heljakka

4 ms·

1. 1.▲
   by heljakka 10mo ago

   I can confirm this after hundreds of talks about the topic over the last 2 years. 90% of cases are simply not high-volume or high-stakes enough for the devs to care enough. I'm a founder of an evaluation automation startup, and our cha
2. 2.▲
   by heljakka 10mo ago

   What are the main shortcomings of the solutions you tried out? We believe you need to both automatically create the evaluation policies from OTEL data (data-first) and to bring in rigorous LLM judge automation from the other end (intent-fir
3. 3.▲
   Pioneer Networks(arxiv.org)

   3 points by heljakka 8y ago|0 comments
4. 4.▲
   GANs for Simulation, Representation and Inference(medium.com)

   3 points by heljakka 9y ago|0 comments