News for hackers. Stuff that matters.

austinpilot

Re:Harnessing the Universal Geometry of Embeddings on Wednesday May 21, 2025 @08:52PM (#44057709)
There has been plenty of work on alignment of embeddings, a lot of it cited in the paper. This paper solves the problem of translation, where (unlike in word alignment) there isn't already a set of candidate vectors in the target embedding space. It's generation, not matching.
Re:Harnessing the Universal Geometry of Embeddings on Wednesday May 21, 2025 @08:49PM (#44057702)
The problem solved in this paper is strictly harder than alignment. Alignment works with multiple, unmatched representations of the same inputs (e.g, different embeddings of the same words). The goal is to match them up.

The goal here is harder: given an embedding of an unknown text in one space, generate a vector in another space that's close to the embedding of the same text -- but, unlike in the word alignment problem, the texts are not known in advance.

Neither unsupervised transport, nor optimal alignment can solve this problem. Their input sets must be embeddings of the same texts. The input sets here are embeddings of different texts.

FWIW, this is all explained in the paper, including even the abstract. The comparisons with optimal assignment explicitly note that this is an idealized pseudo-baseline, and in reality OA cannot used for embedding translation (as opposed to matching, alignment, correspondence, etc.)

Re:The most dangerous code in the world on Wednesday October 24, 2012 @07:42PM (#4695593)
One of the authors here... The paper is accompanied by a FAQ, which among other things explains the impact on specific software:

https://docs.google.com/document/pub?id=1roBIeSJsYq3Ntpf6N0P...

For example, broken SSL in Amazon FPS allows a MITM attacker to forge instant payment notifications and defraud merchants who use vulnerable SDKs. This is a real vulnerability, acknowledged by Amazon.