Attribute or Abstain: Large Language Models as Long Document Assistants

Jan Buchmann (maintainer), Xiao Liu, Iryna Gurevych

Ubiquitous Knowledge Processing Lab (UKP), TU Darmstadt

Contact: [email protected]

---

This repository contains the evaluation and analysis code for LAB, the Long document Attribution Benchmark introduced in the paper "Attribute or Abstain: Large Language Models as Long Document Assistants" (Arxiv link will follow soon). The benchmark consists of 6 diverse long document tasks:

• QASPER (scientific QA, Dasigi et al., 2021)
• Natural Questions (QA on Wiki pages, Kwiatkowski et al., 2019)
• Evidence Inference (classification of clinical trial reports, DeYoung et al., 2020)
• Wice (fact checking, Kamoi et al., 2023)
• ContractNLI (legal NLI, Koreeda and Manning, 2021) tests the ability of LLMs (in combination with retrievers) to produce responses
• GovReport (summarization, Huang et al., 2021)

To perform well on the benchmark, models need to provide correct responses and extract evidence from the documents that support their response (attribute). If there is not enough information to provide a useful response, models should explicitly state this (abstain).

To download the pre-formatted datasets, please see here.

Setup

Data

It is assumed that there is a data folder at ../data (relative to this repository). This folder should contain the following:

attribution
    datasets <- Datasets to test attributability evaluation models
    results <- Test results for attribution evaluation models
datasets <- Datasets to evaluate attribution capabilities of LLMs 
openai_cost_tracking <- Only needed when using openAI models
results <- All predictions, results and models will be stored here
tensorboard <- Logging

You can simply download the data from here, unzip it and place it in the same folder as this repository. If you want to use a different location, adapt the paths in config/location/local

Install requirements

conda create -n lab_env python=3.11
conda activate lab_env
# There is a requirements conflict because we are using cuda 11.8,
# so  installation is done in this bash script
bash install.sh

Basic Usage

The run_evaluation.py script serves as the entry point for all experiment runs. All important hyperparameters are controlled by the files in the config/ directory. See below for an explanation of the config and the most important parameters.

To evaluate the Longchat on the dev set of QASPER, you can run

python run_evaluation.py task=qasper model=longchat-7b-v1.5-32k use_dev_as_test_data=True

The Config

Hyperparameter management is done using the hydra package. All hyperparameters are documented in the corresponding classes in the config_lib/ directory. These classes document all config parameters and serve as typecheckers for the input values.

The default values of the parameters can be seen in the yaml files in the config/ directory. The defaults can be overridden by adding <param_name>=<param_value> to the command.

Important config parameters

All config parameters are documented in the respective module in the config_lib/ directory.

description: A free text description of the experiment model: The model to use. See config/model for all available models. task: The task to evaluate on. See evaluation/tasks for all available tasks. required_aspects: str, one of

• "answer_and_segments" (used when doing citation)
• "answer_only" (used when doing post-hoc and retrieve-then-read)
• "segments_only" do_train: Whether to fine-tune the model on the train set of the task. use_dev_as_test_data: Whether to evaluate on the dev data. use_first_n_test_instances: Determine the number of instances to evaluate on. If set to -1, all available test / dev instances are used.

Running Experiments

Post hoc

To run post hoc experiments, we first run LLM inference, and then do post hoc evidence retrieval

# Run inference on QASPER with GPT-4, using the first 100 instances from the test set
# This will produce a new folder in the results directory, named with a unique hash
python run_evaluation.py description='post_hoc' task=qasper model=gpt-4-turbo-128k use_dev_as_test_data=False required_aspects=answer_only use_first_n_test_instances=100

# Get the hash and evaluate. This will do post hoc retrieval and evaluate attributability
python evaluate_predictions.py <hash> --post_hoc_extract

Retrieve-then-read

# Run inference on QASPER with GPT-4, using the first 100 instances from the test set
# This will produce a new folder in the results directory, named with a unique hash
python run_evaluation.py description='post_hoc' task=qasper model=gpt-4-turbo-128k use_dev_as_test_data=False required_aspects=answer_only use_first_n_test_instances=100 do_retrieve_then_read=short

# Get the hash and evaluate. This will evaluate attributability
python evaluate_predictions.py <hash>

Citation

# Run inference on QASPER with GPT-4, using the first 100 instances from the test set
# This will produce a new folder in the results directory, named with a unique hash
python run_evaluation.py description='post_hoc' task=qasper model=gpt-4-turbo-128k use_dev_as_test_data=False required_aspects=answer_and_segments use_first_n_test_instances=100

# Get the hash and evaluate. This will evaluate attributability
python evaluate_predictions.py <hash>

Reduced Post Hoc

# Run inference on QASPER with GPT-4, using the first 100 instances from the test set
# This will produce a new folder in the results directory, named with a unique hash
python run_evaluation.py description='post_hoc' task=qasper model=gpt-4-turbo-128k use_dev_as_test_data=False required_aspects=answer_and_segments use_first_n_test_instances=100 do_retrieve_then_read=long

