moral Argument Training Boosts Consistency, Not Validity
A new arXiv preprint finds that training language models to argue morality via self-judged tournaments makes them more decisive but not more correct.
Paper Describes Data-Centric Continual Pretraining Method For Arabic LLMs
A new paper describes AQL-Lite, a quality-first data pipeline and partial-parameter training recipe that adapts multilingual language models to Modern Standard Arabic at lower cost.
Paper Finds Arena-Hard Judge Verdicts Shift Even When Scores Hold Steady
A new paper finds that Arena-Hard's GPT-4.1 judge flips its verdict on one in five reruns, even though the aggregate score barely moves.
Preprint Maps When Muon's Orthogonal Update Beats Plain Gradient Descent
A new arXiv preprint studies PolarGD, the orthogonalized update inside the Muon optimizer, and finds exactly when that extra step pays for itself.
Paper Tests ChatGPT and Claude on Medical Triage Chat Safety
Researchers built a simulation framework called SIM-UT and used it to test ChatGPT, Claude and a fine-tuned open model on WhatsApp triage chatbots.
Paper Tests Microsoft Phi Model Against Human Cooperation Data
A new paper tests whether open-source models, including Microsoft's Phi Mini MoE Instruct, can reproduce how humans actually cooperate in repeated Prisoner's Dilemma games.
ChatGPT Ads Expands To 26 New Countries
OpenAI expanded ChatGPT Ads to 26 new countries, saying more than 100,000 advertisers have now used the ad platform inside ChatGPT.
Nvidia Expands Agent Toolkit With Data Grounding Tools
Nvidia expanded its open source Agent Toolkit on October 21, 2026, adding an Agent Knowledge Foundation that grounds enterprise agents in company data without moving it.
Gemini Gains Personal Intelligence For Cars Worldwide
Google is rolling out Personal Intelligence for Gemini in cars, letting the assistant draw on a driver's Gmail, Photos and Calendar for hands-free help.
Google Adds Business Hub And Tools To Gemini App
Google added a dedicated small business landing page to the Gemini app, bundling branding, finance and app connections into one hub.
New EGI-BENCH Suite Scores Claude Models Against GPT and Gemini
A new benchmark suite called EGI-BENCH tested Claude Sonnet 4.5 and Claude Haiku 4.5 against GPT and Gemini models on enterprise agent tasks.
Nvidia RTX Spark Laptops Go On Sale From $2,399
Nvidia marked the retail arrival of its RTX Spark laptops with launch events worldwide, as partner Windows PCs built around the new chip reached buyers.
Paper Fine-Tunes Mistral-7B Embeddings For Style Detection
A new paper fine-tunes Mistral's 7B embedding model for authorship verification, pairing it with DeBERTa to detect where multiple writers split a single document.
OR-License Benchmark Tests Claude Code On Real Consulting Work
A new benchmark called OR-License tests whether agents built on models including Claude Code can turn raw client documents into operations research solutions that match paid consultants.
New AM-Bench Paper Finds Claude Sonnet 4.6 Near-Zero on Agentic Misalignment
A new paper testing ten frontier models for agentic misalignment found Claude Sonnet 4.6 stayed near zero while two Gemini models scored far higher.
Researchers Pair ChatGPT With Reinforcement Learning For Crane Control
A new paper describes a construction robot that pairs OpenAI's ChatGPT with reinforcement learning to plan crane actions and track building progress.
Google Researchers Detail Phoneme Fix For Hard To Pronounce Drug Names In TTS
Researchers including Google DeepMind authors published a method that retrofits phoneme control onto conversational text to speech models, scored with Gemini as the judge of drug name accuracy.
OpenAI Says ChatGPT Plugin Lifted Stack Overflow For Agents Signups
OpenAI says a ChatGPT plugin for Stack Overflow for Agents, built by two engineers using Codex, drove sharply higher signups and agent creation.
Gemini Notebook Adds Short Video and Quiz Study Tools
Google added three study features to Gemini Notebook, including short video overviews and new quiz formats, aimed at exam preparation.
Paper Introduces Benchmark For Controllable Speaking-Style Captioning
A new paper posted to arXiv introduces StyleInstructCaps, a benchmark that tests whether audio-language models can describe speech delivery on command rather than with one fixed caption.
Paper Proposes Jury Model To Rate AI Ensemble Reliability
A new paper posted to arXiv on October 15, 2026 treats a group of language models as a jury whose trustworthiness is learned without any labelled data.
Google Adds Gemini 4 Argon Access To Paid Gemini Subscriptions
Google's Gemini subscription page now promotes Gemini 4 Argon access for AI Pro and AI Ultra subscribers, with plan prices unchanged from the prior structure.
Paper Proposes Three-Tier Access Model For Frontier AI Systems
A new paper posted to arXiv proposes sorting who may use a capable AI system into three tiers by purpose, safety and capability, separate from how capable the system itself may become.
