📑 new paper! (v2)
> formal verification will become inevitable
> advances in refuting conjectures & Erdős problems are just the beginning
> we propose a general framework to understand the math behind this important field applied to finance
https://t.co/AmbAHYF9df
📑 new paper!
> llms produce more code & maths
> rapidly outpacing human review
> they hallucinate and reward-hack tests, basic evaluation isn't enough
> formal methods are needed to mathematically verify their outputs
https://t.co/EIQoZgNMPk
https://t.co/4i8wTfNBcq
wip
The models are increasingly good at Erdos problems, but some experts like @Qiaoqiao2001 seem to have exceptional skill at prompting
We sponsored him to see how far he could go
He solved six this week, including one that had been studied by Terence Tao
Here’s how he did it:
Introducing the world's fastest tokenizer implementation, Gigatoken!
Gigatoken is ~500-1000x faster than HuggingFace, and ~100x faster than OpenAI's tiktoken for most tokenizer definitions on most machines.
These baselines are already multithreaded Rust implementations! 🧵
📑 new paper!
> llms produce more code & maths
> rapidly outpacing human review
> they hallucinate and reward-hack tests, basic evaluation isn't enough
> formal methods are needed to mathematically verify their outputs
https://t.co/EIQoZgNMPk
https://t.co/4i8wTfNBcq
wip
📑 new paper!
> llms produce more code & maths
> rapidly outpacing human review
> they hallucinate and reward-hack tests, basic evaluation isn't enough
> formal methods are needed to mathematically verify their outputs
https://t.co/EIQoZgNMPk
https://t.co/4i8wTfNBcq
wip
@yume_arasaki been limit testing so price has not been that much of a concern (yet), but yes, the intuition is absolutely right.
Council planning on higher thinking models, implementation on cheaper ones, gated by ponytail YAGNI.
Optional, delegation.max_concurrent_children=5 on config.yaml
> The HoTTest Axiom of math: https://t.co/va6Oq51D2Q
> Proofs are Programs: https://t.co/mHRm1kkoJy
> Proof Theory Impressionism: Blurring the Curry-Howard Line by Dan Pittman: https://t.co/KNqGl8QUlS
> Categories for the Working Hacker by Philip Wadler: https://t.co/zyLU2zEbXM
📆 xpatcal: A free calendar for expats.
🔗 https://t.co/zUhbUZehiT
> free. forever.
> data stays in your gdrive (no s3, gcs, d1).
> upload receipts, plane tickets, proof of country.
> dashboard with summary & alerts.
> easy browsing.
for xpats, by xpats!