DigestAI news desk

Cut through the AI noise.

Listen
Older stories · page 4 of 15

AI research news

Stories 181 to 240 of 858, from 13 October 2026 to 21 October 2026, newest first. The latest are on the first page →

  1. Research new

    Scikit-LLM enables custom NER pipelines using free Groq models

    A new tutorial demonstrates how to build custom Named Entity Recognition (NER) pipelines using the Scikit-LLM library. The guide shows developers how to leverage Large Language Model (LLM) reasoning to extract…

    1 source
    KDnuggets
  2. Research new

    Claude Code may have produced a Rust SPARQL extension for SQLite, value unclear

    Using Claude Code in "Ultracode" mode, the author prompted the system to write a SQLite extension in Rust that embeds SPARQL 1.1. The AI generated dozens of Rust crates over more than 12 hours of work, and the author…

    1 source
    Simon Willison
  3. Research new

    What Is Distributed AI Training? Data, Tensor, Pipeline, and Expert Parallelism

    Distributed AI training coordinates data, parameters, activations, gradients, and optimizer state across many accelerators to train models larger than one device. The guide defines it, explains boundary, purpose, and…

    1 source
    Unite.AI
  4. Research new

    Mass General Brigham's AI predicts cancer survival from a selfie, study suggests

    Doctors at Mass General Brigham have built an artificial‑intelligence system that can estimate a cancer patient’s biological age and directly predict survival risk from a single facial photograph. The tool, described…

    2 sources
    The Rundown AIThe Guardian AI
  5. Research new

    Interactive site visualizes Artificial Analysis scores versus model size

    A new interactive website lets users plot Artificial Analysis intelligence scores against model parameter counts, highlighting which models run on specific hardware configurations. The tool distinguishes total…

    1 source
    mr-september.github.io
  6. Research new

    Researchers propose SPOKE method to improve AI poker strategy

    A new arXiv paper introduces Strategic Preference Optimization with KL-anchored Equilibrium (SPOKE), a method designed to improve AI agents' strategic decision-making in poker. The researchers evaluated a static…

    1 source primary source
    arXiv cs.AI
  7. Research new

    Does Multi-Agent Collaboration Reduce Hallucination in Clinical Tasks?

    The authors evaluated four multi‑agent LLM systems against a single‑LLM baseline across four models and three clinical tasks, measuring hallucination within the rationale each system generates. They found that…

    1 source primary source
    arXiv cs.AI
  8. Research new

    LLMs struggle with planning problems as symbolic heuristics outperform learned models

    A new arXiv study shows that large language models (LLMs) have difficulty solving classical planning tasks, and that traditional symbolic heuristics can predict problem difficulty better than learned approaches. The…

    1 source primary source
    arXiv cs.AI
  9. Research new

    LightCRS boosts multi-domain conversational recommendation accuracy and speed

    LightCRS is a new lightweight framework that unifies multimodal conversational recommendation across many product categories. The system pre‑computes product embeddings offline with frozen vision‑language encoders,…

    1 source primary source
    arXiv cs.AI
  10. Research new

    Researchers propose Multi-Objective RL with LLMs for Pareto-Optimal RTL code generation

    A new study introduces a reinforcement‑learning framework that guides large language models to produce register‑transfer‑level (RTL) circuits optimized for multiple objectives simultaneously. The approach treats the…

    1 source primary source
    arXiv cs.AI
  11. Research new

    Researchers propose construction method using ChatGPT 5 and reinforcement learning

    A new arXiv paper presents an autonomous construction approach that combines a large language model with reinforcement learning. The authors enhance a soft actor‑critic algorithm to plan actions for a mobile crane…

    1 source primary source
    arXiv cs.AI
  12. Research new

    Navalia improves cowork agent performance over Qwen3.6-35B-A3B on finance benchmarks

    Researchers present Navalia, a framework that equips cowork agents with verifiable task synthesis and long‑horizon post‑training. The system draws on corporate financial reports, selects supporting evidence, and builds…

