Skip to content

THUDM/LongReward

Repository files navigation

LongReward: Improving Long-context Large Language Models with AI Feedback

🤗 HuggingFace • 📃 Paper

🔍 Table of Contents

LongReward

cof

We open-source LongReward under long_reward/auto_scorer.py, a novel method that utilize an off-the-shelf LLM to automatically provide rewards for model responses in long-context scenarios, considering four human-valued dimensions: helpfulness, logicality, faithfulness, and completeness. Given a long-context-based model response, LongReward assigns a score ranging from 0 to 10 for each dimension, and takes their average as the final reward.

Please configure your API key in the utils/llm_api.py before running the code. You can also run the following scripts under batch_inference/ for large-scale inference: 1_get_chunk_info.py, 2_get_score.py. 3_get_pairs.py. The results will be stored in ./data.

Released Datasets & Models

SFT Datasets & SFT Models

Our SFT dataset contains 10k long-context QA instances, whose context lengths range from 8k to 64k tokens. The QA pairs are generated by GLM-4-0520, following the self-instruct method in LongAlign. Using this dataset, we supervised fine-tune two models: LongReward-glm4-9b-SFT and LongReward-llama3.1-8b-SFT, which are based on GLM-4-9B and Meta-Llama-3.1-8B, respectively, and support up to 64k context.

Preference Datasets & DPO Models

We utilize LongReward and prompts in the SFT dataset to construct the preference datasets for each SFT model, and train their DPO version: LongReward-glm4-9b-DPO and LongReward-llama3.1-8b-DPO. More Details can be found in our paper.

All Available Datasets and Models

Here is the full list of datasets and models we released:

Name Download Path
LongReward-10k (SFT & DPO Datasets) 🤗 HuggingFace
LongReward-glm4-9b-SFT 🤗 HuggingFace
LongReward-glm4-9b-DPO 🤗 HuggingFace, 🤖 ModelScope, 🟣 WiseModel
LongReward-llama3.1-8b-SFT 🤗 HuggingFace
LongReward-llama3.1-8b-DPO 🤗 HuggingFace, 🤖 ModelScope, 🟣 WiseModel

QuickStart

Try our model with following step:

  1. install requirements
pip install -r requirement.txt
  1. run with model
from transformers import AutoModelForCausalLM, AutoTokenizer

MODEL_PATH = 'THUDM/LongReward-glm4-9b-DPO'

tokenizer = AutoTokenizer.from_pretrained(MODEL_PATH)
model = AutoModelForCausalLM.from_pretrained(MODEL_PATH, device_map="auto")

message = [
    {
        "role": "user",
        "content": "W. Russell Todd, 94, United States Army general (b. 1928). February 13. Tim Aymar, 59, heavy metal singer (Pharaoh) (b. 1963). Marshall \"Eddie\" Conway, 76, Black Panther Party leader (b. 1946). Roger Bonk, 78, football player (North Dakota Fighting Sioux, Winnipeg Blue Bombers) (b. 1944). Conrad Dobler, 72, football player (St. Louis Cardinals, New Orleans Saints, Buffalo Bills) (b. 1950). Brian DuBois, 55, baseball player (Detroit Tigers) (b. 1967). Robert Geddes, 99, architect, dean of the Princeton University School of Architecture (1965–1982) (b. 1923). Tom Luddy, 79, film producer (Barfly, The Secret Garden), co-founder of the Telluride Film Festival (b. 1943). David Singmaster, 84, mathematician (b. 1938). \n\n What was Robert Geddes' profession?"
    }
]

inputs = tokenizer.apply_chat_template(
    message,
    return_tensors='pt',
    add_generation_prompt=True,
    return_dict=True,
).to(model.device)

input_len = inputs['input_ids'].shape[1]
generate_kwargs = {
    "input_ids": inputs['input_ids'],
    "attention_mask": inputs['attention_mask'],
    "max_new_tokens": 128,
    "do_sample": False,
}
out = model.generate(**generate_kwargs)
print(tokenizer.decode(out[0][input_len:], skip_special_tokens=True))

Evaluation

We provide our evaluation code for LongBench and LongBench-Chat under evaluation/. Details can be found in evaluation/README.md and evaluation/LongBench_Chat/README.md. Remember to configure your OpenAI API key in utils/llm_api.py since we adopt GPT-4o as the judge.

To reproduce our results on other benchmarks, we refer to the code in FastChat and alpaca_eval for evaluating on MT-Bench and AlpacaEval2.

Here are our evaluation results:

eval

Citation

If you find our work helpful, please consider citing the following paper:

@article{zhang2024longreward,
  title = {LongReward: Improving Long-context Large Language Models
with AI Feedback} 
  author={Jiajie Zhang and Zhongni Hou and Xin Lv and Shulin Cao and Zhenyu Hou and Yilin Niu and Lei Hou and Yuxiao Dong and Ling Feng and Juanzi Li},
  journal={arXiv preprint arXiv:2410.21252},
  year={2024}
}

License

  • The use of GLM-4 model weights must follow the Model License.

  • The code in this open source repository follows the Apache 2.0 license.

Please strictly follow the open source license.

About

No description, website, or topics provided.

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages