In case the straight lines seem puzzling: the code intelligence values are plotted on a logarithmic scale. Is Compression Really Linear with Code Intelligence?
Paper: https://
arxiv.org/pdf/2505.11441
@jiqizhixin
-
Compression and Code Intelligence: Linear Relationship Analysis
By
–
-
Code LLM Compression: Linear Relationship with Intelligence?
By
–
Is compression really linear with code intelligence?
A research team from China dives deep into this question.
They evaluated a wide range of open-source Code LLMs across multi-language, multi-task benchmarks — and introduced Format Annealing for fair comparison. Key -
Tencent Hunyuan Game: Try AI-Powered Gaming Experience
By
–
In case if you wanna give it a try: https://
hunyuan.tencent.com/game/home -
Tencent Launches Hunyuan Game: Industrial AIGC Engine for Game Development
By
–
🎮 Tencent just launched Hunyuan Game — the first industrial-grade AIGC engine built for game production!
— 机器之心 JIQIZHIXIN (@jiqizhixin) 20 mai 2025
Powered by Tencent’s Hunyuan AI, it turns ideas into production-ready assets 10× faster ⚡
Game devs, this could change everything. pic.twitter.com/p4QkDlFVcQTencent just launched Hunyuan Game — the first industrial-grade AIGC engine built for game production! Powered by Tencent’s Hunyuan AI, it turns ideas into production-ready assets 10× faster Game devs, this could change everything.
-
Microsoft Magentic-UI and Flowith Neo: New Web Agents Launched
By
–
Today is a day of Agents. 🤖
— 机器之心 JIQIZHIXIN (@jiqizhixin) 20 mai 2025
Microsoft launched Magentic-UI, an experimental human-centered web agenthttps://t.co/qB7nGoL2bb
Flowith introduced Flowith Neo, the world #1 Infinite Agent – infinite steps, infinite context, infinite output on cloud.https://t.co/dnRQaT5Xaj… pic.twitter.com/0ZoQNDGlaQToday is a day of Agents. Microsoft launched Magentic-UI, an experimental human-centered web agent https://
microsoft.com/en-us/research
/blog/magentic-ui-an-experimental-human-centered-web-agent/
… Flowith introduced Flowith Neo, the world #1 Infinite Agent – infinite steps, infinite context, infinite output on cloud. https://
x.com/flowith_ai/sta
tus/1924453961538670755
… -
Continuous Chain-of-Thought Reasoning Outperforms Discrete Methods
By
–
Why do continuous Chain-of-Thoughts (CoTs) outperform discrete ones in complex reasoning?
This new paper might have the answer. Reasoning by Superposition: A Theoretical Perspective on Chain of Continuous Thought Paper: https://
arxiv.org/pdf/2505.12514
v1
… -
Which TTS speaks Chinese best? Audio Turing Test
By
–
Which TTS speaks Chinese best? Audio Turing Test: Benchmarking the Human-likeness of Large Language Model-based Text-to-Speech Systems in Chinese Paper: https://
arxiv.org/pdf/2505.11200 -
Physical AI Foundation of Robotics Revolution
By
–
"Physical AI is the foundation of the robotics revolution," says Jensen Huang. The future isn’t just digital — it’s embodied.
-
Test-Time Compute: Why Thinking Time Matters in AI
By
–
Lilian Weng just dropped a new blog post: Why We Think. It explores recent advances in using test-time compute (aka 'thinking time')—and why it matters.
-
CVPR 2025 Paper Sharing Session Beijing June 8
By
–
Join us in Beijing on June 8 (09:00–17:30, UTC+8) for the CVPR 2025 Paper Sharing Session! Keynotes | Paper talks | Roundtables | Posters | Corporate booths Topics: Multimodality, Generation, and more. Meet top experts & authors for a full day of