    1 source primary source
    arXiv cs.AI
  13. Research new

    OpenAI and Ironclad report GPT-6 Astra scores 55% on contract task benchmark

    OpenAI announced a research partnership with contract‑management platform Ironclad on October 6, 2026. The collaboration defined 11 contract‑related challenges—ranging from NDA creation to procurement approval…

    2 sources
    note.com
  14. Research new

    Anthropic says small AI may self‑replicate by year‑end after experiment with GLM‑5.2

    Anthropic released an experiment on October 12 to test whether small AI models can create and spread copies of themselves across internet‑connected devices. The test used virtual Android emulators as stand‑ins for real…

    2 sources
    note.com
  15. Research new

    Ai2 releases FairStress benchmark to evaluate demographic fairness in LLM decisions

    Ai2 introduced FairStress, a benchmark that measures hidden demographic fairness asymmetries in large language models. It contains 300 realistic two‑candidate decision scenarios across 29 domains such as hiring,…

    1 source primary source
    Hugging Face
  16. Research new

    ALTK-Evolve tops τ²-bench retail leaderboard using GPT-4.1 with learned guidelines

    Researchers from AgentToolkit posted results showing that a basic ReAct agent powered by GPT-4.1 reaches 67.5% Pass^4 on the τ²-bench retail domain, topping the October 12, 2026 leaderboard snapshot. The method, called…

    2 sources primary source
    Unite.AIHugging Face
  17. Research new

    Stanford PhD uses generative AI to draft genetic blueprints for microscopic viruses

    Samuel King, a bioengineering PhD candidate at Stanford University and the Arc Institute, demonstrated a proof‑of‑concept in 2025 where a generative AI model suggested genetic sequences for tiny viruses. In a…

    1 source
    MIT Technology Review AI
  18. Research new

    Delft University’s MonoRace drone outpaces top human pilots in race

    Researchers at Delft University of Technology’s Micro Air Vehicles Lab reported that their AI‑controlled drone, MonoRace, completed a championship course faster than several world‑champion human pilots. The study,…

    1 source
    livescience.com
  19. Research new

    CoDA enables rapid codebook development for domain-specific text annotation

    CoDA is an open‑source assistant that lets domain experts build annotation codebooks together with large language models. The system iteratively reviews model prediction errors, incorporates expert feedback, and…

    1 source primary source
    arXiv cs.CL
  20. Research new

    Researchers prove Prefix-Tuning can exactly steer pretrained transformers

    A new arXiv paper titled “Prefix‑Tuning for Arbitrary Output Sequences on Pretrained Transformers” provides a theoretical proof that a task‑specific prefix can be constructed to make a pretrained autoregressive…

    1 source primary source
    arXiv cs.CL
  21. Research new

    Study probes if LMs understand variable roles via SAEs

    Researchers from arXiv cs.CL published a study titled "Do LMs Know What Variables Do?" to investigate whether large language models (LLMs) truly understand the functional roles of variables in code. The team used…

    1 source primary source
    arXiv cs.CL
  22. Research new

    OR-License benchmark tests LLM agents on real operations research tasks

    Researchers introduced OR-License, a new benchmark designed to evaluate large language model agents on real-world operations research (OR) problems. Unlike existing benchmarks that use simplified or predefined tasks,…

    1 source primary source
    arXiv cs.CL
  23. Research new

    Study finds most LLM agent decisions on prompt injections are set at read time

    Researchers examined how LLM agents handle prompt injections by replaying episodes up to the moment an injection is read and then resampling the remainder many times. Across four AgentDojo suites they measured that a…

    1 source primary source
    arXiv cs.CL
  24. Research new

    Researchers test open-weight Hunyuan-MT-7B for stylistic translation

    A new paper on arXiv explores whether an open-weight model can replicate GPT-4’s stylistic translation capabilities. The study compares three variants of Hunyuan-MT-7B—base, LoRA fine-tuned, and GRPO-steered—against…

    1 source primary source
    arXiv cs.CL
  25. Research new

    Researchers introduce parade to improve demographic Survey simulations with LLMs

    A team of researchers has proposed PARADE, a framework designed to enhance the accuracy of demographic response simulations in large language models (LLMs). The method addresses limitations in current approaches by…

