d2ou

See the excellent "LLM can't jump" position paper on this : with the main illustration being the discovery of general relativity.

https://openreview.net/forum?id=klU4737opt

People in my lab (sklearn people) developped something that I feel close to jev but focused on tabular data : https://tabicl.readthedocs.io/en/latest/

This is a transformer based classifier with massive pretraining on synthetic datasets and it outperforms boosting classifiers on many benchmarks without the need of more gradient descent steps (the forward pass on X_train, y_train IS the training).

I understand that jev focus on text entry. But I feel that it is a similar kind of model but trained on text. Did someone test it on tabular data as well ?

And for a similar dplyr-like API on duckdb I recommand the excellent https://ibis-project.org/ !

> everyone wants more

Some evidence of satiation have been shown in five western countries. Satiation is defined in the following work as an income level where more money dont increase subjective well being. The threshold varies and could very well depend on the generosity of the social safety net : https://www.insee.fr/en/statistiques/8199185

In France we already have https://hal.science/ which is quite well integrated with arxiv.

However I don't think that it supports open review which this new cern initiative seems to promote. Great if followed and promoted by big national labs

Is there any report on the distribution of growth per firm ? (Interested for example by the median of firm growth). Could be a good evidence to support this inequal repartition of growth argument).

Would it make sense for the big players to buy them? Seems to be a huge avenue here to kill inference costs which always made me dubious on LLMs in general.