Are LLM Uncertainty and Correctness Encoded by the Same Features? A Functional Dissociation via Sparse Autoencoders

• URL: http://arxiv.org/abs/2604.19974v1
• Date: Tue, 21 Apr 2026 20:34:14 GMT
• Title: Are LLM Uncertainty and Correctness Encoded by the Same Features? A Functional Dissociation via Sparse Autoencoders
• Abstract summary: Large language models can be uncertain yet correct, or confident yet wrong, raising the question of whether their output-level uncertainty and their actual correctness are driven by the same internal mechanisms or by distinct feature populations.<n>We introduce a 2x2 framework that partitions model predictions along correctness and confidence axes, and uses sparse autoencoders to identify features associated with each dimension independently.
• Score: 10.172598963520961
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Large language models can be uncertain yet correct, or confident yet wrong, raising the question of whether their output-level uncertainty and their actual correctness are driven by the same internal mechanisms or by distinct feature populations. We introduce a 2x2 framework that partitions model predictions along correctness and confidence axes, and uses sparse autoencoders to identify features associated with each dimension independently. Applying this to Llama-3.1-8B and Gemma-2-9B, we identify three feature populations that play fundamentally different functional roles. Pure uncertainty features are functionally essential: suppressing them severely degrades accuracy. Pure incorrectness features are functionally inert: despite showing statistically significant activation differences between correct and incorrect predictions, the majority produce near-zero change in accuracy when suppressed. Confounded features that encode both signals are detrimental to output quality, and targeted suppression of them yields a 1.1% accuracy improvement and a 75% entropy reduction, with effects transferring across the ARC-Challenge and RACE benchmarks. The feature categories are also informationally distinct: the activations of just 3 confounded features from a single mid-network layer predict model correctness (AUROC ~0.79), enabling selective abstention that raises accuracy from 62% to 81% at 53% coverage. The results demonstrate that uncertainty and correctness are distinct internal phenomena, with implications for interpretability and targeted inference-time intervention.