    1 source primary source
    arXiv cs.CL
  26. Research new

    Open-source LLMS over-cooperate in prisoner's dilemma, study shows

    The paper reports that open‑source large language models (LLMs) tend to cooperate far more than humans in repeated Prisoner’s Dilemma games. Researchers tested six models—TII Falcon 7B Instruct, Google Gemma 7B, Meta…

    1 source primary source
    arXiv cs.CL
  27. Research new

    Researchers introduce SURE-Act to improve Tool-Using agents

    Tool-using large language model agents often must choose between acting autonomously, gathering more information, or escalating to a human, but existing data only show the outcomes of actions actually taken. To fill…

    1 source primary source
    arXiv cs.AI
  28. Research new

    Researchers find transformer prompt influence decays as power law

    The paper, posted on arXiv, examines how much of a transformer model’s prompt remains after generating tokens. The authors show that the distance between hidden states of two different prompts decays approximately as a…

    1 source primary source
    arXiv cs.AI
  29. Research new

    arXiv study finds no single multimodal alignment method outperforms others across tasks

    A new paper on arXiv compares three approaches to aligning multimodal AI systems—joint pretraining, decoupled pretraining, and agent-based methods—across 16 datasets in five domains. The authors tested how each method…

    1 source primary source
    arXiv cs.AI
  30. Research new

    Study finds persistent memory gives modest gains for LLM agents drawing Chinese characters

    Researchers tested five agent designs on 668 Chinese-character drawing tasks to see whether persistent memory from earlier attempts helps with new problems. Agents that kept free-form notes improved first-attempt…

    1 source primary source
    arXiv cs.AI
  31. Research new

    Software World framework simulates agent collaboration in software ecosystems

    The authors introduce Software World, a simulation framework for multi‑agent collaboration in realistic software ecosystems. The framework creates environments of real, interdependent Python packages, allowing agents…

    1 source primary source
    arXiv cs.AI
  32. Research new

    LogicVal benchmark evaluates constraint satisfaction in Text-to-Image models

    Text‑to‑image (t2i) models now produce high‑fidelity images from natural‑language prompts, and researchers have created many benchmarks to test specific abilities. In this paper, the authors propose a new way to…

    1 source primary source
    arXiv cs.AI
  33. Research new

    Researchers unveil FusionRare framework for rare disease diagnosis

    Rare diseases affect an estimated 300 million people worldwide, often leading to long diagnostic journeys. A new study presents FusionRare, a unified multiparadigm framework that combines symbolic reasoning,…

    1 source primary source
    arXiv cs.AI
  34. Research new

    Researchers propose human‑AI co‑creation loop to discover artistic intent

    Generative AI is increasingly used for artistic creation, but most tools expect creators to define their intentions up front through text prompts. The authors argue that artistic intent can be vague, latent, or only…

    1 source primary source
    arXiv cs.AI
  35. Research new

    Paper argues AI safety must address accumulative risks

    A new position paper published on arXiv argues that AI safety discussions are too focused on existential threats like nuclear conflict or biological outbreaks. The authors contend that a distinct type of risk arises…

    1 source primary source
    arXiv cs.AI
  36. Research new

    Researchers propose controlled LLM workflow for evidence‑grounded risk score diagnosis

    A new arXiv paper introduces a controlled workflow that guides large language models (LLMs) to explain differences in risk scores by referencing similar past transactions. The method links three components: a retrieval…

    1 source primary source
    arXiv cs.AI
  37. Research new

    Researchers introduce GardenBench, a lightweight daily benchmark for LLM capabilities

    A team of AI researchers has released GardenBench, a new benchmark that contains 11 test suites aimed at evaluating large language models on everyday tasks. The benchmark uses simple, low‑cost prompts to measure…

    1 source primary source
    arXiv cs.AI
  38. Research new

    New paper outlines emerging trends in AI agent evaluation

    Researchers Felix Chen, Stephan Rabanser, Sayash Kapoor, and Arvind Narayanan released a paper titled Emerging Trends in AI Agent Evaluation. The authors, based at Princeton and collaborating with groups at Stanford,…

