Evaluations, or “evals” refer to the execution of datasets for custom use cases against foundational models.
They are often used after training or fine tuning to determine how good a model at specific tasks.
We are starting an educational series on evals for LLMs.
The terms "benchmarking", "evals" and "testing" are often conflated with one another. Today, we are doing a deep dive into benchmarking. What follows is a 🧵on what benchmarking is and why it is important.
In line with our private betas for the Atlas Application, we have updated our website to reflect our vision.
Learn more about LayerLens, what we are building, and the long term roadmap here: https://t.co/1nthh30uXq
If you are a developer, you can also book time directly with one of the founders for a private beta!
We can accelerate safety.
Earlier this month, @leopoldasch published SITUATIONAL AWARENESS, a series calling for the nationalization of AI Labs and AGI efforts.
Today, I am publishing my response, Rational Accelrationism, calling for AGI development to remain in our hands.
Also, I technically published this in an effort to get deep into the strategic realities underpinning AI, which is a side-project that I have been pursuing as part of BuildSpace S5. Look for more technical-oriented releases as part of S5 soon.
CC @_buildspace@_nightsweekends
Rational Accelrationism is a multi-part series arguing that we can be trusted with AGI; the reality is that today's models are far away from simulating human cognition, and that winning governments prioritize free-markets over regulation.
Really cool examples of the potential of autonomous agents in the @faktoryai platform. Herman uses connectivity to MySQL and Gmail, planning skills and a helpful personality https://t.co/gFB1yWq2hi #generativeAI#LLMs
Excited to announce that @lituusfinance, a project spun out of a microgrant given to developers @polyblockchain from @AdamniteLabs , has officially started building! Be sure to give them a follow and check out their resources!
There was an attempt to hijack the Adamnite Discord Server today. As always, be sure to enable two-factor authentication and do not click on any malicious links!
Excited to announce that @beryl_finance will be building a decentralized freelancing and rewards platform on #Adamnite. They have also received a microgrant from @AdamniteLabs! Be sure to follow them, and join their Discord here: https://t.co/ttzXKLr9yQ
#Web3#blockchain
We are looking for smart contract developers who are interested in creating new projects/trying out a new programming language to try out A1 and create small demos! If you have an idea, join our Discord: https://t.co/Zr13JSQlMZ
Anyone on #CryptoTwitter interested in learning how to build a dApp? A1, Adamnite's own smart contract programming language, is now live and open-source. Check it out here: https://t.co/icOxtS6Rpp
New Blog Post! Check it out here: https://t.co/iwkMD2MFdl
Also, be sure to follow @AdamniteLabs for continuous updates on Passport and other infrastructure!