fine tuning
4 stories and discussions about fine tuning, aggregated from every source we track.
1.
2.
On real multi-year Comcast support traffic, a fine-tuned verifier's gray-zone positive rate drifted 5x and the model went from helping to actively hurting. Here's the failure, and the three-detector monitor we built…
3.
Empirical notes on fine-tuning π0.5 for a real manufacturing task: training duration, LoRA vs full fine-tuning, batch size, data quantity and quality, and how much your evaluation can actually resolve.
4.
Empirical notes on fine-tuning π0.5 for a real manufacturing task: training duration, LoRA vs full fine-tuning, batch size, data quantity and quality, and how much your evaluation can actually resolve.