Decoding Compressed Trust: Scrutinizing the Trustworthiness of Efficient LLMs Under Compression

• URL: http://arxiv.org/abs/2403.15447v3
• Date: Tue, 4 Jun 2024 05:40:12 GMT
• Title: Decoding Compressed Trust: Scrutinizing the Trustworthiness of Efficient LLMs Under Compression
• Authors: Junyuan Hong, Jinhao Duan, Chenhui Zhang, Zhangheng Li, Chulin Xie, Kelsey Lieberman, James Diffenderfer, Brian Bartoldson, Ajay Jaiswal, Kaidi Xu, Bhavya Kailkhura, Dan Hendrycks, Dawn Song, Zhangyang Wang, Bo Li,
• Abstract summary: This study conducts the first, thorough evaluation of leading Large Language Models (LLMs) We find that quantization is currently a more effective approach than pruning in achieving efficiency and trustworthiness simultaneously.
• Score: 109.23761449840222
• License: http://creativecommons.org/licenses/by-nc-sa/4.0/
• Abstract: Compressing high-capability Large Language Models (LLMs) has emerged as a favored strategy for resource-efficient inferences. While state-of-the-art (SoTA) compression methods boast impressive advancements in preserving benign task performance, the potential risks of compression in terms of safety and trustworthiness have been largely neglected. This study conducts the first, thorough evaluation of three (3) leading LLMs using five (5) SoTA compression techniques across eight (8) trustworthiness dimensions. Our experiments highlight the intricate interplay between compression and trustworthiness, revealing some interesting patterns. We find that quantization is currently a more effective approach than pruning in achieving efficiency and trustworthiness simultaneously. For instance, a 4-bit quantized model retains the trustworthiness of its original counterpart, but model pruning significantly degrades trustworthiness, even at 50% sparsity. Moreover, employing quantization within a moderate bit range could unexpectedly improve certain trustworthiness dimensions such as ethics and fairness. Conversely, extreme quantization to very low bit levels (3 bits) tends to reduce trustworthiness significantly. This increased risk cannot be uncovered by looking at benign performance alone, in turn, mandating comprehensive trustworthiness evaluation in practice. These findings culminate in practical recommendations for simultaneously achieving high utility, efficiency, and trustworthiness in LLMs. Code and models are available at https://decoding-comp-trust.github.io.

Related papers

• Thinking by Subtraction: Confidence-Driven Contrastive Decoding for LLM Reasoning [58.331709210563616]
  Thinking by Subtraction is a confidence-driven contrastive decoding approach.<n>A small subset of low-confidence tokens disproportionately contributes to reasoning errors and unnecessary output expansion.<n>Our method, Confidence-Driven Contrastive Decoding, detects low-confidence tokens during decoding and intervenes at these positions.
  arXiv  Detail & Related papers  (2026-02-20T14:13:22Z)
• UniComp: A Unified Evaluation of Large Language Model Compression via Pruning, Quantization and Distillation [23.560232846931456]
  We introduce UniComp, a unified evaluation framework for comparing pruning, quantization, and knowledge distillation.<n>UniComp evaluates compressed models along three dimensions: performance, reliability, and efficiency.
  arXiv  Detail & Related papers  (2026-02-09T19:20:56Z)
• Enhancing Trustworthiness with Mixed Precision: Benchmarks, Opportunities, and Challenges [12.438306093697]
  Large language models (LLMs) have shown promising performance across various tasks.<n>LLMs' autoregressive decoding process poses significant challenges for efficient deployment on existing AI hardware.
  arXiv  Detail & Related papers  (2025-11-27T14:17:43Z)
• Decomposed Trust: Exploring Privacy, Adversarial Robustness, Fairness, and Ethics of Low-Rank LLMs [2.1311014724439845]
  We study how low-rank factorization affects trustworthiness across privacy, adversarial robustness, fairness, and ethical alignment.<n>Low-rank compression preserves or improves training data privacy but weakens PII protection during conversation.<n> adversarial robustness is generally preserved and often enhanced, even under deep compression.<n>Ethical reasoning degrades in zero-shot settings but partially recovers with few-shot prompting.
