nimitkalra

One method to get a better estimate is to extract the token log-probabilities of "YES" and "NO" from the final logits of the LLM and take a weighted sum [1] [2]. If the LLM is calibrated for your task, there should be roughly a ~50% chance of sampling YES (1) and ~50% chance of NO (0) — yielding 0.5.

But generally you wouldn't use a binary outcome when you can have samples that are 50/50 pass/fail. Better to use a discrete scale of 1..3 or 1..5 and specify exactly what makes a sample a 2/5 vs a 4/5, for example

You are correct to question the weaknesses of a panel. This class of methods depends on diversity through high-temperature sampling, which can lead to spurious YES/NO responses that don't generalize well and are effectively noise.

[1]: https://arxiv.org/abs/2303.16634 [2]: https://verdict.haizelabs.com/docs/concept/extractor/#token-...

There are technical quirks that make LLM judges particularly high variance, sensitive to artifacts in the prompt, and positively/negatively-skewed, as opposed to the subjectivity of human judges. These largely arise from their training distribution and post-training, and can be contained with careful calibration.

Some other known distributional biases include self-preference bias (gpt-4o prefers gpt-4o generations over claude generations for eg) and structured output/JSON-mode bias [1]. Interestingly, some models have a more positive/negative-skew than others as well. This library [2] also provides some methods for calibrating/stabilizing them.

[1]: https://verdict.haizelabs.com/docs/cookbook/distributional-b... [2]: https://github.com/haizelabs/verdict

In order to see download counts for specific packages, you'll need to navigate to:

  bintray.com/homebrew/bottles/{PACKAGE}/view#statistics

(where {PACKAGE} is installed with `brew install {PACKAGE}`)

From what I've noticed, these "transpilers" output code that is readable (the code itself is written as though a human wrote it) where as "compilers" output code that has been optimized and show effects of name mangling in the code itself, etc. Just an observation.

I think it makes sense to use a different term for this "compiler"-esque behavior. For example, I might edit the output of CoffeeScript generated Javascript whereas I wouldn't know how to modify the output of gcc.

Demo: https://wells.ee/trix-demo/

Duplicate: https://news.ycombinator.com/item?id=10081105

Risks to getting solar power (according to Project Sunroof) are something to consider.

> As with any investment, there are some risks, though a well-installed system will make most risks extremely rare. Risks include PV systems catching fire, installations leading to roof leaks, theft, obsolescence, and hail damage and/or wind damage to the solar system itself.

> Fast-growing trees can shade solar installations, reducing production over time. Utilities can change how much they charge their customers for electricity, changing the savings from solar.

The code in the GitHub repository [1] is pretty interesting to look around in.

[1] https://github.com/01org/acat