stared

36 ms·

781.
by stared 2y ago
It is up for testing, but you likely get the effect of "don't think about a pink elephant." So I guess that for most embedding models, "articles about San Francisco that don't mention cars" are closest to artic
782.
by stared 2y ago
It's nice to hear that! And from this thread, it is not us only two—otherwise, the title wouldn't have resonated with the Hacker News community. This blog post stemmed from my frustration that people use cosine distance without a
783.
by stared 2y ago
If you ask, "Is '5+5' similar to '10'?" it depends on which notion of similarity you have - there are multiple differences: different symbols, one is an expression, the other is just a number. But if you ask, &
784.
by stared 2y ago
Technically speaking, cross encoders are LLMs - they use the last layer to predict similarity (a single number) rather than the probability of the next token. They are faster than generative models only if they are simpler - otherwise, ther
785.
by stared 2y ago
According to text-embedding-3-large, "french press" and "espresso" have a cosine similarity of 0.506. More than half, so we are fine, right? Right?
786.
by stared 2y ago
Well, "Is {sentence_a} similar to {sentence_b}?" is the correct query when we care about some vague similarity of statements. In this case, we should go with something in the line "Is {answer} a plausible answer to the questi
787.
by stared 2y ago
Thx, fixed!
788.
Don't use cosine similarity carelessly(p.migdal.pl)
439 points by stared 2y ago|96 comments
789.
by stared 2y ago
The illustrated guide (which I had known during my PhD) focuses on progress—however, niche progress. Yet, PhD students' life experiences are usually not centered around progress but frustration, disillusionment, and depression. If you
790.
by stared 2y ago
I couldn't believe it, so I tried myself with GPT-4o, and the result was the same. Interestingly, the answer is correct when asked which hand is being used in an image. For watches, it doesn't even notice when hands are incorrect.
791.
Ned Block: Why ChatGPT Struggles with Watch Faces [video](youtube.com)
1 points by stared 2y ago|1 comments
792.
Show HN: Currex, Pythonic currency calculator with real-time exchange rates(github.com)
3 points by stared 2y ago|0 comments
793.
by stared 2y ago
Thanks! Right now, it does not. I focus on a common case of a single product without any other parts. If there is a summation, it is managed via indices. If you are interested in arbitrary drawings, you are probably aware of TikZ (if you li
794.
by stared 2y ago
I had been working on a library for visualizing tensor contractions (so-called Penrose tensor diagrams), https://github.com/Quantum-Flytrap/tensor-diagrams , with an example (a tutorial stub) https://p.migdal
795.
Poland's minimum wage exceeds US federal rate for first time(notesfrompoland.com)
12 points by stared 2y ago|0 comments
796.
by stared 2y ago
There is a lot of good advice about details! Especially one that got my heart: > Some authors design their tutorials the way you’d give instructions for an origami structure. It’s a mysterious sequence of twists and folds until you get t
797.
by stared 2y ago
Location: Warsaw, Poland Remote: Yes (experienced in remote work) Willing to relocate: No (but travels for business) Technologies: Python deep learning data science stack (PyTorch, Scikit-learn, Pandas, NumPy, Keras), TypeScript (mostly for
798.
WebDev Arena: AI Battle to build the best website(web.lmarena.ai)
1 points by stared 2y ago|0 comments
799.
Began a pay-on-results coaching experiment, made $40,300 since July(chrislakin.blog)
1 points by stared 2y ago|0 comments
800.
Polish linguistic and cultural competency benchmark for LLMs(huggingface.co)
2 points by stared 2y ago|0 comments
801.
by stared 2y ago
Well, quite a few things can be described using the same mathematics - including switches, resistors, capacitors, and inductive coils. Sure, it is not everything, but well - fluid is fluid, electricity is electricity. For a small simulation
802.
by stared 2y ago
I often find myself revisiting old posts and stories. As with any human artifacts, most things aren't worth revisiting or are only meaningful in the moment. If they're gone, few people miss them. I'm a link hoarder myself (ov
803.
by stared 2y ago
I was thinking of developing something similar, but it ended up being one of the thousands of ideas that never end with a line of code. I'm glad to see it here. Visual programming is a tempting idea I love. It rarely works, but this mi
804.
Slithering liquid metal motivated by chemistry [video](youtube.com)
1 points by stared 2y ago|0 comments
805.
Moonshine – open-source, real-time speech-to-text in the browser(huggingface.co)
5 points by stared 2y ago|1 comments
806.
by stared 2y ago
How does it compare to OpenAI and Anthropic models, on the same benchmarks?
807.
by stared 2y ago
And it is reflected in numbers. The fertility rate has a non-monotonous relationship with income. It is the middle-income fertility that plummets. https://x.com/theHauer/status/1222514313723875332/photo/1
808.
OpenAI o1 pro reproduces Coinbase frontend from a screenshot(twitter.com)
3 points by stared 2y ago|1 comments
809.
by stared 2y ago
It was not my point. The point is that it is (almost) always about politics. Just the Church changed it is the position from a firm one to one for which dissidents pose political danger. I found it surprising, but in the Middle Ages, there
810.
Highly Cited Researchers – 2024 Analysis(clarivate.com)
1 points by stared 2y ago|0 comments

More ›