Semiconductor Engineering collects several new chip technical papers: oxygen-doped WSe2 monolayer ballistic p-type transistors achieve low contact resistance and high current density; UC Santa Cruz explores the design space of backside clock mesh networks for 2nm GAAFET backside power delivery technology.
Mercor's research had 12 licensed accountants with an average of five and a half years of experience complete simplified tasks from the APEX Accounting Benchmark; results showed today's AI models outperform accountants on speed, accuracy, and cost, whereas 18 months ago the best models scored about 37% below the accountants' average.
TechCrunch reported on Graphite's new research: comparing 10,000 human articles written before ChatGPT's release with rewritten versions by various models, it identified 13,000 phrases that appear at least 2x more frequently in AI text.
A research team from Carnegie Mellon, MIT, NYU, and Stanford has developed an AI called Ataraxos that defeated Pim Niemeijer, considered the strongest Stratego player in history, with 15 wins, 1 loss, and 4 draws, using only 16 GPUs and a few thousand dollars for training. Stratego is an imperfect-information game with large amounts of hidden information and very long game timescales; even DeepMind had previously failed to build a machine that could reliably beat top human players.
Ataraxos, developed by a research team, defeated four-time world champion and widely acknowledged greatest Stratego player Niemeijer with an 85% effective win rate over three weeks of play, and won 38 of 40 games in an exhibition match at the 2025 world championship, using less than $8,000 of compute to crack imperfect-information games.
Why it matters: The article compares Ataraxos and DeepNash training costs and methods, giving readers insight into the latest progress in imperfect-information game research.
OpenAI disclosed a distillation theft campaign targeting its models' reasoning content: from July 24-25, more than 16,000 requests were made by over 4,000 users, involving more than 15,000 linked accounts. OpenAI said it shut them all down on July 28 and linked the core group to people associated with Moonshot AI (developer of the Kimi model), while noting these were attempted extractions.
A Google DeepMind team published a paper proposing a protein watermarking scheme called SynthIDBio, which can embed watermarks into AI-designed protein sequences without breaking their function. Built on SynthID technology and integrated into the popular protein design tool ProteinMPNN, the system introduces watermark-consistent amino acids as it places each residue, only at positions where function remains viable.
Google Research introduced the Diffusion Controller framework, which reframes the denoising process as a continuous control problem, using a lightweight "steering damper" network to guide image generation without touching the main model's weights. In experiments based on Stable Diffusion v1.4, its fully unlocked version achieved a 90% win rate over the baseline on the Human Preference Score (HPS-v2) benchmark, and supports black-box/gray-box control of closed-source models.
A Hugging Face team has proposed ProvenanceGuard, a source-aware factuality verification layer for MCP agents, which checks after generation whether the sources supporting each claim match the sources the answer states or implies, identifying cross-source confusion.
Import AI issue 474 covers three items: Michael Levin has published a paper proposing that the relationship between mind and brain is like that between mathematical patterns and physical outcomes—the mind is a non-physical pattern from "Platonic space" and the body is merely its interface into the physical world, citing experiments with xenobots, anthrobots, and more as evidence. The issue also discusses how robotics is preparing for an LLM-like moment but still lacks a general-purpose algorithm, and Zhipu AI's progress in starting an outer RSI (recursive self-improvement) loop.
A new working paper by CESifo economists Robert Fairlie and Jane Wu in Munich finds no evidence, in absolute or relative terms, of significant or widespread job displacement or hiring cutbacks among recent college graduates.
Google Research unveiled its AI video co-director research, building a multi-agent orchestration layer based on Gemini and Veo with four frameworks — Co-Director, CANVAS, A²RD and VQQA — that reformulates long-form video generation as a global optimization and world-state tracking problem.
A Microsoft Research study shows that running physical AI inference solely on robots' onboard GPUs limits performance, battery life and scalability, and offloading inference to edge or cloud GPUs can significantly improve it.
