yorwba

19 ms·

391.
by yorwba 6mo ago
Vietnam hardly has any malaria cases anymore: https://ourworldindata.org/grapher/incidence-of-malaria?tab=...
392.
by yorwba 6mo ago
You would probably have to completely rewrite its internals.
393.
by yorwba 6mo ago
Happy Horse is a model that has been announced, but not released to the general public except for testing through Arena, which makes it a feast for copycats looking to scam people. https://xcancel.com/HappyHorseATH
394.
by yorwba 6mo ago
You can build an unlimited-order Markov chain by, instead of pre-computing a table of counts for all possible contexts, using a substring-search index on the training data to count possible continuations on the fly: https://arxiv
395.
by yorwba 6mo ago
It's a good-per-byte-but-not-in-absolute-terms quantization of Qwen3-8B that's comparable in accuracy to Qwen3.5-4B at 4-bit quantization (which makes the 4B model larger in terms of storage, though the lower number of parameters
396.
by yorwba 6mo ago
The 1-bit Bonsai and Ternary Bonsai models are all based on the corresponding Qwen3 model: https://raw.githubusercontent.com/PrismML-Eng/Bonsai-demo/re... (page 4)
397.
by yorwba 6mo ago
Rather than letting the model generate arbitrary code and type-checking it afterward, the author wants to pre-restrict the output with templates that are well-typed by construction and only let the model make choices between valid alternati
398.
by yorwba 6mo ago
Because there's pre-existing demand for lottery tickets that provides some plausible deniability for the money-laundering use case. If prediction markets were primarily used to buy insurance against hard-to-avoid expensive events, that
399.
by yorwba 6mo ago
Yes, that's why insider trading is illegal.
400.
by yorwba 6mo ago
"enforcing Temperature=1.0 and TopP=0.95 in Thinking mode, with mandatory validation that thinking content is correctly passed back."
401.
by yorwba 6mo ago
Anthropic was already special-casing case-folding in their tokenizers before this recent change: https://transformer-circuits.pub/2025/attribution-graphs/met... "The tokenizer the model was trained with uses
402.
by yorwba 6mo ago
Don't worry, they'll also use AI to add more pictures, which will all look strangely similar while bearing at best passing resemblance to anything you might actually receive after placing an order.
403.
by yorwba 6mo ago
It's an LLM account, as you can tell from the hallucinated comment in the thread on Japanese penguins: https://news.ycombinator.com/item?id=47816256
404.
by yorwba 6mo ago
Small LLMs are certainly restricted, but they're not reliably restricted in a way that aligns with educational goals. Even Qwen3.5-0.8B, the smallest, dumbest LLM I have lying around, can solve quadratic equations. In math class, learn
405.
by yorwba 6mo ago
So, scientific calculators - made tasks easy that were a necessary prerequisite for advanced math (basic arithmetic), but not what the lesson was supposed to be about - could in theory also let students skip over what they were supposed to
406.
by yorwba 6mo ago
Well, Japan isn't much different in terms of the share of income that goes to housing: https://housingpolicytoolkit.oecd.org/2.H_conso.html
407.
by yorwba 6mo ago
Here's a sample-size calculator that may help illustrate the issue: https://sample-size.net/sample-size-proportions/ Put in the benchmark score of one model as p₀ and of the other model as p₁ (as a fraction betwee
408.
by yorwba 7mo ago
Creating an alias is not the same as putting an executable in your $PATH.
409.
by yorwba 7mo ago
Yes, MarginLab only tests 50 tasks a day, which is too few to give a narrower confidence interval. On the other hand, this really calls into question claims of performance degradation that are based on less intensive use than that. Variance
410.
by yorwba 7mo ago
1. Take a large collection of text, e.g. from https://opus.nlpl.eu/corpora-search/zh-CN&en 2. Split into sentences and tokenize sentences into words, e.g. using https://github.com/fxsjy/jieba
411.
by yorwba 7mo ago
> A guy on a forum had hired a calligrapher to write three thousand characters in ballpoint pen A shame that this amazing resource is not linked.
412.
by yorwba 7mo ago
https://www.canirun.ai can help you find a model that'll run on your hardware.
413.
by yorwba 7mo ago
I'm pretty sure this will just call itself in a loop. You need to use the absolute path to the wrapped binary to distinguish it from the wrapper.
414.
by yorwba 7mo ago
If all models are trained on the benchmark data, you cannot extrapolate the benchmark scores to performance on unseen data, but the ranking of different models still tells you something. A model that solves 95/98 benchmark problems may
415.
by yorwba 7mo ago
The previous major war was in Afghanistan with 150,000 German soldiers: https://www.dw.com/en/germany-honors-soldiers-who-fought-in-...
416.
by yorwba 7mo ago
I think you took away the wrong lesson from that podcast: I think there is work to be done on scaffolding the models better. This exponential right now reminds me of the exponential from CPU speeds going up until let’s say 2000 or somethin
417.
by yorwba 7mo ago
The suspension doesn't change whether you get drafted or not, it just reduces peacetime bureaucracy at the expense of making a future draft more chaotic if it does happen.
418.
by yorwba 7mo ago
It's massive in the sense of people hyping it on social media and grifters trying to profit from it. Pure FOMO, not dissimilar from the earlier "earning a side income using ChatGPT" hype. I doubt there are many people using i
419.
by yorwba 7mo ago
Yes, typically the tags used for tool calls get their own special tokens, e.g. https://huggingface.co/google/gemma-4-E4B-it/blob/main/token...
420.
by yorwba 7mo ago
Each text token already represents the activation of certain neurons. There is nothing "more direct." And you cannot fully separate data and metadata if you want them to influence the output. At best you can clearly distinguish th

More ›