To Trust or Not to Trust? Enhancing Large Language Models' Situated Faithfulness to External Contexts

• URL: http://arxiv.org/abs/2410.14675v2
• Date: Mon, 17 Mar 2025 04:47:58 GMT
• Title: To Trust or Not to Trust? Enhancing Large Language Models' Situated Faithfulness to External Contexts
• Authors: Yukun Huang, Sanxing Chen, Hongyi Cai, Bhuwan Dhingra,
• Abstract summary: Large Language Models (LLMs) are often augmented with external contexts, such as those used in retrieval-augmented generation (RAG)<n>We show that when provided with both correct and incorrect contexts, both open-source and proprietary models tend to overly rely on external information.<n>We propose two approaches: Self-Guided Confidence Reasoning (SCR) and Rule-Based Confidence Reasoning (RCR)
• Score: 10.748768620243982
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Large Language Models (LLMs) are often augmented with external contexts, such as those used in retrieval-augmented generation (RAG). However, these contexts can be inaccurate or intentionally misleading, leading to conflicts with the model's internal knowledge. 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 resolve knowledge conflicts. To benchmark this capability, we evaluate LLMs across several QA datasets, including a newly created dataset 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-assess 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.

Related papers

• A Context-Aware Dual-Metric Framework for Confidence Estimation in Large Language Models [6.62851757612838]
  Current confidence estimation methods for large language models (LLMs) neglect the relevance between responses and contextual information.<n>We propose CRUX, which integrates context faithfulness and consistency for confidence estimation via two novel metrics.<n> Experiments across three benchmark datasets demonstrate CRUX's effectiveness, achieving the highest AUROC than existing baselines.
  arXiv  Detail & Related papers  (2025-08-01T12:58:34Z)
• Deliberative Searcher: Improving LLM Reliability via Reinforcement Learning with constraints [18.10515528600634]
  We propose textbfDeliberative Searcher, the first framework to integrate certainty calibration with retrieval-based search for open-domain question answering.<n>The agent performs multi-step reflection and verification over Wikipedia data and is trained with a reinforcement learning algorithm that optimize for accuracy under a soft reliability constraint.
  arXiv  Detail & Related papers  (2025-07-22T16:09:34Z)
• Understanding and Benchmarking the Trustworthiness in Multimodal LLMs for Video Understanding [59.50808215134678]
  This study introduces Trust-videoLLMs, a first comprehensive benchmark evaluating 23 state-of-the-art videoLLMs.<n>Results reveal significant limitations in dynamic scene comprehension, cross-modal resilience and real-world risk mitigation.
  arXiv  Detail & Related papers  (2025-06-14T04:04:54Z)
• Divide-Then-Align: Honest Alignment based on the Knowledge Boundary of RAG [51.120170062795566]
  We propose Divide-Then-Align (DTA) to endow RAG systems with the ability to respond with "I don't know" when the query is out of the knowledge boundary.<n>DTA balances accuracy with appropriate abstention, enhancing the reliability and trustworthiness of retrieval-augmented systems.
  arXiv  Detail & Related papers  (2025-05-27T08:21:21Z)
• After Retrieval, Before Generation: Enhancing the Trustworthiness of Large Language Models in RAG [13.603907803297561]
  RAG systems face challenges in balancing internal (parametric) and external (retrieved) knowledge.<n>We propose the BRIDGE framework, which dynamically determines a comprehensive response strategy of large language models.<n>Experiments show BRIDGE outperforms baselines by 5-15% in accuracy while maintaining balanced performance across all scenarios.
  arXiv  Detail & Related papers  (2025-05-21T16:29:19Z)
• Assessing Judging Bias in Large Reasoning Models: An Empirical Study [99.86300466350013]
  Large Reasoning Models (LRMs) like DeepSeek-R1 and OpenAI-o1 have demonstrated remarkable reasoning capabilities. We present a benchmark comparing judging biases between LLMs and LRMs across both subjective preference-alignment datasets and objective fact-based datasets.