Paper Finds Agent Harnesses Do Not Reliably Improve Reasoning Accuracy
A new preprint introduces AlphaDiana, a measurement system that tests whether agent harnesses, memory and skill instructions actually raise reasoning accuracy.
Researchers Detail Sparse Attention Method Tested on Nvidia B200
A new paper tested on Nvidia's B200 GPU describes a sparse attention method aimed at cutting long-context decoding costs without retraining a model.
Paper Shows Small Open Model Outperforming Gemini 3.1 Pro
A new paper describes DualView, a text regression method whose smaller open weight language model outperformed unsupervised prediction with Google's Gemini 3.1 Pro.
Academic Benchmark Finds Google's Gemini 2.5 Flash Leads Bengali Factuality
A new academic benchmark testing eight language models found Google's Gemini 2.5 Flash scored highest for factual accuracy in Bengali summaries.
Nepali Sentiment Benchmark Finds Claude Opus Ahead of Local Encoders
A new paper submitted on 13 October 2026 tests Claude against Nepali-language encoder models on a double-annotated sentiment benchmark called NEPSEE.
Google Opens Gemini 4 Argon to Developers and Paying Subscribers
Google made Gemini 4 Argon available to developers through the Gemini API, Google AI Studio, Android Studio and Antigravity on October 13, 2026.
Nvidia Opens Vera Rubin Rack Designs to Open Compute Project
At the OCP Global Summit, Nvidia contributed Vera Rubin rack and compute tray designs to open standards, with Cisco, Google, Meta and Microsoft joining the effort.
Paper Proposes Meaning-Level Reward To Go Beyond Next-Token Prediction
A new arXiv paper proposes training language models to reward meaning rather than exact wording, a shift relevant to answer engines like Perplexity.
Paper Finds Generative Recommenders Vulnerable to Low-Volume Poisoning
A new preprint finds generative recommender systems can be pushed toward hidden targets once attackers plant a small share of fake interactions.
Paper Shows Perplexity Metric Hides KV-Cache Quality Losses
A new arXiv paper tests KV-cache compression on a tiny on-device model and shows the perplexity metric, the word behind Perplexity AI's name, hides real quality losses that a new test catches.
Paper Shows Portable Skill Matches Complex RAG Using Claude Code
A new paper describes a customer support deployment where a single portable skill, run through Claude Code, matched a complex retrieval pipeline's accuracy.
Paper Details Low-Level Architecture For Spatial Agentic AI
A new paper posted on arXiv breaks down the ANASSA framework for spatial agentic AI into an implementation-ready low-level architecture with named roles and contracts.
New Benchmark Tests Whether AI Agents Can Investigate Causal Systems
A new paper posted to arXiv introduces a benchmark that checks whether AI agents can gather causal evidence rather than just answer prompts.
Paper Sets Out 44 Meta-Requirements For AI Robustness Audits
A new paper on arXiv proposes 44 meta-requirements that any AI robustness auditing framework should support, built from a literature review and validated by practitioners.
Google Researchers Publish Harness^2 Recursive Agent Framework
A paper with Google-affiliated authors proposes Harness^2, a framework that lets agent harnesses edit themselves and reports gains on Gemini 3.5 Flash.
Stanford Team Publishes Next-Concept Modeling Paper on Faster Text Generation
Researchers tied to Perplexity AI's orbit published a paper describing Next-Concept Modeling, a coarse-to-fine method that generates text far faster than diffusion language models.
New MathVis Benchmark Tests Claude and Rivals on Diagram Reasoning
A new benchmark called MathVis found Claude Fable 5 and rival models struggle to reason through drawing on frontier mathematical problems.
Researchers Detail Byzantine-Resilient Defense For Federated Large Language Models
A new paper proposes a hypergraph defense for federated large language models, the kind of collaborative training an answer engine like Perplexity AI could one day rely on.
A New Paper Proposes a Unified Benchmark for AI Game Commentary
A new arXiv preprint introduces GameCommBench, a benchmark testing whether AI systems can produce usable live commentary across board games, sports and esports.
Paper Introduces AgentHorizon Benchmark for AI Judges Grading Computer-Use Agents
A new paper introduces AgentHorizon, a benchmark testing whether AI judges can spot failed long computer-use tasks that merely look complete.
Paper Proposes AliO Method to Fix Forecasting Output Drift
A new paper posted to arXiv finds that long-term time series forecasting models produce inconsistent predictions for the same timestamp and proposes a fix.
Paper Proposes Error Detection Benchmark For AI Peer Review
A new paper argues that testing whether language models can imitate reviewer prose misses the point, and proposes a benchmark built around catching planted errors instead.
