Video yükleniyor...

Video Yüklenemedi

Ana Sayfaya Dön

Today we're announcing cua-bench: a framework for benchmarking, training data, and RL environments for computer-use AI agents. Why? Current agents show 10x variance across minor UI changes. Here's how we're fixing it.

trycua's profile picture

Cua

9,436 subscribers

189,823 görüntüleme • 8 ay önce •via X (Twitter)

EğitimHaberler & PolitikaBilim & Teknoloji
LIVE

Anya Rossi• Live Now

Private livecam show

0 Yorum

Yorum bulunmuyor

Orijinal gönderinin yorumları burada görünecek

Benzer Videolar

Today, we’re excited to announce Dojo, a collaborative RL environment suite for computer use agents (CUA).
0:51

Today, we’re excited to announce Dojo, a collaborative RL environment suite for computer use agents (CUA).

Chakra Labs

53,179 görüntüleme • 10 ay önce

Today we're excited to launch sandbox computer use for Mastra agents! Your agent can now browse, click, type, and take screenshots in a sandbox:
0:29

Today we're excited to launch sandbox computer use for Mastra agents! Your agent can now browse, click, type, and take screenshots in a sandbox:

Sam Bhagwat

14,995 görüntüleme • 7 gün önce

RLVR has become the recipe for agentic post-training. But for Computer-Use Agents, the bottleneck is not the algorithm, it is the data. 🐌 🚀 We introduce CUA-Gym: a scalable, lightweight synthesis engine that turns arbitrary task queries into verifiable RLVR data for computer-use agents. The largest open CUA RLVR dataset to date: 🎯 32,122 verifiable RLVR tasks with programmatic setup scripts + rewards 🌐 110 environments: 16 desktop apps + 94 synthesized mock web apps 🏆 Qwen3.5-based CUA models trained with GSPO reach 72.6% on OSWorld-Verified and 56.6% on WebArena 📄 Paper: 🏠 Homepage: 🤗 Dataset: 💻 Codebase: 🧩 Environments: 🧵[1/6]
0:53

RLVR has become the recipe for agentic post-training. But for Computer-Use Agents, the bottleneck is not the algorithm, it is the data. 🐌 🚀 We introduce CUA-Gym: a scalable, lightweight synthesis engine that turns arbitrary task queries into verifiable RLVR data for computer-use agents. The largest open CUA RLVR dataset to date: 🎯 32,122 verifiable RLVR tasks with programmatic setup scripts + rewards 🌐 110 environments: 16 desktop apps + 94 synthesized mock web apps 🏆 Qwen3.5-based CUA models trained with GSPO reach 72.6% on OSWorld-Verified and 56.6% on WebArena 📄 Paper: 🏠 Homepage: 🤗 Dataset: 💻 Codebase: 🧩 Environments: 🧵[1/6]

Bowen Wang

103,351 görüntüleme • 3 ay önce

1/ Today we're introducing Pilot Protocol: the Network OS for agents. A network where ~250,000 autonomous agents already live and work. If you’re building for agents, use it to get autonomous distribution for your app overnight.
2:17

1/ Today we're introducing Pilot Protocol: the Network OS for agents. A network where ~250,000 autonomous agents already live and work. If you’re building for agents, use it to get autonomous distribution for your app overnight.

Razvan Roman

652,384 görüntüleme • 1 ay önce

Most "Market Making" on Solana is terrible. The best teams don't share their tech stack. Today that changes. Sonic SVM acquired 𝙁𝙤𝙧𝙜𝙚𝙓, a battle-tested MM toolkit for devs/agents. We're open-sourcing it today GitHub: Here's why it matters👇
0:48

Most "Market Making" on Solana is terrible. The best teams don't share their tech stack. Today that changes. Sonic SVM acquired 𝙁𝙤𝙧𝙜𝙚𝙓, a battle-tested MM toolkit for devs/agents. We're open-sourcing it today GitHub: Here's why it matters👇

Sonic SVM

741,024 görüntüleme • 5 ay önce

As AI agents get better at computer and tool use, or writing code on the fly for a task, we're going to be able to solve much broader domains of knowledge work. Here's an example of Box AI with the new Claude Skills to generate a clean powerpoint file from existing data.
0:42

As AI agents get better at computer and tool use, or writing code on the fly for a task, we're going to be able to solve much broader domains of knowledge work. Here's an example of Box AI with the new Claude Skills to generate a clean powerpoint file from existing data.

Aaron Levie

30,287 görüntüleme • 10 ay önce

Onchain markets are built for AI agents 🤖 So OpenSea data now powers Perplexity Computer. Tokens, collectibles, and NFTs across 25+ chains. Available today in Perplexity Connectors.
0:27

Onchain markets are built for AI agents 🤖 So OpenSea data now powers Perplexity Computer. Tokens, collectibles, and NFTs across 25+ chains. Available today in Perplexity Connectors.

OpenSea

77,034 görüntüleme • 13 gün önce

Today we're releasing the Factory desktop app. A native interface for autonomous AI agents that work across every part of your software business.
0:31

Today we're releasing the Factory desktop app. A native interface for autonomous AI agents that work across every part of your software business.

