Model performance comparisons are tricky without knowing exact sizes; it could be an apples-to-oranges case. Plus, you can always push numbers with inference scaling. Normalizing by tokens/sec would be more meaningful.
MACHINE LEARNING
-
AI Transforms Customer Experience, Travel Industry Leads
By
–
AI Is Transforming Customer Experience. Travel Could Be The First Stop
#AI #AIio #BigData #ML #NLU #Futureofwork @HaroldSinnott @fogoros @iainljbrown @NandoDF @katecrawford @drhassanrashidi @YuHelenYu -
Graph Neural Networks: Message Passing and Aggregation Explained
By
–
What are GNNs? Graph Neural Networks work with graph data, where relationships between nodes matter as much as the data itself. How it works: Message Passing Aggregation Feature Update Learn more https://
spr.ly/6019Ie3B1 -
Kimi AI beats DeepSeek V3 in multimodal reasoning comparison
By
–
Results: • Kimi AI – 5 points.
• DeepSeek V3 – 1 point. Kimi excels in multimodal reasoning, advanced image analysis, and chain of thought. It can analyze up to 50 PDFs with a 200k context window. DeepSeek is good at content creation and structured output, but the source and -
Kimi AI excels in image analysis over DeepSeek-V3
By
–
2/ Image Analysis
— God of Prompt (@godofprompt) 19 février 2025
I asked both to analyse the picture
Here's what happened:
• Kimi AI: Kimi efficiently finds and understands information, analyzing stock prices deeply and from various angles.
• DeepSeek-V3: Couldn’t analyze the image.
Winner: Kimi AI (2-0) pic.twitter.com/6M7khINcg52/ Image Analysis I asked both to analyse the picture Here's what happened: • Kimi AI: Kimi efficiently finds and understands information, analyzing stock prices deeply and from various angles. • DeepSeek-V3: Couldn’t analyze the image. Winner: Kimi AI (2-0)
-
Research@ 2025: Sydney Team Shares AI and Language Research
By
–
That's a wrap on our first Research@ 2025! Amazing research shared by our Sydney team: Aboriginal English Voices, Multilinguality, Hearing, Operations Research, & Open Models. Thank you to all who participated.
-
New Coding Benchmark Released, But Uses Older Sonnet Version
By
–
Cool new coding benchmark! I always love to see new evals out in the world. Note though that the testing here is on the June version of Sonnet, not the latest version, so technically not "current frontier models."
-
Lilian Weng announces Thinking Machines Lab AI research startup
By
–
This is something we have been cooking together for a few months and I'm very excited to announce it today. Thinking Machines Lab is my next adventure and I'm feeling very proud and lucky to start it with a group of talented colleagues. Learn more about our vision at thinkingmachines.ai/ Thinking Machines (@thinkymachines) Today, we are excited to announce Thinking Machines Lab (thinkingmachines.ai/), an artificial intelligence research and product company. We are scientists, engineers, and builders behind some of the most widely used AI products and libraries, including ChatGPT, Character.ai, PyTorch, and Mistral. Our mission is to make artificial intelligence work for you by building a future where everyone has access to the knowledge and tools to make AI serve their unique needs. We are committed to open science through publications and code releases, while focusing on human-AI collaboration that serves diverse domains. Our approach embraces co-design of research and products to enable learning from real-world deployment and rapid iteration. This work requires three core foundations: state-of-the-art model intelligence, high-quality infrastructure, and advanced multimodal capabilities. We are committed to building models at the frontier of capabilities to deliver on this promise. If you’re interested in joining our team, consider applying here: 6wajk07p.paperform.co/ — https://nitter.net/thinkymachines/status/1891919141151572094#m
→ View original post on X — @lilianweng, 2025-02-18 18:48 UTC
-
Open-Source AI Benchmark for Software Engineering Assessment
By
–
As AI research advances, more realistic software engineering benchmarks are critical to assess model performance and understand socioeconomic implications. To facilitate future research, we open-source a unified Docker image and a public evaluation split, SWE-Lancer Diamond.