Precision at scale is one of the hardest asks for LLMs today. Great work by the team in building real time self-correction to push the boundaries of accuracy and comprehensiveness.

RT @maithra_raghu: 📉 𝗪𝗵𝗮𝘁 𝗵𝗮𝗽𝗽𝗲𝗻𝗲𝗱 𝗮𝗻𝗱 𝘄𝗵𝗮𝘁 𝗰𝗵𝗮𝗻𝗴𝗲𝗱 𝗶𝗻 𝘆𝗲𝘀𝘁𝗲𝗿𝗱𝗮𝘆'𝘀 𝗙𝗲𝗱 𝗺𝗲𝗲𝘁𝗶𝗻𝗴:. 📝 𝗢𝘂𝘁𝗰𝗼𝗺𝗲𝘀:.• Rate stayed the same at 𝟰.𝟮𝟱–𝟰.𝟱%.• Projecti….

📉 𝗪𝗵𝗮𝘁 𝗵𝗮𝗽𝗽𝗲𝗻𝗲𝗱 𝗮𝗻𝗱 𝘄𝗵𝗮𝘁 𝗰𝗵𝗮𝗻𝗴𝗲𝗱 𝗶𝗻 𝘆𝗲𝘀𝘁𝗲𝗿𝗱𝗮𝘆'𝘀 𝗙𝗲𝗱 𝗺𝗲𝗲𝘁𝗶𝗻𝗴:. 📝 𝗢𝘂𝘁𝗰𝗼𝗺𝗲𝘀:.• Rate stayed the same at 𝟰.𝟮𝟱–𝟰.𝟱%.• Projecting 𝘁𝘄𝗼 𝗿𝗮𝘁𝗲 𝗰𝘂𝘁𝘀 later this year.• No change to 𝗯𝗮𝗹𝗮𝗻𝗰𝗲 𝘀𝗵𝗲𝗲𝘁 policies

Wait for 9 minutes or get it instantly and in more depth on @samaya_AI . DMs open if you want access!

Standard AI benchmarks are hitting a wall. As model capabilities grow, so do our expectations. It's no longer impressive for an AI to summarize a single document or regurgitate facts. In real-world use cases—like expert financial analysis—AI must deliver much more, e.g. 1)

🚀 Thrilled to share that @SamayaAI has raised $43.5M in funding led by @NEA to build Expert AI Agents for financial services and transform knowledge work at scale. We started Samaya in 2022 — before ChatGPT — with a belief:. 💡 AI could revolutionize sophisticated financial

Full post: We've put it to use at @samaya_AI in testing our Expert Financial Knowledge Agents, finding Samaya's Knowledge Agents substantially outperform general purpose LLMs, Grounded RAG, and more.

AI Knowledge Agents have an evaluation problem. As we continue to improve AI capabilities, differentiating "generic" from "expert" becomes increasingly important and increasingly harder. Today we're introducing "Reality Bench", a new evaluation for expert level AI. Reality

Check out the full post at

The AI traces through historical and recent sources to identify important factors, how they relate to each other, and even produces an interactive quantitative model predicting effects. We can use this to predictively answer incredibly complex questions, such as the economic

Excited to share our work on Causal World Models! Human experts have the remarkable ability to hone in on the critical, top 0.1% of information. We've been developing AI that can "see the world like a human expert", and distills top insights from the complex, noisy real world.

True Expert Level AI is hard to achieve when staying general purpose. Take this recent AI generated Tesla stock analysis:  At first, it looks impressive. But digging deeper, we see that:. (i) The financials are out of date (no details from 2024). (ii)

RT @samaya_AI: 🚀 Samaya AI is hiring 1 to 2 PhD summer research interns in Mountain View, CA!. Work on cutting-edge topics: knowledge-groun….

Discussion on AI progress today conflates "beauty" and "truth". "Beauty" is the amazing technical progress we see on challenging but carefully constructed, idealized benchmarks. "Truth" is the tangible impact AI has on the real world --- tackling ambiguity, noisiness, and.

Key takeaways from reading the DeepSeek R1 paper:. AI training matters, a lot! A lot of the focus in AI has been on the data — how to collect it, generate it, ensure high quality. The DeepSeek paper also shows the critical role in the right multi-stage training process in

🎆🎉 Happy New Year!! 🎆🎉. Grateful to have wrapped up an incredible year at @samaya_AI seeing huge growth in usage and customers! It is a privilege to work with a world-class AI team that makes this momentum possible! 🚀. Over the break, I took some time to dive into o1's.

RT @nathanbenaich: News! @modal_labs is supporting @RAAIS fellowships with $10,000 credits on their platform, alongside the $5,000 GPU cred….