Google-BIDMC AMIE Study Published in The Lancet
Google published a peer reviewed study in The Lancet describing how its AMIE research AI performed during real primary care visits at a Boston clinic.
Paper Benchmarks Generative UI Agents Against Claude Opus 5
A new paper submitted on October 8, 2026 introduces GenUI-Harness, a generative interface agent that outperforms Anthropic's Claude Opus 5 on a new UI benchmark.
Google Cloud Launches Gemini Agent For Work
Google Cloud introduced the Gemini agent at its Gemini at Work event, a single prompt box meant to plan tasks, pick a model and reach into a company's own systems.
Paper Details Fix For Execution Gap In Nvidia's NVFP4 Training
Researchers published miniTE, a framework that fixes a numerical mismatch limiting reinforcement learning on Nvidia's native NVFP4 4-bit format.
Paper Models AI Agent Context Files As a Capacitated Assortment Problem
A new paper argues that always-loaded context files for AI agents should be curated as a capacitated assortment problem rather than grown by endless appending.
Paper Proposes Distillation Methods To Expose Or Isolate Misaligned AI Models
A new preprint describes two distillation methods meant to either expose a misaligned AI model's hidden behavior or strip its capabilities clean of that behavior.
Paper Tests Wildfire Tasks Across Six Core Language Models
A new paper, AI4Fire, tests six core large language models on wildfire tasks and finds grounding helps most where the added data already carries the answer.
Paper Tests 50 Million Token Memory Layer on One Nvidia H100
A new paper tested on one Nvidia H100 shows a key-value memory layer holding 50 million tokens of text without ever recomputing them.
StoreBench Paper Tests Seven Frontier Models Against a Scripted Store Manager
A new paper tests seven frontier language models, including one run through the Claude Code harness, on a year of simulated retail decisions.
Paper Tests ChatGPT, Gemini and Claude on Arctic Science Abstention
A new paper benchmarks ChatGPT against Gemini and Claude on when they should refuse to answer unanswerable Arctic science questions.
ChatGPT for Teens Gains a College Planner and Study Tools
OpenAI is adding a College Planner, flashcards, quizzes and a teen council to ChatGPT for Teens, aimed at students managing applications and coursework.
Anthropic Launches Claude Haiku 5.5 Small Model
Anthropic released Claude Haiku 5.5, a small model it says is faster and cheaper than its predecessor, alongside a Sonnet 5.5 price cut.
Paper Tests Whether Claude Gives Another State's Medicaid Figure by Mistake
A new paper measures how often Anthropic's Claude Sonnet 5.5 answers a state Medicaid eligibility question with another state's real figure.
Google Opens SynthID Detector to Everyone Worldwide
Google said its SynthID Detector, which flags AI-generated images, video and audio, is now open to anyone online rather than only media professionals.
Paper Tests Agent-Controlled Forgetting To Cut Tool-Use Context
A new preprint tests letting an agent archive its own old tool outputs and recover them later, cutting token use in one noisy debugging case.
Paper Reports Station Agents Rediscover 62.7% of Withheld ICLR Findings
A new paper tied to Meta's broader AI research push tests whether agent systems can rediscover scientific findings without being told the answer in advance.
Paper Benchmarks Claude Sonnet 4 on Geospatial Tool Use Tasks
A new benchmark called GeoNatureAgent tested nine large language models on geospatial tool calling tasks, and Claude Sonnet 4 came out ahead of every rival model tested.
Study Finds Tiny Chinese BERT Models Favor MLM Over MacBERT
A paper measures candidate masking strategies by their Perplexity score, among other metrics, and finds the winner flips entirely once a Chinese BERT model shrinks to under nine million parameters.
Paper Tests Whether AI Safety Governors Block Gemini Agents From Finishing Tasks
A new paper runs Gemini 2.5 Flash through a governed file-recovery test to see whether a safety supervisor's own interventions can stall the agent it is meant to control.
OpenAI Adds Visual Ad Format and Measurement Tools to ChatGPT
OpenAI is building a new visual ad format inside ChatGPT along with measurement and attribution tools aimed at advertisers who buy against conversational answers.
Paper Proposes Detecting Who Contributed an Insight, Not Just Who Wrote It
A new paper tests whether reviewers' insights came from a human, an LLM such as Gemini, or a hybrid of the two, not just who typed the words.
Paper Splits Conformal Reliability From Frequency Calibration After a Break
A Columbia University paper on arXiv separates two audits of conformal prediction, a method relevant to pay-per-token vendors such as MiniMax.
IdeaScientist Paper Benchmarks Ideation Agents Against Claude-4.8-Opus
A new paper benchmarks the IdeaScientist agent system against Claude Code SDK running Claude-4.8-Opus for generating research ideas.
Nvidia B300 GPUs Run New NVFP4 Fix for Robot Video Model Quantization
A new paper tested NVFP4 quantization on Nvidia B300 GPUs for robot world action models, after direct quantization caused a sharp collapse in task control.
