Wrote a literature review on various automated evals for measuring linguistic diversity in LLM generated synthetic data. Useful to systematically test impact of various techniques on diversity Link in first reply.

Based on surveys from @ChantalShaib et al. & @yanzhu_guo et al. 🙏

@prasann_pandya @abbe_hyde Malta also had this; gives you an address in EU/US/UK and just forwards the item to your address in Malta. Charged by weight + customs if ordered from non-EU. You could even have it delivered to nearest locker and collect whenever you're free.

@aarongraves @max_paperclips it's also backed by science 😂

@NirantK @abhi_agarwal4 Social blade should help:

@sergeykarayev Diarium does this. It shows you your journal entry along with photos taken that date, sleep/steps (via google fit), calendar events, google timeline integration etc. Has lots of integrations with third party apps

@mitsuhiko @rustlang @huggingface @OpenAI @pydantic @astral_sh @pinecone @qdrant_engine Do you find it easier compared to Cython? ML libraries conventionally leveraged cython (@scikit_learn, @spacy_io etc.) to offload intensive parts, so this move to Rust is quite interesting to see

@levelsio For Chrome's search box, this extension works for me: Once installed, on search box, type gpt > tab > query

@nishparadox @bg_learning @GoogleColab @code I have copilot enabled through the GitHub education pack. Pretty useful and surprises me sometimes with exact suggestions of what I was thinking to do next. But if I am already in the flow, I usually type faster than the time it takes to show me the completion.

@Towzeur @overleaf It's not possible currently. Overleaf doesn't provide an API for that

@eugeneyan What heuristic do you use to parse only the article part in the HTML document?

@nishparadox Yeah, tinkering with this approach for fun. I agree, API would be cleanest solution and is easy to reverse engineer. My only concern is legality of that. I think browser automation wouldn't violate TOS. Also, possible to hide tab and render directly ↓

Behind the scenes, @overleaf provides an endpoint to access latex zip by url but arxiv exports a tar.gz So, I built a @gradio app in the @huggingface spaces to host tar.gz-to-zip conversion. Every gradio space exposes an API endpoint, so you can use it like a serverless function