A finetune of Qwen3.5-Antirep-27B to reduce infinite looping and repetition in multiturn conversations.

Trained using int8 LoRA with DPO on 1x A100.

Safetensors

Model size

27B params

Tensor type

BF16

Inference Providers NEW

This model isn't deployed by any Inference Provider. 🙋 Ask for provider support

Model tree for ConicCat/Qwen3.5-Antirep-27B

Base model

Finetuned

(31)

this model

Dataset used to train ConicCat/Qwen3.5-Antirep-27B