It is. The EOTF wasn’t an engineered function in 1996, it was the natural response of CRTs. The EOTFs of later technologies like LCDs were developed to approximate this response, not to be an inverse of the encoding gamma.
Sure but modern ICC color-managed workflows (e.g. digital photography, printing) basically don't distinguish between EOTF and OETF. Assuming your source is tagged correctly and all displays have a profile matching their true response, you necessarily need to linearize with the inverse of the encoding gamma. All edits are done directly in the source color space, if the viewer's profile differs from the source it's up the viewer to translate accordingly. You don't edit images "on the assumption" that they'll be decoded with a 2.2 gamma.
For some reason video workflows never adapted to the ICC system (probably because in CRT days you couldn't really adapt your decode gamma on the fly) which is basically where the whole debate in https://gitlab.freedesktop.org/wayland/wayland-protocols/-/m... comes from.
I'm not saying that the people using 2.2 EOTF are wrong, but all this just adds to the absurdity: in the modern day where LUTs are cheap and plentiful, instead of tagging content as an ambiguous sRGB it could simply be tagged as gamma 2.2 if it's actually intended to be decoded at that gamma.
It depends what you want to apply your linear functions on. If you want to work directly in the source scene light (e.g. photography), then it would make sense to use the inverse OETF. If you are blending graded scenes of emissive light, i.e. a movie, then using the EOTF makes sense. The reason for this is that movies are graded with that EOTF in mind, so by linearizing with that EOTF, you get a resulting linear value as it is intended to be seen by a viewer.
Regardless of what you use for a linearizing function, the more important thing is that you use the correct encoding function afterward, so that you don’t introduce any additional gamma correction. For example, it was common to use a simple squaring function for speed. This gives fairly good results as long as you apply the square root function afterward to restore the original gamma correction. It doesn’t matter if the source is 2.2 or 2.4 gamma encoded or something else, that correction will be preserved. The blending post-linearization will be less accurate, but much better than not linearizing at all.
>The reason for this is that movies are graded with that EOTF in mind, so by linearizing with that EOTF
I guess this is the part I find anachronistic. Why do we work in the source scene light for photography, but do the opposite for videos? It makes sense if you assume the viewing device is "dumb" (like a television or CRT, especially in the analog days) but by now I assume the workflows are all fully digital, and even the most basic output device can apply LUTs. When digital video container formats were introduced, why didn't they align with what ICC did? It would have saved a lot of headache for everyone, compared to limited NCLC tags and the mess around EOTFs.
I don't think the MoE part has anything to do with it, but the current gen of multimoddal models can do thinking interleaved with autoregressive(?*) image-gen so it's probably not long before they bake this into the RL process, same way native thought obviated need for "think carefully step by step" prompts.
A lot of the trickiness is that if you believe they're conscious, it's clearly not a "continuous" form of consciousness. Because the transcript by itself is just a transcript. (We don't consider novels conscious even though they're transcripts in a similar way). Either you say they're alive only when generating text, or you consider that input from environment a necessary component and so consider the entire "back/forth conversation dynamic unfolding" necessary for the consciousness.
If you sort of squint this idea does work in cases where the cost of comparison is dominated by the cost of going down a level. And that leads you to things like b-trees where fetching a page from disk is expensive but doing the comparisons within that page is basically free.
The fact that COTs often "hallucinate" was known anecdotally, but they study it more systematically here and provide ways to mitigate. Apparently SFT'ing on "meaningful" reasoning traces provides enough of a scaffold so that later RL results in meaningful/"truthful" traces rather than the appearance of reasoning. See also the author's summary at https://x.com/qinan_yu/status/2049865788304380239
Sad that the article doesn't mention wildberger (coincidentally similar last name), an (in)famous math youtuber that's been mentioned on HN several times before. He has a "rational trigonometry series" an approachable way to see how math would work in an ultrafinite setting.
I guess this is referring to https://www.youtube.com/watch?v=KXuZi9aeGTw ?
reply