Related papers

• From Passive Response to Proactive Correction: Enhancing LLM Robustness Against Input Fact Perturbations [12.021107313829082]
  Large language models (LLMs) produce confident yet factually incorrect responses when user inputs contain misleading premises.<n>We propose DEDUCE, a three-stage framework that transforms LLMs from passive responders into proactive error correctors.<n> Experiments on TruthfulQA, FalseQA, and our MisFactQA benchmark demonstrate that DEDUCE significantly improves both accuracy and error correction capability.
  arXiv  Detail & Related papers  (2026-08-26T15:13:50Z)
• Perplexity Can Miss SAE Feature Damage Under Quantization [0.16921396880325779]
  Quantized models are judged acceptable when perplexity or downstream accuracy remains close to the full-precision original.<n>We test this by using a frozen sparse-autoencoder (SAE) as a fixed measurement basis.
  arXiv  Detail & Related papers  (2026-06-02T01:17:05Z)
• Unveiling the Entropy Dynamics of Chain-of-Thought Reasoning [81.57028614960576]
  We show that the Confidence Region possesses two critical properties: 1) High Reliability -- answers in the confidence region become highly accurate and stable, and 2) High Redundancy -- models generate unnecessary tokens long after reaching the correct answer.<n>These properties unlock more efficient and reliable inference strategies.
  arXiv  Detail & Related papers  (2026-06-01T10:11:14Z)
• On the Limits of LLM Adaptability: Impact of Model-Internalized Priors on Annotation Task Performance [0.7303825588015752]
  We investigate how model-internalized priors interact with user-provided instructions.<n>We find that nearly two-thirds of zero-shot errors are resistant to correction.<n>High-confidence errors prove especially resistant to correction.
  arXiv  Detail & Related papers  (2026-05-30T01:21:14Z)
• Discovery of Hidden Miscalibration Regimes [52.452902154360565]
  Models may be systematically overconfident on some kinds of inputs and underconfident on others.<n>We define the corresponding miscalibration field and propose a diagnostic framework for estimating it.<n>Our approach learns a calibration-aware representation of the input space and estimates signed local miscalibration by kernel smoothing in the learned geometry.
  arXiv  Detail & Related papers  (2026-05-13T13:07:50Z)
• Evolving Multi-Channel Confidence-Aware Activation Functions for Missing Data with Channel Propagation [1.3999481573773072]
  In neural networks, activation functions significantly affect performance yet do not account for missingness indicators or confidence scores.<n>We propose Three-Channel Evolved Activations (3C-EA) and ChannelProp, an algorithm that deterministically propagates missingness and confidence values.<n>Results indicate that integrating missingness and confidence inputs into the activation search improves classification performance under missingness.
  arXiv  Detail & Related papers  (2026-02-14T19:52:10Z)
• Same Answer, Different Representations: Hidden instability in VLMs [65.36933543377346]
  We introduce a representation-aware and frequency-aware evaluation framework that measures internal embedding drift, spectral sensitivity, and structural smoothness.<n>We apply this framework to modern Vision Language Models (VLMs) across the SEEDBench, MMMU, and POPE datasets.
  arXiv  Detail & Related papers  (2026-02-06T12:24:26Z)
• Geometric Calibration and Neutral Zones for Uncertainty-Aware Multi-Class Classification [0.0]
  This work bridges information geometry and statistical learning, offering formal guarantees for uncertainty-aware classification in applications requiring rigorous validation.<n> Empirical validation on Adeno-Associated Virus classification demonstrates that the two-stage framework captures 72.5% of errors while deferring 34.5% of samples, reducing automated decision error rates from 16.8% to 6.9%.
  arXiv  Detail & Related papers  (2025-11-26T01:29:49Z)
• DUAL: Dynamic Uncertainty-Aware Learning [19.100858792977807]
  We propose DynamicUncertainty-Aware Learning (DUAL), a unified framework that effectively handles feature uncertainty in both single-modal and multi-modal scenarios.<n>DUAL introduces three key innovations: Dynamic Feature Uncertainty Modeling, Adaptive Distribution-Aware Modulation, and Uncertainty-aware Cross-Modal Relationship.
  arXiv  Detail & Related papers  (2025-05-21T18:50:15Z)
• On the Efficacy of Generalization Error Prediction Scoring Functions [33.24980750651318]
  Generalization error predictors (GEPs) aim to predict model performance on unseen distributions by deriving dataset-level error estimates from sample-level scores. We rigorously study the effectiveness of popular scoring functions (confidence, local manifold smoothness, model agreement) independent of mechanism choice.
  arXiv  Detail & Related papers  (2023-03-23T18:08:44Z)
• Robustness and Accuracy Could Be Reconcilable by (Proper) Definition [109.62614226793833]
  The trade-off between robustness and accuracy has been widely studied in the adversarial literature. We find that it may stem from the improperly defined robust error, which imposes an inductive bias of local invariance. By definition, SCORE facilitates the reconciliation between robustness and accuracy, while still handling the worst-case uncertainty.
  arXiv  Detail & Related papers  (2022-02-21T10:36:09Z)
• Uncertainty-Aware Camera Pose Estimation from Points and Lines [101.03675842534415]
  Perspective-n-Point-and-Line (Pn$PL) aims at fast, accurate and robust camera localizations with respect to a 3D model from 2D-3D feature coordinates.
  arXiv  Detail & Related papers  (2021-07-08T15:19:36Z)
• Removing Spurious Features can Hurt Accuracy and Affect Groups Disproportionately [83.68135652247496]
  A natural remedy is to remove spurious features from the model. We show that removal of spurious features can decrease accuracy due to inductive biases. We also show that robust self-training can remove spurious features without affecting the overall accuracy.
  arXiv  Detail & Related papers  (2020-12-07T23:08:59Z)
• Unlabelled Data Improves Bayesian Uncertainty Calibration under Covariate Shift [100.52588638477862]
  We develop an approximate Bayesian inference scheme based on posterior regularisation. We demonstrate the utility of our method in the context of transferring prognostic models of prostate cancer across globally diverse populations.
  arXiv  Detail & Related papers  (2020-06-26T13:50:19Z)
• Triple Wins: Boosting Accuracy, Robustness and Efficiency Together by Enabling Input-Adaptive Inference [119.19779637025444]
  Deep networks were recently suggested to face the odds between accuracy (on clean natural images) and robustness (on adversarially perturbed images) This paper studies multi-exit networks associated with input-adaptive inference, showing their strong promise in achieving a "sweet point" in cooptimizing model accuracy, robustness and efficiency.
  arXiv  Detail & Related papers  (2020-02-24T00:40:22Z)

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.