# Get the hash and evaluate. This will do post hoc retrieval and evaluate attributability
python evaluate_predictions.py <hash> --post_hoc_extract

Reduced Citation

# Run inference on QASPER with GPT-4, using the first 100 instances from the test set
# This will produce a new folder in the results directory, named with a unique hash
python run_evaluation.py description='post_hoc' task=qasper model=gpt-4-turbo-128k use_dev_as_test_data=False required_aspects=answer_and_segments use_first_n_test_instances=100 do_retrieve_then_read=long

# Get the hash and evaluate. This will evaluate attributability
python evaluate_predictions.py <hash>

Running Attribution Evaluation Models on Test data

python evaluate_attribution.py --description 'test' --model_name true_nli --task_name qasper --partition dev --is_csv

Adding new tasks / datasets

• Add new datasets to ../data/datasets. Create a new directory <dataset_name>. In that new directory, create one directory <dataset_name>_itg for the dataset in itg format and optionally another directory for the raw dataset.
• Add a new <task-name>.py file to evaluation/tasks. Implement the required task class (...Task) in that file, inheriting from the BaseTask classe in evaluation/common.py. You can use the other tasks for reference
• If your dataset exists as individual json files, you might want to use evaluation/common/SingleFileDataset to implement data loading
• Add your new classes to run_evaluation.py
• Add new yaml file to config/tasks that specifies hyperparameters for the task. If new hyperparameters are necessary, add them to config_lib/base_task/BaseTaskConfig
• Add new task_explanation and example to config/prompts.py

Logging

To view the logs in your browser, follow these steps:

1. Install tensorboard on your local machine pip install tensorboard
2. Set logging directory in config/location/local
3. Start tensorboard tensorboard --logdir /path/to/logging/dir

Structure of this repository

├── attribution_eval                # Functions and classes for attributability evaluation
    ├── attribution_dataset.py      # Contains code to construct claims from responses for attributability evaluation
    ├── attribution_model.py        # Contains code for the attributability evaluation models
    └── util.py                     # utility functions 

├── config                          # contains yaml files with default values for hydra config parameters
    ├── location                    # Default paths for models, result files, predictions, datasets, ...
    ├── model                       # defaults for currently implemented models
    ├── task                        # defaults for currently implemented tasks
    ├── config.yaml                 # top-level config file with general parameters
    └── prompts.yaml                # Hard coded prompt elements
    
├── config_lib                      # contains config classes for documentation and type checking

├── dataset_creation
    ├── make_attributed_dataset.py  # Contains code to automatically annotate evidence via post hoc retrieval (Used for GovReport in the paper).

├── evaluation
    ├── tasks                       # Contains task classes that implement instance loading
    ├── callbacks.py                # Contains callbacks that are called during evaluation, important for storing predictions and logging
    ├── common.py                   # Base classes for tasks, instances, predictions. New tasks should inherit from these base classes.
    ├── metrics.py                  # Metrics for evaluation (e.g. Answer F1)
    ├── run.py                      # Contains run() which is called in run_evaluation.py
    ├── validator.py                # Basic implementation of evaluation loop.
    └── util.py                     # Utility functions for evaluation
    
├── models
    ├── api_lm.py                   # General implementation for API-served models (e.g. ChatGPT).
    ├── base_model.py               # Base class for models
    ├── causal_lm.py                # General implementation of decoder-only models that can be accessed through Huggingface's AutoModelForCausalLM class.
    ├── oracle.py                   # A simple oracle model that returns the ground truth as predictions. Can be useful for debugging.
    ├── retrieve.py                 # Retrievers for reducing, retrieve then read and post hoc
    └── seq2seq_lm.py               # General implementation of encoder-decoder LMs.

├── structformer
    ├── collation_utils.py          # Helper functions for data collation
    ├── example_formatting.py       # Helper functions for formatting of in-context examples.
    ├── format_enforcement.py       # Creates functions for constrained generation
    ├── input_preparation.py        # Functions for data processing
    ├── input_sequence.py           # Functions to produce a single string from an Intertext Document
    └── sequence_alignment.py       # Functions to map between tokenized ITGs and ITGs.
    
├── evaluate_predictions.py         # Main script to evaluate runs, e.g. for attributability. Post hoc retrieval is also done here
├── evaluate_attribution.py         # Main script to evaluate attributability evaluation models
├── install.sh                      # Script to install dependencies
└── run_evaluation.py               # Main script for experiments

