#AI News

1721 articles

Eseandre Mordi Eseandre Mordi Industry
10 minutes ago

MIT Study Questions How LLMs Learn Reasoning: 2 Key Findings That Could Cut Training Costs

New MIT CSAIL research explores how large language models build reasoning skills - and whether massive human-text datasets are really necessary.

Usman Salis Usman Salis Industry
3 hours ago

ACT Trains AI Agents to Judge Their Own Decisions - With 5-Point Accuracy Gains

A new reinforcement learning framework teaches AI agents to evaluate their own choices, not just copy expert behavior - delivering measurable gains over existing baselines.

Victoria Bazir Victoria Bazir Startups
3 hours ago

Anthropic Researchers Launch $1B Startup Mirendil to Speed Up Scientific Discovery

Ex-Anthropic team raises $175M for Mirendil, an AI startup targeting breakthroughs in biology and materials science.

Eseandre Mordi Eseandre Mordi Industry
3 hours ago

AI Humanoid Robots Move Toward Homes as Reflex Shows 80 kg Robot Handling Daily Tasks

Reflex demo shows an 80 kg humanoid robot cooking, cleaning, and folding clothes inside real homes - teleoperation bridges the gap to full autonomy.

Marina Lyubimova Marina Lyubimova Industry
22 hours ago

OpenRAG Launch: Langflow Packages Full RAG Pipeline Into 1 Command

Langflow open-sources OpenRAG, an all-in-one RAG platform with semantic search, doc ingestion, and AI chat in one deployable package.

Usman Salis Usman Salis Industry
22 hours ago

Meta Study: Reasoning LLM Judges Reach 89.6% Score, Yet Reward Hacking Persists

Meta and Yale researchers show reasoning-based judges cut reward hacking in RL alignment pipelines, but challenges remain in non-verifiable training settings.

Peter Smith Peter Smith Industry
23 hours ago

AI Research Warns of "Artificial Hivemind" Across 70+ Language Models

Study of 70+ AI models finds they give nearly identical answers to open-ended prompts, raising concerns about systemic bias.

Usman Salis Usman Salis AI models
23 hours ago

NVIDIA's AutoGaze Cuts 99% of Video Tokens, Speeds Up AI Models 19x

Researchers from NVIDIA, MIT, UC Berkeley, and Clarifai built AutoGaze - an AI filter that slashes video data by up to 99% without hurting model accuracy.

Saad Ullah Saad Ullah Benchmarks
1 day ago

GENIUS Benchmark Tests 12 AI Models - and Most Fall Short on Reasoning

A new multimodal benchmark reveals that most leading AI systems struggle with context-driven reasoning, relying on memorization rather than adaptive thinking.

Saad Ullah Saad Ullah AI models
1 day ago

Lightpanda Browser Claims 11x Faster AI Automation with 9x Less Memory

Lightpanda open-source headless browser runs 11x faster than Chrome with 9x less memory for AI agents.

Peter Smith Peter Smith AI models
1 day ago

Nvidia Nemotron Super 120B Achieves 2.2x Speed Gains Over Rivals

Nvidia Nemotron Super 120B delivers 2.2x faster inference with 1M-token context and hybrid MoE architecture.

Marina Lyubimova Marina Lyubimova AI models
1 day ago

GPT-5.4 Scores 77.7% on WeirdML - Trailing Claude Opus 4.6 by a Thin Margin

GPT-5.4 scores 77.7% on WeirdML, placing just behind Claude Opus 4.6. Results reveal how cost, token use, and code efficiency now define AI model rankings.

Saad Ullah Saad Ullah Industry
1 day ago

MiroFish Swarm Engine Surges Past 17,000 GitHub Stars

MiroFish swarm AI engine hits 17K GitHub stars, simulating thousands of autonomous agents for market and sentiment modeling.

Usman Salis Usman Salis Industry
2 days ago

OpenClaw-RL: 145K Stars Framework Trains AI Agents from Live Conversations

OpenClaw-RL trains AI agents from real-world interactions using async RL across 5 agent types.

Saad Ullah Saad Ullah Industry
2 days ago

SiliconMind-V1: Multi-Agent AI Cuts RTL Design Errors With Fewer Training Resources

SiliconMind-V1 is a multi-agent AI that generates, tests, and debugs Verilog code for chip design with fewer training resources.

Eseandre Mordi Eseandre Mordi Industry
2 days ago

NVIDIA Launches Nemotron 3 Super: 120B-Parameter Open AI Model With 1M-Token Context

NVIDIA unveils Nemotron 3 Super, a 120B-parameter open AI model with 1M-token context for efficient multi-agent systems.

Alex Dudov Alex Dudov AI models
2 days ago

FlashAttention-4 Delivers 2.7x Speed Gains and 71% GPU Utilization on Nvidia Blackwell

FlashAttention-4 hits 71% GPU utilization and 2.7x speed gains on Nvidia Blackwell B200s.

Peter Smith Peter Smith AI models
2 days ago

AI Tool Verification Boosts Qwen and Llama Accuracy by 31.6% on Hard Math Benchmarks

Stanford and University of Munich researchers unveiled a framework that uses external code checks to verify AI reasoning, lifting model accuracy by up to 31.6% on AIME, AMC, and MATH-500 tests.

Marina Lyubimova Marina Lyubimova Industry
3 days ago

Google Drives 21x More Traffic, But AI Visitors Spend 20% More

AI discovery traffic converts at a higher value than Google search, with average LTV reaching $325 vs. $271 - despite a 37x gap in traffic share.

Usman Salis Usman Salis AI models
3 days ago

xAI Grok 4.20 API Launches With 2M Token Context and 3 Model Variants

xAI quietly launched the Grok 4.20 API with three model variants and a 2-million token context window, opening large-context AI to developers at tiered pricing.

Posts feed
ACT Trains AI Agents to Judge Their Own Decisions - With 5-Point Accuracy Gains
3 hours ago
Anthropic Researchers Launch $1B Startup Mirendil to Speed Up Scientific Discovery
3 hours ago
AI Humanoid Robots Move Toward Homes as Reflex Shows 80 kg Robot Handling Daily Tasks
3 hours ago
Claude Opus 4.6 Scores 78.3% on MRCR v2 With 1M Token Context Window
3 hours ago
Grok 4.20-Beta1 Claims #2 Spot on Search Arena With Score of 1225
4 hours ago
New JTok Scaling Method Boosts LLM Efficiency With 35% Less Compute
5 hours ago
ReMix Fixes LoRA Routing Collapse with Reinforcement Learning
19 hours ago
Cursor Benchmarks GPT-5.4 Coding Efficiency: 60%+ Score on CursorBench
19 hours ago
OpenRAG Launch: Langflow Packages Full RAG Pipeline Into 1 Command
22 hours ago
Meta Study: Reasoning LLM Judges Reach 89.6% Score, Yet Reward Hacking Persists
22 hours ago
Follow us on Twitter

Disclaimer: Aigazine provides content for informational and educational purposes only. We do not offer financial, investment, or legal advice. Any decisions made based on the information published on this website are solely at the reader’s own discretion and risk. We encourage all readers to conduct their own research and seek professional guidance when necessary.

Aigazine is a news platform focused on artificial intelligence, covering global AI trends, technology, and innovation. Aigazine is based in Tbilisi (0179, Georgia, Tbilisi City, Vake District, 49 Besarion Zhghenti Street, VAT 305786600).

© 2025 aigazine.com

We use cookies to improve your experience on our site and to show you relevant advertising. To find our more, read our privacy policy and cookie policy