  arXiv  Detail & Related papers  (2025-11-27T04:40:56Z)
• BrowseConf: Confidence-Guided Test-Time Scaling for Web Agents [58.05949210993854]
  We investigate whether search agents have the ability to communicate their own confidence through verbalized confidence scores after long sequences of actions.<n>We propose Test-Time Scaling (TTS) methods that use confidence scores to determine answer quality, encourage the model to try again until reaching a satisfactory confidence level.
  arXiv  Detail & Related papers  (2025-10-27T15:58:51Z)
• Can Confidence Estimates Decide When Chain-of-Thought Is Necessary for LLMs? [32.02698064940949]
  Chain-of-thought (CoT) prompting has emerged as a common technique for enhancing the reasoning abilities of large language models.<n>We present the first systematic study of training-free confidence estimation methods for CoT gating.
  arXiv  Detail & Related papers  (2025-10-23T21:33:28Z)
• ReFIne: A Framework for Trustworthy Large Reasoning Models with Reliability, Faithfulness, and Interpretability [23.70973331911138]
  We argue that usable reasoning systems must be trustworthy, characterized by three properties: interpretability, faithfulness, and reliability.<n>We propose ReFIne, a new training framework that integrates supervised fine-tuning with GRPO to encourage models to improve interpretability.<n>Our experimental results show that ReFIne models generate clearer and better-structured reasoning traces.
  arXiv  Detail & Related papers  (2025-10-10T07:08:44Z)
• Mind the Generation Process: Fine-Grained Confidence Estimation During LLM Generation [63.49409574310576]
  Large language models (LLMs) exhibit overconfidence, assigning high confidence scores to incorrect predictions.<n>We introduce FineCE, a novel confidence estimation method that delivers accurate, fine-grained confidence scores during text generation.<n>Our code and all baselines used in the paper are available on GitHub.
  arXiv  Detail & Related papers  (2025-08-16T13:29:35Z)
• 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)
• ConCISE: Confidence-guided Compression in Step-by-step Efficient Reasoning [75.1101108949743]
  Large Reasoning Models (LRMs) perform strongly in complex reasoning tasks via Chain-of-Thought (CoT) prompting.<n>LRMs often suffer from verbose outputs caused by redundant content, increasing computational overhead, and degrading user experience.<n>We propose ConCISE, a framework that simplifies reasoning chains by reinforcing the model's confidence during inference.
  arXiv  Detail & Related papers  (2025-05-08T01:40:40Z)
• Theoretical Insights in Model Inversion Robustness and Conditional Entropy Maximization for Collaborative Inference Systems [89.35169042718739]
  collaborative inference enables end users to leverage powerful deep learning models without exposure of sensitive raw data to cloud servers. Recent studies have revealed that these intermediate features may not sufficiently preserve privacy, as information can be leaked and raw data can be reconstructed via model inversion attacks (MIAs) This work first theoretically proves that the conditional entropy of inputs given intermediate features provides a guaranteed lower bound on the reconstruction mean square error (MSE) under any MIA. Then, we derive a differentiable and solvable measure for bounding this conditional entropy based on the Gaussian mixture estimation and propose a conditional entropy algorithm to enhance the inversion robustness
  arXiv  Detail & Related papers  (2025-03-01T07:15:21Z)
• On Verbalized Confidence Scores for LLMs [25.160810008907397]
  Uncertainty quantification for large language models (LLMs) can establish more human trust into their responses. This work focuses on asking the LLM itself to verbalize its uncertainty with a confidence score as part of its output tokens. We assess the reliability of verbalized confidence scores with respect to different datasets, models, and prompt methods.
  arXiv  Detail & Related papers  (2024-12-19T11:10:36Z)
• PUFFLE: Balancing Privacy, Utility, and Fairness in Federated Learning [2.8304839563562436]
  Training and deploying Machine Learning models that simultaneously adhere to principles of fairness and privacy poses a significant challenge. We introduce PUFFLE, a high-level parameterised approach that can help in the exploration of the balance between utility, privacy, and fairness in FL scenarios. We prove that PUFFLE can be effective across diverse datasets, models, and data distributions, reducing the model unfairness up to 75%, with a maximum reduction in the utility of 17% in the worst-case scenario.