Citation

If you use our data or code in your research, please cite

TODO: add citation

References

@inproceedings{dasigi-etal-2021-dataset,
    title = "A Dataset of Information-Seeking Questions and Answers Anchored in Research Papers",
    author = "Dasigi, Pradeep  and
      Lo, Kyle  and
      Beltagy, Iz  and
      Cohan, Arman  and
      Smith, Noah A.  and
      Gardner, Matt",
    editor = "Toutanova, Kristina  and
      Rumshisky, Anna  and
      Zettlemoyer, Luke  and
      Hakkani-Tur, Dilek  and
      Beltagy, Iz  and
      Bethard, Steven  and
      Cotterell, Ryan  and
      Chakraborty, Tanmoy  and
      Zhou, Yichao",
    booktitle = "Proceedings of the 2021 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies",
    month = jun,
    year = "2021",
    address = "Online",
    publisher = "Association for Computational Linguistics",
    url = "https://aclanthology.org/2021.naacl-main.365",
    doi = "10.18653/v1/2021.naacl-main.365",
    pages = "4599--4610",
}

@article{kwiatkowski2019natural,
    author = {Kwiatkowski, Tom and Palomaki, Jennimaria and Redfield, Olivia and Collins, Michael and Parikh, Ankur and Alberti, Chris and Epstein, Danielle and Polosukhin, Illia and Devlin, Jacob and Lee, Kenton and Toutanova, Kristina and Jones, Llion and Kelcey, Matthew and Chang, Ming-Wei and Dai, Andrew
                        M. and Uszkoreit, Jakob and Le, Quoc and Petrov, Slav},
    title = "{Natural Questions: A Benchmark for Question Answering
                    Research}",
    journal = {Transactions of the Association for Computational Linguistics},
    volume = {7},
    pages = {453-466},
    year = {2019},
    month = {08},
    issn = {2307-387X},
    doi = {10.1162/tacl_a_00276},
    url = {https://doi.org/10.1162/tacl\_a\_00276},
    eprint = {https://direct.mit.edu/tacl/article-pdf/doi/10.1162/tacl\_a\_00276/1923288/tacl\_a\_00276.pdf},
}

@inproceedings{deyoung-etal-2020-evidence,
    title = "Evidence Inference 2.0: More Data, Better Models",
    author = "DeYoung, Jay  and
      Lehman, Eric  and
      Nye, Benjamin  and
      Marshall, Iain  and
      Wallace, Byron C.",
    editor = "Demner-Fushman, Dina  and
      Cohen, Kevin Bretonnel  and
      Ananiadou, Sophia  and
      Tsujii, Junichi",
    booktitle = "Proceedings of the 19th SIGBioMed Workshop on Biomedical Language Processing",
    month = jul,
    year = "2020",
    address = "Online",
    publisher = "Association for Computational Linguistics",
    url = "https://aclanthology.org/2020.bionlp-1.13",
    doi = "10.18653/v1/2020.bionlp-1.13",
    pages = "123--132"
}

@inproceedings{kamoi-etal-2023-wice,
    title = "{W}i{CE}: Real-World Entailment for Claims in {W}ikipedia",
    author = "Kamoi, Ryo  and
      Goyal, Tanya  and
      Diego Rodriguez, Juan  and
      Durrett, Greg",
    editor = "Bouamor, Houda  and
      Pino, Juan  and
      Bali, Kalika",
    booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing",
    month = dec,
    year = "2023",
    address = "Singapore",
    publisher = "Association for Computational Linguistics",
    url = "https://aclanthology.org/2023.emnlp-main.470",
    doi = "10.18653/v1/2023.emnlp-main.470",
    pages = "7561--7583",
}

@inproceedings{koreeda-manning-2021-contractnli-dataset,
    title = "{C}ontract{NLI}: A Dataset for Document-level Natural Language Inference for Contracts",
    author = "Koreeda, Yuta  and
      Manning, Christopher",
    editor = "Moens, Marie-Francine  and
      Huang, Xuanjing  and
      Specia, Lucia  and
      Yih, Scott Wen-tau",
    booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2021",
    month = nov,
    year = "2021",
    address = "Punta Cana, Dominican Republic",
    publisher = "Association for Computational Linguistics",
    url = "https://aclanthology.org/2021.findings-emnlp.164",
    doi = "10.18653/v1/2021.findings-emnlp.164",
    pages = "1907--1919"
}

@inproceedings{huang-etal-2021-efficient,
    title = "Efficient Attentions for Long Document Summarization",
    author = "Huang, Luyang  and
      Cao, Shuyang  and
      Parulian, Nikolaus  and
      Ji, Heng  and
      Wang, Lu",
    editor = "Toutanova, Kristina  and
      Rumshisky, Anna  and
      Zettlemoyer, Luke  and
      Hakkani-Tur, Dilek  and
      Beltagy, Iz  and
      Bethard, Steven  and
      Cotterell, Ryan  and
      Chakraborty, Tanmoy  and
      Zhou, Yichao",
    booktitle = "Proceedings of the 2021 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies",
    month = jun,
    year = "2021",
    address = "Online",
    publisher = "Association for Computational Linguistics",
    url = "https://aclanthology.org/2021.naacl-main.112",
    doi = "10.18653/v1/2021.naacl-main.112",
    pages = "1419--1436",
}

Disclaimer

This repository contains experimental software and is published for the sole purpose of giving additional background details on the respective publication.