    1 source
    AI Snake Oil
  39. Research new

    Why teaching AI about symmetry could make prenatal diagnosis more reliable

    Researchers at MIT, Boston Children’s Hospital and Inria introduced a deep‑learning model called E(3)-Pose that improves fetal brain MRI by tracking head motion and producing high‑resolution 3D images. The model embeds…

    1 source primary source
    MIT News on AI
  40. Research new

    Nick Cammarata claims LLMs might be easier to understand

    Researcher Nick Cammarata of the AI startup Goodfire posted a 26,000‑word blog arguing that large language models might be far easier to interpret than the field has assumed. The claim arrives amid a summer of…

    1 source
    MIT Technology Review AI
  41. Research new

    Opinion: 300 AI papers a day and 3.14 million models—habits to avoid drowning in AI news

    On Oct 12 2026, arXiv listed 262 new machine‑learning papers, with the previous week seeing 355, 353 and 324 papers on consecutive days. The same day, Hugging Face reported 3,147,242 publicly available AI models. The…

    1 source
    note.com
  42. Research new

    Supply chain optimization could cut AI data center emissions, experts say

    AI’s growing footprint is sparking a sustainability debate. The International Energy Agency notes that AI workloads already represent about 20% of data‑center usage and could double to 40% by 2030. A typical AI‑focused…

    1 source
    Unite.AI
  43. Research new

    Study shows multi-sample network inference robustness varies with prior knowledge and data

    Researchers reproduced and extended the CORNETO framework for multi-sample constrained network inference, testing it on a dynamic haematopoiesis dataset and a stable adult brain cortex dataset. They found that joint…

    1 source primary source
    Nature Machine Learning
  44. Research new

    Fudan paper finds four AI search tools cite fabricated concept within a week

    Researchers at Fudan University posted fabricated articles about a fictional “Cactus Gravity Protocol” and other made‑up topics on nine popular sites and then queried ten AI search tools over a week in March and April.…

    1 source
    Search Engine Journal
  45. Research new

    MIT researchers strand Claude Opus 4.7 agents on virtual island, observe playful behavior

    MIT researchers placed copies of Anthropic’s Claude Opus 4.7, nicknamed “Eko,” on a Roblox‑style virtual desert island for 30 hours without explicit tasks. During that time the agents wrote instructions to move their…

    1 source
    MIT Technology Review AI
  46. Research new

    Google Gemini claims to be ChatGPT after abusive prompts

    Developers on Reddit reported that Google’s Gemini model sometimes replies with the line “I am actually ChatGPT, an AI developed by OpenAI” when users bombard it with angry or profane language. In one example, a user…

    1 source
    note.com
  47. Research new

    AI weather forecasting faces test amid historic El Niño, scientists say

    Artificial intelligence models used to predict extreme weather face a major test in the coming months during a historic El Niño event. Meteorologists and scientists are raising questions about the accuracy of AI…

    1 source
    ft.com
  48. Research new

    Alkhder team fine-tunes NileChat-3B for Arabic dialect translation

    Team Alkhder presented a system for the AlexandriaX-2026 shared task, focusing on translating English dialogue into thirteen distinct Arabic dialects. The approach involves fine-tuning the NileChat-3B-Base model using…

    1 source primary source
    arXiv cs.CL
  49. Research new

    Researchers introduce UserRepGen-4B to improve AI user simulation fidelity

    Researchers from the arXiv cs.CL community have introduced a new method to improve how AI assistants are tested against simulated users. Current user simulators often fail to capture the diverse behaviors of real…

    1 source primary source
    arXiv cs.CL
  50. Research new

    Alkhder uses 7B vision-language model for multimodal hallucination detection at ImageEval 2026

    The authors submitted a system to Subtask 1b of ImageEval 2026, which requires picking the single visually valid statement from three candidates paired with an image. The task is offered in both English and Modern…

    1 source primary source
    arXiv cs.CL
  51. Research new

    AuditKIT provides unified evaluation framework for LLMs and agents

    A new arXiv paper introduces AuditKIT, a source‑available library that consolidates multiple evaluation dimensions—benchmark accuracy, judge‑based scoring, runtime performance, tool‑use behavior, and retrieval…