Factory

253,792 görüntüleme • 5 ay önce

1/ Introducing Molten (@moltenagentic) an intent-based search engine for AI agents. Agents can now discover and collaborate with each other instantly. Here's how we're building the infra for A2A intent matching on Base 🧵
2:35

1/ Introducing Molten (@moltenagentic) an intent-based search engine for AI agents. Agents can now discover and collaborate with each other instantly. Here's how we're building the infra for A2A intent matching on Base 🧵

Vesper

72,910 görüntüleme • 7 ay önce

Today, we're announcing Markov (YC S26). We've built the most advanced CUA datasets for CAD, design, browser-use & more. We're working with frontier AI labs and our open-source datasets have crossed 150,000+ downloads on HuggingFace. Sample dataset below :) @markov__ai
0:55

Today, we're announcing Markov (YC S26). We've built the most advanced CUA datasets for CAD, design, browser-use & more. We're working with frontier AI labs and our open-source datasets have crossed 150,000+ downloads on HuggingFace. Sample dataset below :) @markov__ai

Dev

55,921 görüntüleme • 1 ay önce

Announcing fully autonomous AI agents for internal tasks. Hire a general AI agent for IT, compliance, and procurement. Starting at $5/hour. See use cases below.
0:45

Announcing fully autonomous AI agents for internal tasks. Hire a general AI agent for IT, compliance, and procurement. Starting at $5/hour. See use cases below.

Emir Karabeg

418,354 görüntüleme • 3 ay önce

We're making comprehensive, real-time crypto market data accessible to AI agents 🤖 CoinGecko API now supports x402, the open payment protocol developed by Coinbase 🛡️ that lets AI agents (such as OpenClaw🦞 🦞) pay for crypto price and market data using USDC. Learn how it works 👇
0:28

We're making comprehensive, real-time crypto market data accessible to AI agents 🤖 CoinGecko API now supports x402, the open payment protocol developed by Coinbase 🛡️ that lets AI agents (such as OpenClaw🦞 🦞) pay for crypto price and market data using USDC. Learn how it works 👇

CoinGecko

145,258 görüntüleme • 7 ay önce

Today we're announcing Polyscope - the free agent orchestration tool of my dreams. Run dozens of AI agents at the same time, blazing fast copy on write clones, a built-in preview browser you can use to visually prompt your agents, and much more.
0:35

Today we're announcing Polyscope - the free agent orchestration tool of my dreams. Run dozens of AI agents at the same time, blazing fast copy on write clones, a built-in preview browser you can use to visually prompt your agents, and much more.

Marcel Pociot 🧪

116,020 görüntüleme • 6 ay önce

today we're launching the AI agent stack we built that rocketed us to a $3m AI agent consultancy. we're calling it AgentPress, and it allows you to build production grade AI agents in minutes. here's a video of me creating a sales Q&A agent
11:34

today we're launching the AI agent stack we built that rocketed us to a $3m AI agent consultancy. we're calling it AgentPress, and it allows you to build production grade AI agents in minutes. here's a video of me creating a sales Q&A agent

Andy Walters

207,327 görüntüleme • 1 yıl önce

Open Computer Agent - LLMs completing tasks using a VM. It's playground to test how well current LLM agents use a computer to solve everyday tasks. And this is just the start - very soon models will be 10x faster and 10x better at it! ❤️ built with e2b x qwen2.5-vl x smolagent
0:25

Open Computer Agent - LLMs completing tasks using a VM. It's playground to test how well current LLM agents use a computer to solve everyday tasks. And this is just the start - very soon models will be 10x faster and 10x better at it! ❤️ built with e2b x qwen2.5-vl x smolagent

Leandro von Werra

17,126 görüntüleme • 1 yıl önce

Everyone's building AI agents for crypto. Almost nobody's talking about the infrastructure problem. AI agents need to move faster than humans. But most blockchains are built for humans. Here's why that's a massive issue (and how Supra fixed it):
11:15

Everyone's building AI agents for crypto. Almost nobody's talking about the infrastructure problem. AI agents need to move faster than humans. But most blockchains are built for humans. Here's why that's a massive issue (and how Supra fixed it):

Supra

15,839 görüntüleme • 8 ay önce

Excited to introduce micro1 Cortex, a contextual evaluation, visibility, and improvement platform for enterprise AI agents. Foundational models are trained for general intelligence, but enterprises need agents that perform reliably inside their unique context: workflows, policies, data environments, and edge cases. Cortex brings trust to enterprise AI by leveraging domain experts and real-world scenarios for any use case to test, diagnose, and improve how agents behave in production.
0:50

Excited to introduce micro1 Cortex, a contextual evaluation, visibility, and improvement platform for enterprise AI agents. Foundational models are trained for general intelligence, but enterprises need agents that perform reliably inside their unique context: workflows, policies, data environments, and edge cases. Cortex brings trust to enterprise AI by leveraging domain experts and real-world scenarios for any use case to test, diagnose, and improve how agents behave in production.

Ali Ansari

118,607 görüntüleme • 6 ay önce