OSWORLD 2.0: Benchmarking Computer Use Agents on Long-Horizon Real-World Tasks
4:56
OSWORLD 2.0: Benchmarking Computer Use Agents on Long-Horizon Real-World Tasks
10 views3 weeks ago
YouTubePraveen Govindaraj
What do AI Benchmarks Actually Mean?! A Fast Breakdown (MMLU, SWE-bench, & More Explained)
0:58
What do AI Benchmarks Actually Mean?! A Fast Breakdown (MMLU, SWE-bench, & More Explained)
146 views2 months ago
YouTubeNeural AI Flair
OS-World: Improving LLM Agent Operating Systems!
10:04
OS-World: Improving LLM Agent Operating Systems!
8K viewsApr 13, 2024
YouTubeWorldofAI
OSWorld 2.0来了！这是一个针对长时程真实世界任务对计算机使用Agent进行基准测试新特性：🎯 108 个真实世界工作流程，每个对熟练人类来说约 1.6 小时 ⏱️⚙️ 每个任务 ~318 个工具调用 vs. OSWorld 1.0 中的 ~30 个🌍 基于真实工件和有状态用户配置文件⚡ 捕捉真实现象：动态环境、流式交互、跨源推理、隐式状态推断 等更多📊 最佳结果：Claude Opus 4.8 以 20.6% 的准确率达到最高，而 GPT-5.5 令牌效率更高得多，但停留在 13% 附近。没有人接近解决真实的计算机使用问题。
0:22
OSWorld 2.0来了！这是一个针对长时程真实世界任务对计算机使用Agent进行基准测试新特性：🎯 108 个真实世界工作流程，每个对熟练人类来说约 1.6 小时 ⏱️⚙️ 每个任务 ~318 个工具调用 vs. OSWorld 1.0 中的 ~30 个🌍 基于真实工件和有状态用户配置文件⚡ 捕捉真实现象：动态环境、流式交互、跨源推理、隐式状态推断 等更多📊 最佳结果：Claude Opus 4.8 以 20.6% 的准确率达到最高，而 GPT-5.5 令牌效率更高得多，但停留在 13% 附近。没有人接近解决真实的计算机使用问题。
2.3K views4 weeks ago
x.comMax For AI
OSWorld 2.0 Shows Where Computer-Use Agents Break Down
9:54
OSWorld 2.0 Shows Where Computer-Use Agents Break Down
1 views2 weeks ago
YouTubeNeural Trend Hub
the benchmark is called osworld-v
0:06
the benchmark is called osworld-v
6 views4 months ago
YouTubeConnor Gallic
The best AI agent finishes 1 real task in 5 📗 Read #Shorts
0:55
The best AI agent finishes 1 real task in 5 📗 Read #Shorts
2 weeks ago
YouTubePaperCutAI
1:20
For the first time AI outperformed humans at using a real desktop computer. 🖥️
79 views3 months ago
YouTubeFuturescopeAI
1:58
OpenAI Unveils GPT-5.4 — 1M-Token Context, 75 Percent on OSWorld-V, Multi-Step Agents 🧠⚡
117 views2 months ago
YouTubeVibes Uncut Media
17:52
【AI论文】OSWorld: Benchmarking Multimodal Agents for Open-Ended Tasks
743 viewsSep 3, 2024
bilibiliJJ博士a
0:35
GPT-5.4 Scored HIGHER Than Humans — What Does This Mean For You?
7.1K views3 months ago
YouTubeKeshav Krishnan
9:42
Claude Opus 4.8 vs GPT-5.5: 13 Benchmarks, One Clear Winner
1.1K views1 month ago
YouTubeProf Rino and Caroline the boss
2:24:27
GPT-5.5 Day One: Does It Beat Opus?
16K views3 months ago
YouTubeRay Fernando
5:15
GPT-5.4 Just Beat Humans at Their Own Jobs (The Benchmark Nobody Expected)
12K views3 months ago
YouTubeJulia McCoy
0:58
AI Just Beat Humans at Their Own Jobs (Real Test, Real Data)
1.4K views2 months ago
YouTubethecontextwindow
0:38
Build AI Agents That Control Your Desktop
326 views3 months ago
YouTubeGithub Signals
1:14
OpenAI GPT-5.4: +27 Points in One Release. Humans Officially Behind.
3 views3 months ago
YouTubeContentBuffer
4:58
The Upgrade Every AI Agent Has Been Waiting For, 40k stars on github
7.3K views2 months ago
YouTubeBitwise AI
See more
Static thumbnail place holder