This post has no text preview — click the link below to read the original article. This article has been indexed from Google DeepMind News Read the original article: Introducing Gemini 3.8 Live and 3.8 Live Extended Thinking
MANAS-2: Constrained Reconstruction for EEG Foundation Models
arXiv:2609.13717v2 Announce Type: replace Abstract: Masked reconstruction is widely used for EEG foundation models, but optimizing reconstruction on…
Off-Target Effects of Response-Style Alignment in a Korean 27B Language Model
arXiv:2609.11291v2 Announce Type: replace Abstract: We post-train Qwen3.8-27B for Korean response style — verbosity, list and markdown usage, discourse…
Unifying ICL, SFT, KL-Regularized RL Through a Bayesian Lens
arXiv:2609.05111v2 Announce Type: replace Abstract: Supervised fine-tuning (SFT), few-shot in-context learning (ICL), KL-regularized RLHF/RLVR, and…
API Benchmark Scores Do Not Reliably Transfer to Chatbot Interfaces
arXiv:2609.08861v2 Announce Type: replace Abstract: Benchmark scores are a central currency in model releases: they inform purchasing decisions, shape…
EdiTikZ: Scientific Figure Editing from Revision Trajectories
arXiv:2609.01409v2 Announce Type: replace Abstract: Vision-language models (VLMs) have shown strong performance in generating scientific figures from text…
BlueLM-GUI Technical Report: A Real-Device-Centric Flywheel for Self-Improving Mobile GUI Agents
arXiv:2609.12394v3 Announce Type: replace Abstract: Mobile GUI agents are shifting from multi-module frameworks to native models trained end-to-end, yet…
AI News Brief Hourly Summary 2026-09-17 04h : 17 posts
17 posts published in the last hour 01:32AutoResearch: Insight In, Hallucination Out 01:32CASCADE: An Agentic Regulatory Network Framework for Patient-Data-Validated Downstream Perturbation Prediction 01:32HoloAegis: Frozen Representation, Topological Inference — Minimally Parametric Safety Manifolds and Their Capability Boundaries for LLM Guardrails…
AutoResearch: Insight In, Hallucination Out
arXiv:2608.17906v3 Announce Type: replace Abstract: Autonomous research systems are increasingly capable of executing long research workflows, yet…
CASCADE: An Agentic Regulatory Network Framework for Patient-Data-Validated Downstream Perturbation Prediction
arXiv:2608.05359v2 Announce Type: replace Abstract: CASCADE is an agentic framework that predicts downstream transcriptional effects of gene perturbation…
HoloAegis: Frozen Representation, Topological Inference — Minimally Parametric Safety Manifolds and Their Capability Boundaries for LLM Guardrails
arXiv:2608.08485v2 Announce Type: replace Abstract: Current LLM safety guardrails face a fundamental tension: fine-tuning distorts pre-trained…
AI for Societal Impact
Explore this collection to see how experts and local leaders are using AI breakthroughs to ensure everyone can share the opportunity of AI.
CAFE: Self-Improving Search Agents Need Co-Evolving Feedback
arXiv:2608.24794v2 Announce Type: replace Abstract: Reliable search requires more than acquiring external evidence. An agent must also recognize and…
Optimizing cost and latency with Amazon Bedrock prompt caching
Prompt caching in Amazon Bedrock can cut input token costs by up to 90% when you repeatedly send the same context to foundation models. This post walks…
Protecting patient privacy in clinical foundation models: Technical and legal perspectives
arXiv:2608.07705v2 Announce Type: replace Abstract: Clinical foundation models trained on large-scale patient data are increasingly used for decision…
Know Your Agent: Reconnaissance-Driven Pentesting of AI Agents
arXiv:2607.19837v2 Announce Type: replace Abstract: Traditional pentesting uses reconnaissance at each step to uncover unseen weaknesses, build stronger…
Nunchux AI Introduces VC-Attention: A Training-Free Low-Bit Attention Kernel That Speeds Up Video Diffusion Transformers
Nunchux AI has released VC-Attention, a training-free low-bit attention kernel built for video Diffusion Transformers (DiTs). It targets 2 problems at…
The Verifier is the Curriculum: Precision Sets the Return on Search in Code Self-Distillation
arXiv:2607.09709v2 Announce Type: replace Abstract: Post-training a code generator against a learned judge can optimize proxy features that raise the…