Microsoft Research's retrosynthesis model RetroChimera was published in the journal Nature, with its implementation and weights open-sourced. The model combines two sub-models—the Transformer-based R-SMILES 2 and the GNN-based NeuralLoc—and fuses their predictions with a learned reranking strategy, covering both common and rare reaction types.
A new OpenAI economic research report analyzes how workers use AI beyond traditional job roles and which new activities have become a routine part of their work. The report reveals patterns of AI use in real-world work.
Google Research's Retrieve-for-Train framework uses offline RL to discover reward-aligned query fan-out and distills it into supervision signals, bypassing LLM token-by-token chain-of-thought inference latency.
IBM Research has introduced the Consistency Analyzer and consistency guidelines in ALTK-Evolve, targeting the problem of agents producing unstable results across multiple runs of the same task, cutting the agent consistency gap from 24.4pp to 12.0pp.
Why it matters: The article offers a reusable Pass^k metric and consistency diagnostics, letting readers evaluate and improve their own agents' run-to-run reliability.
Google Research proposed ToolGrad at ACL 2026, an "answer-first" method for generating tool-use datasets: it first builds real tool-call chains, then iteratively proposes, executes, selects, and updates API workflows using textual "gradients."
OpenAI shared an AI-generated solution approach to the Navier–Stokes Millennium Prize Problem, including a solution write-up and a Lean formalized proof. The material comes from an official OpenAI announcement; the feed does not include the full text, so specifics of the proof and methods should be checked via the original link.
Why it matters: OpenAI has publicly released an AI-generated approach to the Navier–Stokes Millennium Prize Problem with a Lean formalized proof, letting readers track AI's actual output in mathematical proof work.
Google Research studied how polygenic risk scores (PRS) transfer from UK Biobank European populations to about 200,000 Japanese individuals in Biobank Japan, covering 8 clinical traits including BMI, blood pressure, HDL, LDL, and blood glucose, comparing elastic net, meta-analysis, and PRS-CSx.
Google Research, together with HHMI Janelia, the MRC Laboratory of Molecular Biology, the University of Cambridge and others, has published in Cell a complete connectome map of the male fruit fly brain and central nervous system, containing over 166,000 neurons and 125 million synaptic connections — the largest brain map to date by neuron count.
Why it matters: The article details the connectome's scale, the AI reconstruction methods, and an open-source viewer, so readers can gauge the current limits of connectomics and the resources available.
Hugging Face has launched BenchMIRT, a multidimensional IRT-based method that audits what capabilities LLM benchmarks actually measure at the level of individual prompts.
Google, in collaboration with NASA JPL, launched MAPL-EMIT, a deep learning framework based on EMIT hyperspectral data for automatically detecting and quantifying methane plumes and locating emission sources. The framework uses a Swin-S vision transformer and achieves 84% recall on expert-labeled plumes, trained on 3.6 million synthetic methane plumes.
Google Earth AI has introduced an experimental research capability called the Planetary Prediction Engine (PPE), which autonomously performs the full geospatial prediction workflow—data discovery, cleaning, feature engineering, model training, and evaluation—starting from natural language queries, compressing work that previously took weeks of manual data engineering down to a few minutes.
Google DeepMind launched what it says is the world's first double-blind evaluation for proprietary frontier models, confining external evaluations to cryptographically secured isolated environments to prevent benchmark contamination from models seeing test questions in advance. In partnership with the Singapore AI Safety Institute, OpenMined, AVERI and MLCommons, it tested a Gemini Flash Lite model with confidential benchmarks to improve the credibility and integrity of evaluations.
Google unveils GlucoFM, a self-supervised foundation model with a two-stream design that separates slow glucose baseline trends from short-term deviations. Across four cohorts, seven clinical prediction tasks and 14 evaluations, its mean PR-AUC improves from 54.7 to 58.8 over the strongest baseline CGM baseline, leading on all diabetes risk and beta cell dysfunction evaluations, and achieves the lowest MAE in postprandial glucose response prediction.
Google is presenting AgentHands at CHI 2026, a research prototype that lets AI agents generate speech-synced hand gestures in immersive environments like Android XR, building on Project Astra and Gemini 3.1 Flash Live.
Hugging Face proposes Quantization-Aware Healing (QAH), compressing GPT-OSS 120B to a 60B-parameter structure and quantizing it to MXFP4, then distilling directly from the original pre-compression model rather than from a recovered bfloat16 checkpoint.
Import AI issue 470: METR research finds AI brings significant speedups to cybersecurity discoveries, modest contributions to math, but no measurable acceleration yet for optimizing AI research algorithms. Teams from multiple universities propose the SPADE framework, which co-evolves environment synthesis with agent capabilities via self-play, reaching an average score of 58.3 on a suite of game environments with Qwen3-30B-A3B, 5.3 above the baseline. This issue also introduces Hawkeye, for building better GPU kernels.
Google Research introduces the Biomarker Discovery Framework, a multi-agent system that structures candidate biomarker prioritization into an iterative, human-supervised research pipeline combining hypothesis generation, statistical analysis, model training and adversarial validation. Across three cohorts totaling 9,279 person-observations, the system autonomously identified 41 candidate digital biomarkers for mental health and 25 metabolic candidates, and built novel composite features such as sleep-duration variability.
Google Research introduces the Mobility-Embedded POIs (ME-POIs) framework, which folds aggregated, anonymized visit-mobility patterns into language models' place representations, capturing POIs' temporal activity rhythms.
Hugging Face released research using three types of tests — consistency disagreement probes, digit silence retrieval, and spelling switches — to evaluate benchmark optimization (benchmaxxing) behavior in 11 open-source ASR models.
Why it matters: The article quantifies benchmark optimization in speech recognition with three probes and provides a model list, ratio data, and reusable detection scripts, helping readers avoid the pitfall of selecting models by WER alone.
SK hynix, together with researchers from UVA, UIUC, NTU, MIT and Yonsei University, has published a co-packaged optics (CPO) technology roadmap in Nature Electronics. The corresponding authors are Seunghoon Hong, SK hynix AI Infra Team Lead, and UVA professor Kyusang Lee.
An IBM Research team tested eight models on 585 multi-step tasks in AppWorld using ALTK-Evolve and found that how much memory an agent should get depends on the model's capabilities.
DiG-bench (Discovery in Games) has been released, a benchmark of 70 games for testing AI's ability to autonomously discover hidden rules through exploration; 21 of the games are publicly available.
Google Research has introduced PhotoScan, a deep learning framework that estimates body fat percentage, Android-to-Gynoid fat ratio, and visceral-to-subcutaneous fat ratio from standard 2D smartphone photos to predict insulin resistance.
Microsoft Research introduced the MindTopo benchmark, evaluating multimodal large models' ability to reason about and plan with topological concepts such as connectivity, closure, ordering, separation, and knots. Testing shows models perform markedly better on static image recognition than on interactive planning tasks, and both fall far short of human level, with errors mostly involving lost structural relations during planning or physically infeasible actions. The benchmark aims to provide a controlled diagnostic tool for research on agents in robotics and interactive environments.
Google Research released the WikiProfile benchmark (2,150 Wikipedia facts, each with 10 tasks) and a knowledge profiling framework. After evaluating 13 LLMs, it found that frontier models have 95–98% of facts encoded but still fail to directly recall 26–34%, and 11–12% still fail even with thinking enabled.
Microsoft Research introduces CARE-X, a research model and unified chest X-ray VLM built on SigLIP2-so400M and Phi-4-mini-instruct (3.8B), jointly trained with auxiliary heads to provide report generation and structured predictions with confidence, and aligned to clinical rewards using DAPO reinforcement learning.
Hugging Face has published a paper that enables offline distillation by caching the teacher's top-100 logits, and introduces a fused chunked KL loss that avoids generating full vocabulary-by-sequence-length matrices.