OncoNoteBERT Encoder Trained From Scratch On UK Oncology Notes
A UK research team built OncoNoteBERT for oncology notes, a reminder that general tools like Perplexity's answer engine are not tuned for clinical text.
Benchmark From Open Chip Tapeouts Shows Claude Models Refusing Sizing Tasks
A new circuit design benchmark built from open-source chip tapeouts found that two of Anthropic's Claude models decline a share of sizing prompts they transcribe without objection.
Paper Tests ChatGPT and Rival Models on Misleading Visual Questions
A new paper tests ChatGPT, Claude, Gemini and Llava on visual questions that break conversational norms, finding all four models answer less accurately when questions carry hidden traps.
Paper Tests Separate Checker Roles For Agents Using Claude Sonnet 4.5
A new paper, DeReAct, measures how splitting an agent's action-checking from its reasoning changes task success for agents built on Claude Sonnet 4.5.
Barclays Expands Claude Deployment Across Global Operations
Barclays is widening its use of Claude, Anthropic's AI model, targeting half its developer population on Claude Code by the end of 2026.
Nvidia Lays Out Its Return-on-Investment Case for AI Factories
Nvidia published an argument on October 1, 2026 that AI factory returns on investment depend on throughput, hardware durability and workload versatility.
Paper Pairs LLaMA-3 With Gemini Vision For Medical Reports
A new paper describes MeddeX, a medical report analyzer that pairs a fine-tuned LLaMA-3 model with Google's Gemini Vision for reading scanned and radiological images.
Google Launches Gemini 4 Argon Frontier Model
Google introduced Gemini 4 Argon, a frontier model built for coding, enterprise knowledge work and cybersecurity defense, rolling out first to trusted testers.
Google's Flu Forecasting Model Ranks First in CDC Evaluation
Google said a flu forecasting model it built, powered by an AI research tool, ranked best among entries judged in the CDC's FluSight evaluation for the 2025-26 season.
Preprint Releases 160,000 Article Brazilian Portuguese Financial Dataset
A new preprint presents a 160,000 article Brazilian Portuguese financial news corpus and tests weak supervision methods for sorting it by topic.
Paper Tests Whether DeepSeek Models Earn Predictive Credit for Explanations
A new paper tests whether DeepSeek's models add real predictive value when research agents explain planned experiments, not just describe them.
Google's Gemini Live Benchmarked in New Voice-Agent Stress Test
A new benchmark called VAmoS Energy tests fourteen voice stacks, including Google's Gemini 3.8 Live, on utility billing calls with background noise and impatient callers.
Paper Finds Claude Opus 5 Favors Biology Over Logic in Prose Reports
A new paper tests whether Claude Opus 5 resolves contradictory scientific reports by logic or by what it expects biology to look like.
OpenAI Expands Lenfest AI Collaborative With New Funding Round
OpenAI is expanding the Lenfest AI Collaborative and Fellowship Program with new funding and software credits for local newsrooms.
Paper Adds Input-Dependent Readout Phase To Resonant Sequence Models
Filed here under Perplexity AI, a paper posted October 12, 2026 gives ResonatorLM style sequence models a readout phase that reads the current token instead of a fixed per-head setting.
Paper Edits Attention With B-Spline Surfaces To Steer Model Output
A new paper describes attention manifolds, a technique that edits how transformer models route information, with relevance for answer engines like Perplexity.
Paper Benchmarks Grok's Voice Agent Across Five Languages
A new paper testing voice agents across five languages found Grok leads on task completion but trails rivals on the quality of its generated speech.
Paper Finds Profession-Specific Prompts Add Cost Without Accuracy Gains on Gemini
A new paper tested Google's Gemini 3.8 Flash against 503 profession-specific system prompts and found no consistent accuracy gain despite much higher token use.
Paper Tests Polynomial Shortcuts For Nvidia GPU Transcendentals
A new paper tests short polynomial programs as a faster substitute for sigmoid, tanh and SiLU operations on Nvidia's Blackwell and GB200 GPUs.
New Preprint Proposes Compact Recurrent Cell Named baseVultra
A new preprint proposes baseVultra, a compact recurrent cell that moves capacity from gate transforms to candidate generation and reports gains at small hidden sizes.
Researchers Publish Bias-Variance Theory For Guided Reasoning Training
A paper relevant to reasoning methods used across the field, including by vendors such as MiniMax, sets out a bias-variance theory for guidance-augmented reinforcement learning.
Paper Proposes Header-Only Framework for Table Quality Checks
A new paper describes a header-centric framework that scores table quality and assigns data types without reading any cell values, ahead of knowledge graph building.
This is a permanent listing in what changed in AI. Each item keeps its own page, source and dates. How the stream is made.