README.md

Code for Aligning with Human Judgement: The Role of Pairwise Preference in Large Language Model Evaluators

Link to paper: Aligning with Human Judgement: The Role of Pairwise Preference in Large Language Model Evaluators (arXiv preprint arXiv:2403.16950)
This paper has been accepted by COLM 2024.

If you are interested in pairwise evaluator, please also checkout our latest work on zero-shot automatic prompt optimization for pairwise evaluators.

Code

Ready-to-use Package

We provide a ready-to-use Python library for Pairwise preference ranking (PairS). We show a ranking demonstration below. For an input source text and a sequence of output candidates, PairsGreedy and PairsBeam can be used to rank the output candidates in ascending order. We currently support the following base models: google/gemma-2-9b-it, google/gemma-2-27b-it, meta-llama/Meta-Llama-3-8B-Instruct, microsoft/Phi-3-medium-4k-instruct, microsoft/Phi-3-mini-4k-instruct, mistralai/Mistral-7B-Instruct-v0.1, meta-llama/Llama-2-7b-chat-hf, meta-llama/Llama-2-13b-chat-hf, HuggingFaceH4/zephyr-7b-beta, gpt-3.5-turbo, gpt-4-turbo.

from pairs import PairsGreedy, PairsBeam
from scripts.utils import shuffle_lists, load_summEval


# Load example data
summ_eval_path = 'data/SummEval/model_annotations.aligned.paired.jsonl'
input_doc, output_doc, _ = load_summEval(summ_eval_path, flat_output=False)

doc_id = 42
input, output = input_doc[doc_id], output_doc[doc_id]
input, output = shuffle_lists(input, output)

# The same input source text corresponds to multiple output summaries
print('Number of summary candidates:', len(output))

method = 'PairsGreedy'
if method == 'PairsGreedy':
    # Set hyperparameters
    params = {
        # 'engine': "mistralai/Mistral-7B-Instruct-v0.1",
        'engine': "meta-llama/Llama-2-7b-chat-hf",
        'api_call': 0,
        'with_input': True,   # Use the prompt template for task with context input, e.g. Summarization 
        'calibrate': False,   # For each pairwise comparison, we average the probabilities of both permutations to cancel the positional bias.
    }
    # Rank the output summaries from low to high quality
    indices = PairsGreedy(input[0], output, params)
    print(indices)

elif method == 'PairsBeam':
    # Set hyperparameters
    params = {
        'engine': "mistralai/Mistral-7B-Instruct-v0.1",
        'beam_size': 2000,
        'api_call': 0,
        'prob_gap': 0.1,
        'with_input': True,
        'calibrate': False,
    }
    # Rank the output summaries from low to high quality
    indices = PairsBeam(input[0], output, params)
    print(indices)

Evaluate on Datasets

We also present the original code (in the folder scripts/) to evalute on the datasets reported in the paper.

For NewsRoom and SummEval

bash pairs_run.sh

Notebook Demo

We provide a Notebook demonstrations in notebooks/.

Break downs

Load dataset: We put all datasets loading in scripts/utils.py.

Prompts: We put all prompts and instructions in scripts/prompts.py.

Base models: We supports the following base models, mistralai/Mistral-7B-Instruct-v0.1, meta-llama/Llama-2-7b-chat-hf, all versions of GPT-3.5-turbo and GPT-4-turbo.

Hyper-parameters:

• dataset: We support 3 datasets, 'newsroom', 'SummEval' and 'hanna'.
• eval_method: For all PairS method, we use 'pairwise comparison'.
• engine: The base models.
• with_input: If the data format has input text. For example, the summarization task has source text as input, but story writing task has no input text.
• confidence_beam: True for PairS-beam and False for PairS-greedy.
• prob_gap: The uncertainty tolerance. $0.1$ represents we will create beam candidates for both A and B if $0.5-0.1 < P(A\succ B) < 0.5+0.1$.
• calibrate: LLMs suffer from positional bias. Set this as True will average the probabilities of both permutations of A and B for each pairwise comparison. This will cancel the positional bias.

More details and comments will be added soon.

Algorithm of PairS-Beam

The PairS-Greedy can be understood as a merge sort with pairwise comparison by LLMs, while the PairS-Beam is to do a beam-search for each merge operation. In order to improve the beam search efficiency and limit the search space, we also apply a local uncertainty-based prunning mechanism.

We show the algorithm of the modified merge operation for PairS-Beam below.

A Beam-search Merge Operation Demonstration

For more details please check out our paper.

Citation

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

@article{liu2024aligning,
  title={Aligning with Human Judgement: The Role of Pairwise Preference in Large Language Model Evaluators},
  author={Liu, Yinhong and Zhou, Han and Guo, Zhijiang and Shareghi, Ehsan and Vulic, Ivan and Korhonen, Anna and Collier, Nigel},
  journal={arXiv preprint arXiv:2403.16950},
  year={2024}
}