Retrained with original gen_features.py + backend servers (fp32 GPT2-XL, 8bit Neo/J/LLaMA). Word acc 95.7%, Sent F1 95.6%.

Retrained with paper's exact feature models (gpt2-xl, gpt-neo-2.7B, gpt-j-6B, llama-7B). Word acc: 95.6%, Sent F1: 95.7% on SeqXGPT-Bench test split.

Upload model.py with huggingface_hub

Upload README.md with huggingface_hub

Upload seqxgpt_transformer.pt with huggingface_hub

initial commit