Abstract:Large Language Models (LLMs) are often augmented with external information as contexts, but this external information can sometimes be inaccurate or even intentionally misleading. We argue that robust LLMs should demonstrate situated faithfulness, dynamically calibrating their trust in external information based on their confidence in the internal knowledge and the external context. To benchmark this capability, we evaluate LLMs across several QA datasets, including a newly created dataset called RedditQA featuring in-the-wild incorrect contexts sourced from Reddit posts. We show that when provided with both correct and incorrect contexts, both open-source and proprietary models tend to overly rely on external information, regardless of its factual accuracy. To enhance situated faithfulness, we propose two approaches: Self-Guided Confidence Reasoning (SCR) and Rule-Based Confidence Reasoning (RCR). SCR enables models to self-access the confidence of external information relative to their own internal knowledge to produce the most accurate answer. RCR, in contrast, extracts explicit confidence signals from the LLM and determines the final answer using predefined rules. Our results show that for LLMs with strong reasoning capabilities, such as GPT-4o and GPT-4o mini, SCR outperforms RCR, achieving improvements of up to 24.2% over a direct input augmentation baseline. Conversely, for a smaller model like Llama-3-8B, RCR outperforms SCR. Fine-tuning SCR with our proposed Confidence Reasoning Direct Preference Optimization (CR-DPO) method improves performance on both seen and unseen datasets, yielding an average improvement of 8.9% on Llama-3-8B. In addition to quantitative results, we offer insights into the relative strengths of SCR and RCR. Our findings highlight promising avenues for improving situated faithfulness in LLMs. The data and code are released.

Understanding the Interplay between Parametric and Contextual Knowledge for Large Language Models

Large Language Models and Knowledge Graphs: Opportunities and Challenges

When Context Leads but Parametric Memory Follows in Large Language Models

Resolving Knowledge Conflicts in Large Language Models

Augmented Large Language Models with Parametric Knowledge Guiding

Evaluating the External and Parametric Knowledge Fusion of Large Language Models

Adaptive Chameleon or Stubborn Sloth: Revealing the Behavior of Large Language Models in Knowledge Conflicts

"Merge Conflicts!" Exploring the Impacts of External Distractors to Parametric Knowledge Graphs

Large Knowledge Model: Perspectives and Challenges

Supervised Knowledge Makes Large Language Models Better In-context Learners

Knowledge Conflicts for LLMs: A Survey

Large Language Models Can Better Understand Knowledge Graphs Than We Thought

Large Language Models with Controllable Working Memory

Probing Language Models on Their Knowledge Source

Context-faithful Prompting for Large Language Models

A Principled Framework for Knowledge-enhanced Large Language Model

Self-Updatable Large Language Models with Parameter Integration

Seeking Neural Nuggets: Knowledge Transfer in Large Language Models from a Parametric Perspective

Enhancing Large Language Models' Situated Faithfulness to External Contexts

Challenges and Contributing Factors in the Utilization of Large Language Models (LLMs)