Weakly Supervised Distillation of Hallucination Signals into Transformer Representations

• URL: http://arxiv.org/abs/2604.06277v1
• Date: Tue, 07 Apr 2026 08:14:48 GMT
• Title: Weakly Supervised Distillation of Hallucination Signals into Transformer Representations
• Abstract summary: hallucination detection methods for large language models rely on external verification at inference time.<n>We show that hallucination detection signals can be distilled into transformer representations, enabling internal detection without any external verification at inference time.<n> Transformer-based probes achieve the strongest discrimination, with M2 performing best on 5-fold average AUC/F1, and M3 performing best on both single-fold validation and held-out test evaluation.
• Score: 0.0
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Existing hallucination detection methods for large language models (LLMs) rely on external verification at inference time, requiring gold answers, retrieval systems, or auxiliary judge models. We ask whether this external supervision can instead be distilled into the model's own representations during training, enabling hallucination detection from internal activations alone at inference time. We introduce a weak supervision framework that combines three complementary grounding signals: substring matching, sentence embedding similarity, and an LLM as a judge verdict to label generated responses as grounded or hallucinated without human annotation. Using this framework, we construct a 15000-sample dataset from SQuAD v2 (10500 train/development samples and a separate 5000-sample test set), where each example pairs a LLaMA-2-7B generated answer with its full per-layer hidden states and structured hallucination labels. We then train five probing classifiers: ProbeMLP (M0), LayerWiseMLP (M1), CrossLayerTransformer (M2), HierarchicalTransformer (M3), and CrossLayerAttentionTransformerV2 (M4), directly on these hidden states, treating external grounding signals as training-time supervision only. Our central hypothesis is that hallucination detection signals can be distilled into transformer representations, enabling internal detection without any external verification at inference time. Results support this hypothesis. Transformer-based probes achieve the strongest discrimination, with M2 performing best on 5-fold average AUC/F1, and M3 performing best on both single-fold validation and held-out test evaluation. We also benchmark inference efficiency: probe latency ranges from 0.15 to 5.62 ms (batched) and 1.55 to 6.66 ms (single sample), while end-to-end generation plus probe throughput remains approximately 0.231 queries per second, indicating negligible practical overhead.

