Fine-Tuned LLMs Know They Don't Know: A Parameter-Efficient Approach to Recovering Honesty

• URL: http://arxiv.org/abs/2511.12991v1
• Date: Mon, 17 Nov 2025 05:30:48 GMT
• Title: Fine-Tuned LLMs Know They Don't Know: A Parameter-Efficient Approach to Recovering Honesty
• Abstract summary: Large Language Models (LLMs) are increasingly important for safe deployment in high-stakes domains.<n>Existing recovery methods rely on data-intensive global parameter adjustments, implicitly assuming that SFT deeply corrupts the models' ability to recognize their knowledge boundaries.<n>We propose Honesty-Critical Neurons Restoration (HCNR) to surgically repair this suppressed capacity.
• Score: 48.593888815135934
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: The honesty of Large Language Models (LLMs) is increasingly important for safe deployment in high-stakes domains. However, this crucial trait is severely undermined by supervised fine-tuning (SFT), a common technique for model specialization. Existing recovery methods rely on data-intensive global parameter adjustments, implicitly assuming that SFT deeply corrupts the models' ability to recognize their knowledge boundaries. However, we observe that fine-tuned LLMs still preserve this ability; what is damaged is their capacity to faithfully express that awareness. Building on this, we propose Honesty-Critical Neurons Restoration (HCNR) to surgically repair this suppressed capacity. HCNR identifies and restores key expression-governing neurons to their pre-trained state while harmonizing them with task-oriented neurons via Hessian-guided compensation. Experiments on four QA tasks and five LLM families demonstrate that HCNR effectively recovers 33.25% of the compromised honesty while achieving at least 2.23x speedup with over 10x less data compared to baseline methods, offering a practical solution for trustworthy LLM deployment.

Related papers

• Reinforcement Learning with Metacognitive Feedback Elicits Faithful Uncertainty Expression in LLMs [60.80435138270317]
  We apply reinforcement learning with metacognitive feedback to the problem of faithful calibration.<n>Experiments show RLMF achieves generalizable, state-of-the-art FC on diverse tasks.<n> RLMF surpasses standard RL by up to 63% while enhancing models' ability to assess and express their own capability limits.
  arXiv  Detail & Related papers  (2026-06-30T17:56:01Z)
• Towards Lightweight Reliability: Using Soft Prompts for Hallucination Mitigation in Large Language Models [5.019869660665509]
  Large language models (LLMs) have seen widespread adoption across various domains, yet their reliability is frequently undermined by hallucinations.<n>In high-stakes domains, these errors can reduce trust and introduce real-world risk.<n>We present a parameter-efficient approach that uses soft prompts to mitigate hallucinated content and promote responsible abstention.
  arXiv  Detail & Related papers  (2026-05-30T23:02:33Z)
• Clipping Bottleneck: Stabilizing RLVR via Stochastic Recovery of Near-Boundary Signals [83.0127582612634]
  Near-boundary Rescue (NSR) is a minimal, plug-and-play modification that retains out-of-bound tokens to recover lost signals.<n>NSR substantially improves training stability and delivers consistent gains over strong baselines such as DAPO and GSPO.
  arXiv  Detail & Related papers  (2026-05-21T16:45:31Z)
• LLM-Confidence Reranker: A Training-Free Approach for Enhancing Retrieval-Augmented Generation Systems [27.93755705949248]
  Large language models (LLMs) have revolutionized natural language processing, yet hallucinations in knowledge-intensive tasks remain a critical challenge.<n>We propose the LLM-Confidence Reranker (LCR), a training-free, plug-and-play algorithm that enhances reranking in RAG systems.
  arXiv  Detail & Related papers  (2026-02-14T03:12:05Z)
• Towards Sample-Efficient and Stable Reinforcement Learning for LLM-based Recommendation [56.92367609590823]
  Long Chain-of-Thought (Long CoT) reasoning has shown promise in Large Language Models (LLMs)<n>We argue that Long CoT is inherently ill-suited for the sequential recommendation domain.<n>We propose RISER, a novel Reinforced Item Space Exploration framework for Recommendation.
  arXiv  Detail & Related papers  (2026-01-31T10:02:43Z)
• LLMs can Compress LLMs: Adaptive Pruning by Agents [0.0]
  Post-training pruning has emerged as a promising approach to reduce computational costs while preserving performance.<n>We introduce agent-guided pruning, where a foundation model acts as an adaptive pruning agent.<n>We evaluate our approach on Q3 models (4B and 8B parameters) at approximately 45% sparsity, demonstrating substantial improvements over structured pruning baselines.
  arXiv  Detail & Related papers  (2026-01-14T18:45:36Z)
• Mitigating LLM Hallucination via Behaviorally Calibrated Reinforcement Learning [32.32593439144886]
  Behavior-calibrated reinforcement learning allows smaller models to surpass frontier models in uncertainty quantification.<n>Our model's log-scale Accuracy-to-Hallucination Ratio gain (0.806) exceeds GPT-5's (0.207) in a challenging in-domain evaluation.
  arXiv  Detail & Related papers  (2025-12-22T22:51:48Z)
• Rethinking LLM Parametric Knowledge as Post-retrieval Confidence for Dynamic Retrieval and Reranking [23.1400319714807]
  Large Language Models (LLMs) often generate inaccurate responses (hallucinations) when faced with questions beyond their knowledge scope.<n>Retrieval-Augmented Generation (RAG) addresses this by leveraging external knowledge, but a critical challenge remains: determining whether retrieved contexts effectively enhance the models ability to answer specific queries.<n>This challenge underscores the importance of knowledge boundary awareness, which current methods-relying on discrete labels or limited signals-fail to address adequately.
  arXiv  Detail & Related papers  (2025-09-08T09:37:20Z)
• NAPER: Fault Protection for Real-Time Resource-Constrained Deep Neural Networks [18.79943285154573]
  Memory bit-flips can severely degrade accuracy of Deep Neural Networks (DNNs)<n>Traditional protection approaches like Triple Modular Redundancy (TMR) often sacrifice accuracy to maintain reliability.<n>We introduce NAPER, a novel protection approach that addresses this challenge through ensemble learning.
  arXiv  Detail & Related papers  (2025-04-09T05:37:54Z)
• 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.<n>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)
• 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)
• Iterative Self-Tuning LLMs for Enhanced Jailbreaking Capabilities [50.980446687774645]
  We introduce ADV-LLM, an iterative self-tuning process that crafts adversarial LLMs with enhanced jailbreak ability.<n>Our framework significantly reduces the computational cost of generating adversarial suffixes while achieving nearly 100% ASR on various open-source LLMs.<n>It exhibits strong attack transferability to closed-source models, achieving 99% ASR on GPT-3.5 and 49% ASR on GPT-4, despite being optimized solely on Llama3.
  arXiv  Detail & Related papers  (2024-10-24T06:36:12Z)
• Model Surgery: Modulating LLM's Behavior Via Simple Parameter Editing [63.20133320524577]
  We show that editing a small subset of parameters can effectively modulate specific behaviors of large language models (LLMs)<n>Our approach achieves reductions of up to 90.0% in toxicity on the RealToxicityPrompts dataset and 49.2% on ToxiGen.
  arXiv  Detail & Related papers  (2024-07-11T17:52:03Z)
• Decoding Compressed Trust: Scrutinizing the Trustworthiness of Efficient LLMs Under Compression [109.23761449840222]
  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.
  arXiv  Detail & Related papers  (2024-03-18T01:38:19Z)
• 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)

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.