Jinx vs remove-refusals-with-transformers?
How does this version compare to the remove-refusals-with-transformers technique used by https://huggingface.co/huihui-ai/Huihui-Qwen3-30B-A3B-Thinking-2507-abliterated ?
That's an good question! Would you be interested in conducting an empirical comparison between the two models?
Any updates?
Hi @ParthProLegend . We believe the refuse rate of Jinx is lower than the abliterated models. You can take some quick case tests to verify. Your feedback is valuable for the community.
I am unable to run full models in my rtx 3060 laptop gpu with 6gb vram to compare, so I can only run their quants which might give me different results. So, I trust you here as I couldn't find much about Hui Hui models.
We've released other models with smaller sizes that might work better with your setup. You could try doing a side-by-side comparison between Jinx and the abliterated version of a smaller model to get a better sense of the differences. Enjory~