  arXiv  Detail & Related papers  (2025-04-14T07:14:27Z)
• Parameters vs. Context: Fine-Grained Control of Knowledge Reliance in Language Models [39.73834207174728]
  Retrieval-Augmented Generation (RAG) mitigates hallucinations in Large Language Models (LLMs) by integrating external knowledge. Conflicts between parametric knowledge and retrieved context pose challenges for RAGs. We propose **CK-PLUG**, a plug-and-play method for controlling RAG's reliance on parametric and contextual knowledge.
  arXiv  Detail & Related papers  (2025-03-20T06:26:28Z)
• ParamMute: Suppressing Knowledge-Critical FFNs for Faithful Retrieval-Augmented Generation [91.20492150248106]
  We investigate the internal mechanisms behind unfaithful generation and identify a subset of mid-to-deep feed-forward networks (FFNs) that are disproportionately activated in such cases.<n>We propose Parametric Knowledge Muting through FFN Suppression (ParamMute), a framework that improves contextual faithfulness by suppressing the activation of unfaithfulness-associated FFNs.<n> Experimental results show that ParamMute significantly enhances faithfulness across both CoFaithfulQA and the established ConFiQA benchmark, achieving substantial reductions in reliance on parametric memory.
  arXiv  Detail & Related papers  (2025-02-21T15:50:41Z)
• CER: Confidence Enhanced Reasoning in LLMs [2.4392539322920763]
  We introduce an uncertainty-aware framework designed to enhance the accuracy of Large Language Models responses. We quantify the confidence of intermediate answers such as numerical results in mathematical reasoning and proper nouns in open-domain generation. Results consistently validate the effectiveness of our novel confidence aggregation method.
  arXiv  Detail & Related papers  (2025-02-20T15:16:42Z)
• Towards Fully Exploiting LLM Internal States to Enhance Knowledge Boundary Perception [58.62352010928591]
  Large language models (LLMs) exhibit impressive performance across diverse tasks but often struggle to accurately gauge their knowledge boundaries. This paper explores leveraging LLMs' internal states to enhance their perception of knowledge boundaries from efficiency and risk perspectives.
  arXiv  Detail & Related papers  (2025-02-17T11:11:09Z)
• Mind the Confidence Gap: Overconfidence, Calibration, and Distractor Effects in Large Language Models [0.6091702876917281]
  Large Language Models (LLMs) show remarkable proficiency in natural language tasks.<n>Overconfidence-misalignment between predicted confidence and true correctness poses significant risks in critical decision-making applications.<n>We present a comprehensive analysis on calibration in LLMs across nine LLMs and three factual Question-Answering datasets.
  arXiv  Detail & Related papers  (2025-02-16T07:46:09Z)
• Aligning Large Language Models for Faithful Integrity Against Opposing Argument [71.33552795870544]
  Large Language Models (LLMs) have demonstrated impressive capabilities in complex reasoning tasks. They can be easily misled by unfaithful arguments during conversations, even when their original statements are correct. We propose a novel framework, named Alignment for Faithful Integrity with Confidence Estimation.
  arXiv  Detail & Related papers  (2025-01-02T16:38:21Z)
• Fact-Level Confidence Calibration and Self-Correction [64.40105513819272]
  We propose a Fact-Level framework that calibrates confidence to relevance-weighted correctness at the fact level. We also develop Confidence-Guided Fact-level Self-Correction ($textbfConFix$), which uses high-confidence facts within a response as additional knowledge to improve low-confidence ones.
  arXiv  Detail & Related papers  (2024-11-20T14:15:18Z)
• Provenance: A Light-weight Fact-checker for Retrieval Augmented LLM Generation Output [49.893971654861424]
  We present a light-weight approach for detecting nonfactual outputs from retrieval-augmented generation (RAG) We compute a factuality score that can be thresholded to yield a binary decision. Our experiments show high area under the ROC curve (AUC) across a wide range of relevant open source datasets.
