Rethinking Faithfulness in LLMs: A Pairwise Context-Sensitive Perspective

  

Large language models (LLMs) are expected to answer questions faithfully based on the provided context, abstaining when the context information is insufficient to answer the questions. Existing faithfulness evaluations typically assess each question-context instance in isolation; however, such instance-level evaluation fails to capture a fundamental requirement of faithful behavior: the ability to adapt model responses to changes in available contexts. In particular, a model should provide correct answers when sufficient evidence is present and abstain when it is not. In this work, we propose a Pairwise Faithfulness Benchmark (PFaithBench) that evaluates whether a model can switch between answering and abstaining for the same question under supporting versus non-supporting contexts. Our evaluations across thirty-nine models with seven model families demonstrate that faithfulness fundamentally involves a trade-off between answering and abstaining, and that most current models exhibit a strong bias toward answering, with most faithfulness errors arising from over-answering, i.e., models tend to fabricate a response even when the provided context is insufficient. We further conduct a series of studies on faithfulness training under different data constructions. Our results show that training outcomes are highly sensitive to the specific composition of answering and abstaining data. Constructing answering and abstaining data from mismatched sources can cause models to rely on dataset-specific shortcuts rather than actual context sufficiency. Moreover, increasing answer-supervised data improves answering performance but exacerbates over-answering, while increasing abstaining data reduces hallucination but leads to over-abstention.

Overview

Overview of pairwise faithfulness behavior. Left: Given the same question with sufficient and insufficient contexts, a faithful model should exhibit context-sensitive behavior: answering when sufficient evidence is available and abstaining when key information is missing. Over-abstention and over-answering are two sources of faithfulness errors. Right: Evaluations across seven families with thirty-nine models: most models exhibit bias toward over-answering, where one point is one model.

Pairwise faithfulness evaluation and PFaithBench construction. Left: Existing evaluation protocols assess individual question-context instances, confounding question and context variations. Middle: Pairwise evaluation keeps the question fixed while varying contextual sufficiency, enabling controlled assessment of faithfulness switching. Right: PFaithBench is constructed from multi-hop, single-hop, and counterfactual QA by modifying key information in sufficient contexts.

Guide to Run PFaithBench Evaluation and Faithfulness Training

• PFaithBench is in the data dir
• Training code is in verl

1. Run Evaluation of PFaithBench

# Step 1
# The evaluation scripts are provided in evaluation path

bash evaluate/generate_all_responses.sh

bash evaluate/judge_all_responses.sh

2. Run Faithfulness Training (GRPO)

# Step 1
# Fill-in your Model Path and data path in the run_GRPO.sh

# Run run_GRPO.sh to start the training
# Training data is in the data_train dir
bash run_GRPO.sh

Released Data and Checkpoints

We release the data and all checkpoints trained by us on huggingface:

Citation

If you use our datasets or models, please cite our paper!