2 ms·Lossless Compression of LLMl-Generated Text via Next-Token Prediction2 points by PaulHoule 1y ago