Phemist

6 ms·

31.
by Phemist 2mo ago
I am going off the explanation in the declaude page (and related papers). But I see now anthropic mentions Aaronson's distortion-free watermarking. Random watermarking functions colour the tokens based on (small) contexts and a secret
32.
by Phemist 2mo ago
1. As watermarked text is added to the training data, watermark-related tokens will be associated more with AI outputs and thus lower quality outputs which will hasten model collapse. Especially because every provider has its own secret key
33.
by Phemist 2mo ago
I am interested however in why fine-tuning on reasoning traces of a frontier model is such an effective way of improving an (open-weight) base model. See e.g. https://huggingface.co/hesamation/Qwen3.6-35B-A3B-Claude-4.6
34.
by Phemist 2mo ago
Of course we are all guessing, but both things can be true: they don't want you to hit the cache, because cache writes are more profitable than cache reads, and they are supply limited on the compute. Pre-filling input tokens is very d
35.
by Phemist 2mo ago
Incentives to optimize cache-usage are only aligned between anthropic and you, dear user, when there is not enough compute available to serve the tokens fresh. Apparently enough compute is now available that frugal usage is no longer a requ
36.
by Phemist 2mo ago
Did you try the claude reasoning traces finetune for qwen3.6? I find that it works muuch better. I assume the same 3.8 finetune will be released at some pointas well. Edit: link - https://huggingface.co/rico03/Qwen3.6-2
37.
by Phemist 2mo ago
This kills the isolation.
38.
by Phemist 2mo ago
You let the smarter model explore the traces and figure out where the current harness' bottlenecks are for the current LLM. Then you can adjust prompts or tools to fix those.
39.
by Phemist 2mo ago
Exactly! I am not opposed to cloud-based models, but I do only stick to open-weight models because I know I can move my whole stack to local (given enough hardware) and continue development without any of the LLM interaction contracts being
40.
by Phemist 2mo ago
This would suggest they'd need to control for method by which the alzheimer's was discovered. A quick glance seems to indicate they did not do this. This is not information available I guess.. I would suspect this would eliminate
41.
by Phemist 2mo ago
What if the AI has discovered some new function F that allows it to generate (insanely large) proofs for a ton of theorems in a ton of different fields. Wouldn't you like to know more about this `F`? That seems to be the real innovatio
42.
by Phemist 2mo ago
It also explains the UK representatives for Eurovision in the last.. say 25 years (except Sam Rider in 2022). I've wondered if e.g. this year's very very naff send-in (Eins, Zwei, Drei (although, that was not an R)) was a delibera
43.
by Phemist 2mo ago
Thank you for the additional background! For me as a foreigner coming into the raffle culture, I have no such community associations with the raffles and consider the "naffness" rather more obnoxious. Once I found out that it basi
44.
by Phemist 2mo ago
Not surprising. Brits loooove lotteries. On every scale. My year abroad in the UK, in every social event I attended organized by a student society there would inevitably be left-over budget, which would be taken care of by.. a raffle. Some
45.
by Phemist 2mo ago
Which I do (actually a fork of firefox - waterfox), but it is still worrying as the viability of chromium-base browsers is what's keeping us having "merely" a chromium-monoculture (with viable ad blockers e.g. and website ope
46.
by Phemist 2mo ago
Worrying. Extrapolating a (speculative) future, this means (Google will feel) that soon the chromium base will no longer need the crowd-sourced bug hunting that is open-source. I expect Google to eventually stop working on chromium (in the
47.
by Phemist 2mo ago
You are confusing a lot of different lines of argument, and in the end I'm not even sure what you are arguing against. I think you are mostly agreeing with the proposed solution by echelon? You mentioned AB1856 which seems waaaaay broa
48.
by Phemist 2mo ago
It reads like AB1856 needs website operators to prove they didnt serve their content to the wrong age bracket, which requires way heavier methods than simply trusting the emitted "user is child" header (or rather, trusting that th
49.
by Phemist 2mo ago
Safe to assume I reject things on my first visit as well
50.
by Phemist 2mo ago
Malicious compliance. You do not need a cookie to "store" the fact you have rejected them. They can simply assume you have rejected them from the lack of cookies. They can store a cookie once you (have gone out of your way to) acc
51.
by Phemist 2mo ago
> Tired of misleading cookie banners? The EU Commission has finally proposed a solution: set your privacy preferences in the browser once, and never see another banner. So lawmakers do know how to make legally binding preferences based o
52.
by Phemist 3mo ago
Clearly, NOBUS can work at multiple levels. DJB previously posted about DES (besides it being weak enough): NSA wanted DES to "drive out competitors", to "reduce the field that NSA had to be concerned about". Simply redu
53.
by Phemist 3mo ago
How much relative to 70K is this 12L? Sorry, I am not Indian.
54.
by Phemist 4mo ago
Indeed, a clear Freudian slip. The one where you say one thing, but you mean your mother.
55.
by Phemist 4mo ago
I wasn't familiar with Unclothe, so I had to look it up.. Are you sure you did not mean Unsloth?
56.
by Phemist 4mo ago
If an agent makes a tool call, the LLM provider will receive the full context again after the result of the tool call becomes available in order to decide the next move. Everything up to the point of the tool call being made will no longer
57.
by Phemist 5mo ago
Left-handed people are often excluded from participating in MRI studies. To my personal dismay, as these studies often paid 25 euros per hour ~20 years ago, a significant sum for my student self that I could not partake in. It has however g
58.
by Phemist 5mo ago
Or if you've seen Ted Lasso
59.
by Phemist 5mo ago
Nice! I also found this page very helpful: https://www.nayuki.io/page/creating-a-qr-code-step-by-step This is an interactive guide that will break down the process for your specific QR-code.
60.
by Phemist 5mo ago
I meant the ones you see on those infotainment systems in museums that are super durable, but have terrible ergonomics. The show props those up way too much.

More ›