I've spent months building a diagnostic method for large language models. It catches what standard benchmarks miss – distributional collapse inside tensors, not just loss or perplexity. Gemma 4 26B A4B fails it. I analyzed Gemma 4 26B A4B (Q8_0) qu…
Qwen 3.5 35B A3B Uncensored HauhauCS (repaired) -> (now with KL + ReLU calibration) Model available here: https://huggingface.co/LuffyTheFox/FernflowerAI-35B-A3B-KL-ReLU-GGUF Repair summary: link Extra information about how Qwen 3.5 35B got broken (…
Hello everyone. I found and fixed training bug in Qwen3.5 35B A3B model. Here my fixed version: https://huggingface.co/LuffyTheFox/Qwen3.5-35B-A3B-Uncensored-FernflowerAI-GGUF Upgraded system prompt that unlocks deep thinking (works great with this mod…