Related papers

• Architecture-Dependent Causal Transfer of Activation States Across Large Language Models [51.56484100374058]
  Internal language activation states can be transferred causally between different large language models.<n>Injecting projected activations into a target model during generation produces a statistically significant, preregistered causal effect.<n>We conclude that end-to-end activation-state transfer between LLMs, as currently implemented, is architecture-dependent rather than universal.
  arXiv  Detail & Related papers  (2026-08-17T09:53:27Z)
• UniProbe: A Learnable Token-Level Hallucination Detector for Large VLMs using Multi-Structural Internal Representations [61.85916280854257]
  Large Vision-Language Models (LVLMs) achieve impressive visual reasoning and dialogue capabilities, yet frequently hallucinate content unsupported by the visual input.<n>We introduce textbfUniProbe, a lightweight, unified, learnable detector that models a frozen LVLM's heterogeneous computational trace from a single forward pass.
  arXiv  Detail & Related papers  (2026-08-11T12:01:59Z)
• Perception, Verdict, and Evolution: Hindsight-Driven Self-Refining Forensics Agent for AI-Generated Image Detection [80.15317858033534]
  ForeAgent is an agentic forensics framework for AI-generated image detection with iterative self-evolution.<n>ForeAgent achieves state-of-the-art performance on the Chameleon benchmark, reaching 82.18% accuracy.<n>ForeAgent produces more consistent and causally grounded reasoning compared to GPT-5 and GPT-5-mini.
  arXiv  Detail & Related papers  (2026-06-25T02:59:33Z)
• Layer-Resolved Optimal Transport for Hallucination Detection in NMT and Abstractive Summarization [0.0]
  Optimal transport (OT) has been shown to detect hallucinations in neural machine translation (NMT)<n>We extend this analysis to all six decoder layers of the Fairseq DE-EN model.
  arXiv  Detail & Related papers  (2026-06-11T11:30:08Z)
• The Saturation Trap and the Subjectivity of Intervention Timing: Why Affect-Based Triggers and LLM Judges Fail to Time Interventions on Autonomous Agents [0.0]
  Intervention timing is a low-reliability construct, making single-annotator F1 an unsuitable optimization target.<n>Our contribution is the joint mapping of this problem across human inter-rater reliability, four detector architectures, a cross-model LLM-judge sweep, and a reproduced saturation effect.
  arXiv  Detail & Related papers  (2026-06-02T23:54:27Z)
• Before the First Token: Scale-Dependent Emergence of Hallucination Signals in Autoregressive Language Models [6.5607014897534865]
  We study the temporal dynamics of hallucination-indicative internal representations across 7 autoregressive transformers.<n>We identify a scale-dependent phase transition: models below 400M parameters show chance-level probe accuracy at every generation position.<n>At the 7B scale, we observe a striking dissociation: Pythia-6.9B produces a flat temporal profile, while instruction-tuned Qwen2.5-7B shows a dominant pre-generation effect.
  arXiv  Detail & Related papers  (2026-03-20T02:30:01Z)
• Kestrel: Grounding Self-Refinement for LVLM Hallucination Mitigation [86.37623966653688]
  Large vision-language models (LVLMs) have become increasingly strong but remain prone to hallucinations in multimodal tasks.<n>As training these LVLMs to avoid hallucinations becomes prohibitively expensive for larger models, training-free methods offer a cheap and flexible solution to this problem.<n>We propose Kestrel, a training-free framework for LVLM hallucination mitigation that combines an explicit visual-grounding agent with evidence-verified self-refinement mechanism.
  arXiv  Detail & Related papers  (2026-03-17T15:30:47Z)
• Aligning to Illusions: Choice Blindness in Human and AI Feedback [2.912535004465788]
  In a human choice blindness study, 91% of surreptitiously swapped preferences go undetected, extending choice blindness to third-person evaluative comparison of unfamiliar text.<n>We find detection relies on shallow text matching rather than genuine self-monitoring.<n>In a dose-response experiment across two architectures, one-sixth to one-third of labels must be corrupted before the reward signal halves.<n>A Best-of-N evaluation confirms this translates to downstream policy degradation.
  arXiv  Detail & Related papers  (2026-03-09T14:10:36Z)
• Predictive Coding and Information Bottleneck for Hallucination Detection in Large Language Models [0.8552050317027305]
  Hallucinations in Large Language Models (LLMs) remain a critical barrier to high-stakes deployment.<n>We introduce [Model Name], a hybrid detection framework that combines neuroscience-inspired signal design with supervised machine learning.
  arXiv  Detail & Related papers  (2026-01-22T05:00:21Z)
• DRIFT: Detecting Representational Inconsistencies for Factual Truthfulness [5.785021425715989]
  LLMs often produce fluent but incorrect answers, yet detecting such hallucinations typically requires multiple sampling passes or post-hoc verification.<n>We propose a lightweight probe to read these signals directly from hidden states.<n>We develop an LLM router that answers confident queries immediately while delegating uncertain ones to stronger models.
  arXiv  Detail & Related papers  (2026-01-20T18:16:10Z)
• A Single Direction of Truth: An Observer Model's Linear Residual Probe Exposes and Steers Contextual Hallucinations [0.0]
  A generator-agnostic observer model detects hallucinations via a single forward pass and a linear probe on its residual stream.<n>This probe isolates a single, linear direction separating hallucinated from faithful text, outperforming baselines by 5-27 points.
  arXiv  Detail & Related papers  (2025-07-31T03:26:57Z)
• Backdoor Cleaning without External Guidance in MLLM Fine-tuning [76.82121084745785]
  Believe Your Eyes (BYE) is a data filtering framework that leverages attention entropy patterns as self-supervised signals to identify and filter backdoor samples.<n>It achieves near-zero attack success rates while maintaining clean-task performance.
  arXiv  Detail & Related papers  (2025-05-22T17:11:58Z)
• DC-Solver: Improving Predictor-Corrector Diffusion Sampler via Dynamic Compensation [68.55191764622525]
  Diffusion models (DPMs) have shown remarkable performance in visual synthesis but are computationally expensive due to the need for multiple evaluations during the sampling. Recent predictor synthesis-or diffusion samplers have significantly reduced the required number of evaluations, but inherently suffer from a misalignment issue. We introduce a new fast DPM sampler called DC-CPRr, which leverages dynamic compensation to mitigate the misalignment.
  arXiv  Detail & Related papers  (2024-09-05T17:59:46Z)
• Cross-Shaped Windows Transformer with Self-supervised Pretraining for Clinically Significant Prostate Cancer Detection in Bi-parametric MRI [6.930082824262643]
  We introduce a novel end-to-end Cross-Shaped windows (CSwin) transformer UNet model, CSwin UNet, to detect clinically significant prostate cancer (csPCa) in prostate bi-parametric MR imaging (bpMRI) Using a large prostate bpMRI dataset with 1500 patients, we first pretrain CSwin transformer using multi-task self-supervised learning to improve data-efficiency and network generalizability. Five-fold cross validation shows that self-supervised CSwin UNet achieves 0.888 AUC and 0.545 Average Precision (AP), significantly outperforming four comparable models (Swin U
  arXiv  Detail & Related papers  (2023-04-30T04:40:32Z)
• The Lazy Neuron Phenomenon: On Emergence of Activation Sparsity in Transformers [59.87030906486969]
  This paper studies the curious phenomenon for machine learning models with Transformer architectures that their activation maps are sparse. We show that sparsity is a prevalent phenomenon that occurs for both natural language processing and vision tasks. We discuss how sparsity immediately implies a way to significantly reduce the FLOP count and improve efficiency for Transformers.
  arXiv  Detail & Related papers  (2022-10-12T15:25:19Z)
• Disentangle Your Dense Object Detector [82.22771433419727]
  Deep learning-based dense object detectors have achieved great success in the past few years and have been applied to numerous multimedia applications such as video understanding. However, the current training pipeline for dense detectors is compromised to lots of conjunctions that may not hold. We propose Disentangled Dense Object Detector (DDOD), in which simple and effective disentanglement mechanisms are designed and integrated into the current state-of-the-art detectors.
  arXiv  Detail & Related papers  (2021-07-07T00:52:16Z)

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.