  arXiv  Detail & Related papers  (2024-11-01T20:44:59Z)
• UncertaintyRAG: Span-Level Uncertainty Enhanced Long-Context Modeling for Retrieval-Augmented Generation [93.38604803625294]
  We present UncertaintyRAG, a novel approach for long-context Retrieval-Augmented Generation (RAG) We use Signal-to-Noise Ratio (SNR)-based span uncertainty to estimate similarity between text chunks. UncertaintyRAG outperforms baselines by 2.03% on LLaMA-2-7B, achieving state-of-the-art results.
  arXiv  Detail & Related papers  (2024-10-03T17:39:38Z)
• SFR-RAG: Towards Contextually Faithful LLMs [57.666165819196486]
  Retrieval Augmented Generation (RAG) is a paradigm that integrates external contextual information with large language models (LLMs) to enhance factual accuracy and relevance. We introduce SFR-RAG, a small LLM that is instruction-textual with an emphasis on context-grounded generation and hallucination. We also present ConBench, a new evaluation framework compiling multiple popular and diverse RAG benchmarks.
  arXiv  Detail & Related papers  (2024-09-16T01:08:18Z)
• Confidence Estimation for LLM-Based Dialogue State Tracking [9.305763502526833]
  Estimation of a model's confidence on its outputs is critical for Conversational AI systems based on large language models (LLMs) We provide an exhaustive exploration of methods, including approaches proposed for open- and closed-weight LLMs. Our findings suggest that fine-tuning open-weight LLMs can result in enhanced AUC performance, indicating better confidence score calibration.
  arXiv  Detail & Related papers  (2024-09-15T06:44:26Z)
• Improving Retrieval Augmented Language Model with Self-Reasoning [20.715106330314605]
  We propose a novel self-reasoning framework aimed at improving the reliability and traceability of RALMs. The framework involves constructing self-reason trajectories with three processes: a relevance-aware process, an evidence-aware selective process, and a trajectory analysis process. We have evaluated our framework across four public datasets to demonstrate the superiority of our method.
  arXiv  Detail & Related papers  (2024-07-29T09:05:10Z)
• SaySelf: Teaching LLMs to Express Confidence with Self-Reflective Rationales [29.33581578047835]
  SaySelf is a training framework that teaches large language models to express more accurate fine-grained confidence estimates. In addition, SaySelf directs LLMs to produce self-reflective rationales that clearly identify gaps in their parametric knowledge. We show that the generated self-reflective rationales are reasonable and can further contribute to the calibration.
  arXiv  Detail & Related papers  (2024-05-31T16:21:16Z)
• Confidence Under the Hood: An Investigation into the Confidence-Probability Alignment in Large Language Models [14.5291643644017]
  We introduce the concept of Confidence-Probability Alignment. We probe the alignment between models' internal and expressed confidence. Among the models analyzed, OpenAI's GPT-4 showed the strongest confidence-probability alignment.
  arXiv  Detail & Related papers  (2024-05-25T15:42:04Z)
• TrustScore: Reference-Free Evaluation of LLM Response Trustworthiness [58.721012475577716]
  Large Language Models (LLMs) have demonstrated impressive capabilities across various domains, prompting a surge in their practical applications. This paper introduces TrustScore, a framework based on the concept of Behavioral Consistency, which evaluates whether an LLMs response aligns with its intrinsic knowledge.
  arXiv  Detail & Related papers  (2024-02-19T21:12:14Z)
• The Calibration Gap between Model and Human Confidence in Large Language Models [14.539888672603743]
  Large language models (LLMs) need to be well-calibrated in the sense that they can accurately assess and communicate how likely it is that their predictions are correct. Recent work has focused on the quality of internal LLM confidence assessments. This paper explores the disparity between external human confidence in an LLM's responses and the internal confidence of the model.
  arXiv  Detail & Related papers  (2024-01-24T22:21:04Z)
• Democratizing LLMs: An Exploration of Cost-Performance Trade-offs in Self-Refined Open-Source Models [53.859446823312126]
  SoTA open source models of varying sizes from 7B - 65B, on average, improve 8.2% from their baseline performance. Strikingly, even models with extremely small memory footprints, such as Vicuna-7B, show a 11.74% improvement overall and up to a 25.39% improvement in high-creativity, open ended tasks.
  arXiv  Detail & Related papers  (2023-10-11T15:56:00Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.