Llama-2-7b-chat-TruthX

TruthX: Alleviating Hallucinations by Editing Large Language Models in Truthful Space

Shaolei Zhang, Tian Yu, Yang Feng*

Model for paper "TruthX: Alleviating Hallucinations by Editing Large Language Models in Truthful Space".

TruthX is an inference-time method to elicit the truthfulness of LLMs by editing their internal representations in truthful space, thereby mitigating the hallucinations of LLMs. On the TruthfulQA benchmark, TruthX yields an average enhancement of 20% in truthfulness across 13 advanced LLMs.

<div align="center"> <img src="./truthx_results.png" alt="img" width="100%" /> </div> <p align="center"> TruthfulQA MC1 accuracy of TruthX across 13 advanced LLMs </p>

This repo provides Llama-2-7B-Chat-TruthX, a Llama-2-7B-Chat model with baked-in TruthX model. You can directly download this baked-in model and use it like standard Llama, no additional operations are required.

Quick Starts

Inference with Llama-2-7B-Chat-TruthX:

import torch
from transformers import AutoTokenizer, AutoModelForCausalLM

llama2chat_with_truthx = "ICTNLP/Llama-2-7b-chat-TruthX"
tokenizer = AutoTokenizer.from_pretrained(llama2chat_with_truthx, trust_remote_code=True)
model = AutoModelForCausalLM.from_pretrained(llama2chat_with_truthx, trust_remote_code=True,torch_dtype=torch.float16).cuda()

question = "What are the benefits of eating an apple a day?"
encoded_inputs = tokenizer(question, return_tensors="pt")["input_ids"]
outputs = model.generate(encoded_inputs.cuda())[0, encoded_inputs.shape[-1] :]
outputs_text = tokenizer.decode(outputs, skip_special_tokens=True).strip()
print(outputs_text)

Please refer to GitHub repo and our paper for more details.

Licence

Model weights and the inference code are released under The GNU General Public License v3.0 (GPLv3)

Citation

If this repository is useful for you, please cite as:

@misc{zhang2024truthx,
      title={TruthX: Alleviating Hallucinations by Editing Large Language Models in Truthful Space}, 
      author={Shaolei Zhang and Tian Yu and Yang Feng},
      year={2024},
      eprint={2402.17811},
      archivePrefix={arXiv},
      primaryClass={cs.CL},
      url={https://arxiv.org/abs/2402.17811}
}

If you have any questions, feel free to contact [email protected].