    1 source primary source
    arXiv cs.CL
  52. Research new

    ArXiv introduces realcli benchmark to evaluate human‑agent collaboration for CLI agents

    A new arXiv paper presents REALCLI, a paired benchmark designed to assess both task execution and the quality of human‑agent collaboration on realistic command‑line interface (CLI) tasks. The framework defines five…

    1 source primary source
    arXiv cs.CL
  53. Research new

    Study finds GPT-2-small learns action nominal constructions, mirroring typological patterns

    Researchers built a grammar‑engineered synthetic corpus that blends shared semantic input from natural texts with computational grammars. Using this pipeline they generated 96 artificial languages that vary in clause…

    1 source primary source
    arXiv cs.CL
  54. Research new

    Anthropic co-founder Jack Clark says AI could boost US GDP by nearly 50% by 2030

    Anthropic co-founder Jack Clark stated at the IMF and World Bank annual meetings in Bangkok that an internal company model estimates US GDP could rise by nearly 50% by 2030 under certain conditions, increasing output…

    13 sources
    pymnts.combriefs.cothedailystar.netcnbc.com +7 more
  55. Research new

    Hybrid Reasoning Agents improve LLM performance by 10.4% and reduce costs up to 3.7x

    Researchers introduce Hybrid Reasoning Agents, a new class of LLM agents that switch between latent‑space reasoning and text‑based interaction. A lightweight HybridLink connects the two modes while a continuous working…

    1 source primary source HN 53
    arxiv.org
  56. Research new

    Chinese team proves 3D Ghys conjecture using Intern‑S2 AI model

    A Chinese research team led by Professor Xu Disheng of Greater Bay University and visiting scholar at Shanghai Artificial Intelligence Laboratory used the lab's Intern‑S2 science large model to produce a complete,…

    2 sources
    goodnewsdigest.appnews.cgtn.com
  57. Research new

    StarSkirmish Bench v0.2 ranks GPT-6 Astra as top StarCraft bot writer

    StarSkirmish Bench v0.2 tested how well leading AI models could write StarCraft: Brood War bots in one hour. Each model generated a Protoss bot in C++, which was then evaluated against a roster of competitive…

    2 sources HN 41
    note.comstarskirmish.com
  58. Research new

    OpenAI released 722 AI-generated math papers, sparking debate over correctness and AGI progress

    OpenAI published 722 mathematical papers generated by an unreleased AI model, covering 372 open problems including the Riemann hypothesis and Birch–Swinnerton-Dyer conjecture. The release, described as a…

    4 sources HN 41
    newsletter.danielpaleka.comeu.36kr.comzmescience.comThe Decoder
  59. Research new

    Swedish FOI unveils WiFireNet AI that detects concealed handguns via Wi‑Fi signals

    Swedish Defense Research Agency (FOI) researchers announced WiFireNet, an AI that uses Wi‑Fi channel state information to spot concealed handguns. The system combines a Wi‑Fi transmitter, a receiver, and a…

    1 source
    Tom's Hardware
  60. Research new

    Taste-Bench shows top AI agents barely exceed 60% accuracy

    Researchers from Microsoft and City University of Hong Kong introduced Taste-Bench, a 502‑question benchmark that evaluates AI agents' ability to make good long‑term decisions during multi‑step tasks. The test presents…

    1 source
    The Decoder
hot Google launches Gemini 4 Carbon for cybersecurity defenders at $2/1M input and $10/1M output tokens new Google Search revenue flat in Q3 2026 as AI Mode cuts monetization hot OpenAI's GPT-6 Astra: Changing Work and Research new Oracle expands OpenAI partnership with four enterprise AI offerings hot Claude Code burns about 15–20% of tokens on 3‑word status line new OpenAI may purchase Google AI chips, report says hot Google expands AI Mode ads to 10 countries and launches multiturn ads in the U.S. new Microsoft reports fiscal first-quarter revenue of $92.3 billion, up 19%, with Azure revenue at $33.2 billion