  arXiv  Detail & Related papers  (2024-07-21T17:22:18Z)
• Synchronous Faithfulness Monitoring for Trustworthy Retrieval-Augmented Generation [96.78845113346809]
  Retrieval-augmented language models (RALMs) have shown strong performance and wide applicability in knowledge-intensive tasks. This paper proposes SynCheck, a lightweight monitor that leverages fine-grained decoding dynamics to detect unfaithful sentences. We also introduce FOD, a faithfulness-oriented decoding algorithm guided by beam search for long-form retrieval-augmented generation.
  arXiv  Detail & Related papers  (2024-06-19T16:42:57Z)
• 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)
• TernaryVote: Differentially Private, Communication Efficient, and Byzantine Resilient Distributed Optimization on Heterogeneous Data [50.797729676285876]
  We propose TernaryVote, which combines a ternary compressor and the majority vote mechanism to realize differential privacy, gradient compression, and Byzantine resilience simultaneously. We theoretically quantify the privacy guarantee through the lens of the emerging f-differential privacy (DP) and the Byzantine resilience of the proposed algorithm.
  arXiv  Detail & Related papers  (2024-02-16T16:41:14Z)
• TrustLLM: Trustworthiness in Large Language Models [446.5640421311468]
  This paper introduces TrustLLM, a comprehensive study of trustworthiness in large language models (LLMs) We first propose a set of principles for trustworthy LLMs that span eight different dimensions. Based on these principles, we establish a benchmark across six dimensions including truthfulness, safety, fairness, robustness, privacy, and machine ethics.
  arXiv  Detail & Related papers  (2024-01-10T22:07:21Z)
• U-Trustworthy Models.Reliability, Competence, and Confidence in Decision-Making [0.21756081703275998]
  We present a precise mathematical definition of trustworthiness, termed $mathcalU$-trustworthiness. Within the context of $mathcalU$-trustworthiness, we prove that properly-ranked models are inherently $mathcalU$-trustworthy. We advocate for the adoption of the AUC metric as the preferred measure of trustworthiness.
  arXiv  Detail & Related papers  (2024-01-04T04:58:02Z)
• Can LLMs Express Their Uncertainty? An Empirical Evaluation of Confidence Elicitation in LLMs [60.61002524947733]
  Previous confidence elicitation methods rely on white-box access to internal model information or model fine-tuning. This leads to a growing need to explore the untapped area of black-box approaches for uncertainty estimation. We define a systematic framework with three components: prompting strategies for eliciting verbalized confidence, sampling methods for generating multiple responses, and aggregation techniques for computing consistency.
  arXiv  Detail & Related papers  (2023-06-22T17:31:44Z)
• Just Ask for Calibration: Strategies for Eliciting Calibrated Confidence Scores from Language Models Fine-Tuned with Human Feedback [91.22679548111127]
  A trustworthy real-world prediction system should produce well-calibrated confidence scores. We show that verbalized confidences emitted as output tokens are typically better-calibrated than the model's conditional probabilities.
  arXiv  Detail & Related papers  (2023-05-24T10:12:33Z)
• Trust, but Verify: Using Self-Supervised Probing to Improve Trustworthiness [29.320691367586004]
  We introduce a new approach of self-supervised probing, which enables us to check and mitigate the overconfidence issue for a trained model. We provide a simple yet effective framework, which can be flexibly applied to existing trustworthiness-related methods in a plug-and-play manner.
  arXiv  Detail & Related papers  (2023-02-06T08:57:20Z)
• Locally Valid and Discriminative Confidence Intervals for Deep Learning Models [37.57296694423751]
  Uncertainty information should be valid (guaranteeing coverage) and discriminative (more uncertain when the expected risk is high) Most existing Bayesian methods lack frequentist coverage guarantees and usually affect model performance. We propose Locally Valid and Discriminative confidence intervals (LVD), a simple, efficient and lightweight method to construct discriminative confidence intervals (CIs) for almost any deep learning model.
  arXiv  Detail & Related papers  (2021-06-01T04:39:56Z)

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.