iflp

6 ms·

1.
by iflp 2y ago
The seed is computed using the last 4 tokens, but to get the predictive distribution they should still need the full history?
2.
UK watchdog examining Microsoft's ties to OpenAI(ft.com)
3 points by iflp 3y ago|0 comments
3.
Tuition Fees in the United Kingdom(en.wikipedia.org)
2 points by iflp 3y ago|0 comments
4.
by iflp 3y ago
> The superconducting-like behavior in LK-99 most likely originates from a magnitude reduction in resistivity caused by the first-order structural phase transition of Cu2S. [...] It is important to note that this first-order structural t
5.
First order transition in [LK-99] containing Cu2S(arxiv.org)
57 points by iflp 3y ago|13 comments
6.
“Attention”, “Transformers”, in Neural Network “Large Language Models”(bactra.org)
2 points by iflp 3y ago|0 comments
7.
by iflp 3y ago
This seems to be about easier classification tasks with not too many samples, for which TF-IDF also works well (Table 3). But more generally gzip for text modeling might make sense. Quoting http://bactra.org/notebooks&#x2F
8.
by iflp 3y ago
Content aside, this is terrible editing on BBC’s part: > However, the regulator hit back, saying: "It is the CMA's job to do what is best for the people, businesses and economy of the UK, not merging firms with commercial inter
9.
Alexander and Yudkowsky on AGI Goals(lesswrong.com)
2 points by iflp 3y ago|0 comments
10.
Neurodegenerative disease among male elite football players in Sweden(thelancet.com)
1 points by iflp 4y ago|0 comments
11.
by iflp 4y ago
These are all good reasons, but it’s really a new level of openness from them.
12.
Fordism(en.wikipedia.org)
2 points by iflp 4y ago|1 comments
13.
by iflp 4y ago
No. I was referring to the "standard concentration bound" in that paper, which applies when you have separate validation and test sets. I think the argument can usually be improved by applying small-variance inequalities such as
14.
by iflp 4y ago
If you only care about identically distributed test data, test set overfitting doesn't happen that fast: if you evaluate M models on N test samples, the overfitting error is on the order of sqrt(log M / N). And even as this err
15.
Further delays at ITER are certain, but their duration isn’t clear(physicstoday.scitation.org)
2 points by iflp 4y ago|0 comments
16.
Faculty member issues dire warning to grad students about jobs(insidehighered.com)
33 points by iflp 4y ago|22 comments
17.
by iflp 4y ago
StartAllBack
18.
Zhang, Yitang’s life at Purdue [pdf, 2018](math.purdue.edu)
2 points by iflp 4y ago|0 comments
19.
Idera, Inc(en.wikipedia.org)
2 points by iflp 4y ago|1 comments
20.
The Senatorial Rejection of Leland Olds: A Case Study(jstor.org)
1 points by iflp 4y ago|0 comments
21.
Autodidax: Jax Core from Scratch(jax.readthedocs.io)
1 points by iflp 4y ago|0 comments
22.
by iflp 4y ago
surface go?
23.
by iflp 4y ago
The wake-sleep algorithm is designed to provide generative models with necessary sleep, at least in their developmental stages. (Sorry, can’t resist.)
24.
Florida Python Challenge(flpythonchallenge.org)
1 points by iflp 4y ago|0 comments
25.
by iflp 4y ago
The equation example is artificial. In practice there will be no curly brackets around these single-token sub/superscripts, nor should the \left / \right present in this example. With properly added whitespaces this equation bec
26.
by iflp 4y ago
The lockdown policy was successful in these aspects before, but it’s very unclear if it can still work with omicron.
27.
Book Review: 'A New Kind of Science' (2002)(arxiv.org)
2 points by iflp 4y ago|0 comments
28.
by iflp 5y ago
It is a random variable in this setting, as it is a function of the randomly generated password. Given a deterministic sequence, you find the definition of its Kolmogorov complexity in textbooks/Wikipedia/etc. By saying the Kolmog
29.
by iflp 5y ago
Kolmogorov complexity is only unambiguously defined asymptotically, and "asymptotics is merely a heuristic". It is also uncomputable. So, to use entropy arguments for passwords, the only correct way I could think of is to genera
30.
by iflp 5y ago
Kolmogorov complexity/entropy is more suitable for this purpose, under the implicit assumption that password crackers don't have tailored prior knowledge and are just enumerating "simple" sequences. It only agrees with S

More ›