Can robots master complex manipulation by practicing in their own AI-generated videos? Researchers from Stanford and Tsinghua introduce VLAW, a new framework designed to boost robot learning through a continuous feedback loop. The method uses a co-improvement strategy:
MULTIMODAL AI
-
Elon Musk asks users to share their best Grok Imagine prompts
By
–
Share your best Grok Imagine prompts! Most people don’t know how, so this will help them.
-
Google Releases Gemini Embedding 2
By
–
Google released a new embedding multimodal model, Gemini Embedding 2, with SOTA performance!
— 🚨 AI News | TestingCatalog (@testingcatalog) 10 mars 2026
Unimodal and multimodal 👀 https://t.co/iFhFo2CiEX pic.twitter.com/IFyttjiX83Google released a new embedding multimodal model, Gemini Embedding 2, with SOTA performance! Unimodal and multimodal
-
Google Adds Gemini AI to Docs, Sheets, Slides
By
–
Google is rolling out a new Gemini experience in Docs, Sheets, and Slides, allowing users to offload more tasks to AI.
— 🚨 AI News | TestingCatalog (@testingcatalog) 10 mars 2026
Gemini will be able to pull context from relevant sources and generate or modify the document's content.
I have big hopes on this feature 👀 https://t.co/xo0g2ngMPH pic.twitter.com/piWR9tO4Z1Google is rolling out a new Gemini experience in Docs, Sheets, and Slides, allowing users to offload more tasks to AI. Gemini will be able to pull context from relevant sources and generate or modify the document's content. I have big hopes on this feature
-
Author plans to update AI recommendations list
By
–
btw it took ~5 min to set up this demo, most of which was waiting for Gemini to generate a few Ghibli image of me, while having ChatGPT write a prompt based on what it knows about me (I’ve fed a lot of my granola transcriptions into it) https://t.co/7n2KHtZCYC
— Yohei (@yoheinakajima) 10 mars 2026btw it took ~5 min to set up this demo, most of which was waiting for Gemini to generate a few Ghibli image of me, while having ChatGPT write a prompt based on what it knows about me (I’ve fed a lot of my granola transcriptions into it)
-
Introducing arXivQA: Training retrieval agents for arXiv search
By
–
Introducing arXivQA: Training retrieval agents for arXiv search We curate a multi-hop arXiv dataset based on real queries and use rubric-as-rewards to train a Qwen model for production-grade retrieval Our latest blog details our experience training with both RLVR and rubrics
-
Free AI Agents Course: Learn Text and Audio Models
By
–
Excited to announce our first text + audio AI course. Searching for a solid entry point to learn about AI agents? Look no further. FREE for everyone. Enroll now!
-
Gemini Google Sheets Achieves 70.48% Success Rate SpreadsheetBench
By
–
While we don't have favorites, the evolution of Gemini in Google Sheets might be our most impressive yet. Gemini in Google Sheets has achieved a state-of-the-art benchmark, achieving a 70.48% success rate on the full SpreadsheetBench dataset. This performance not only exceeds
-
Gemini Google Slides Custom Visual Creative Presentation Tool
By
–
Most AI presentation tools rely on their own rigid templates and structured styles, but ask for something custom and they break. With Gemini in Google Slides, we use a custom translation between Gemini and Slides that fully leverages Gemini’s visual creative abilities while
-
World Models Learning Real Data Over Text Hallucination
By
–
World models that learn from real-world data instead of text is a fascinating direction. But 'overcoming hallucination' is a bold claim, we'll see that at scale!