Downloads · 30 days
314
10% of all-time downloads
blascotobasco/Gemma-4-14B-A4B-Heretic-v2-GGUF
Gemma-4-14B-A4B-Heretic-v2-GGUF is a machine learning model from blascotobasco. Use it for the machine learning task on the model card, and read the license before you ship it in a product.
Following up on the previous trained version of my Gemma 4 prune, this version has been trained on MetaMathQA to restore arithmetic capabilities. I have confirmed that the model's roleplaying and instruction following…
Downloads · 30 days
314
10% of all-time downloads
All-time downloads
3.2K
Public
Repo size
38 GB
Likes
2
Public
Click a slice to open those files.
.gguf38 GB · 100%
From the Hugging Face model README
Following up on the previous trained version of my Gemma 4 prune, this version has been trained on MetaMathQA to restore arithmetic capabilities. I have confirmed that the model's roleplaying and instruction following capabilities are fully intact, as is vision.
While the model can do arithmetic properly now, it may still struggle with complex proofs or problems. However for most use cases it should be robust enough.
Since this is a pruned model, I do not recommend relying on it for important, complex work, especially mathematics. Even though it has been trained, a prune is still weaker than the original model, especially a fifty percent prune like this one.
This is based on a Heretic Abliterated version of Gemma 4 26B A4B IT.
After further testing, it seems math was not restored. It's possible the model just got lucky during training or the unquantized weights were more capable. However I noticed that the model still has improved logic following even if degraded.