Image de couverture de livre

Quantisation Reshapes the Metacognitive Geometry of Language Models

Fuente: arXiv

Enregistré dans:

Détails bibliographiques
Auteur principal:	Cacioli, Jon-Paul
Format:	Preprint
Publié:	2026
Sujets:	Computation and Language
Accès en ligne:	Acceder al recurso
Tags:	Ajouter un tag Pas de tags, Soyez le premier à ajouter un tag!

Documents similaires

Before You Interpret the Profile: Validity Scaling for LLM Metacognitive Self-Report
par: Cacioli, Jon-Paul
Publié: (2026)

The Metacognitive Monitoring Battery: A Cross-Domain Benchmark for LLM Self-Monitoring
par: Cacioli, Jon-Paul
Publié: (2026)

K-Way Energy Probes for Metacognition Reduce to Softmax in Discriminative Predictive Coding Networks
par: Cacioli, Jon-Paul
Publié: (2026)

Weber's Law in Transformer Magnitude Representations: Efficient Coding, Representational Geometry, and Psychophysical Laws in Language Models
par: Cacioli, Jon-Paul
Publié: (2026)

Do LLMs Know What They Know? Measuring Metacognitive Efficiency with Signal Detection Theory
par: Cacioli, Jon-Paul
Publié: (2026)

Same Geometry, Opposite Noise: Transformer Magnitude Representations Lack Scalar Variability
par: Cacioli, Jon-Paul
Publié: (2026)

Repetition Without Exclusivity: Scale Sensitivity of Referential Mechanisms in Child-Scale Language Models
par: Cacioli, Jon-Paul
Publié: (2026)

Categorical Perception in Large Language Model Hidden States: Structural Warping at Digit-Count Boundaries
par: Cacioli, Jon-Paul
Publié: (2026)

Beyond the Mean: Within-Model Reliable Change Detection for LLM Evaluation
par: Cacioli, Jon-Paul
Publié: (2026)

Screen Before You Interpret: A Portable Validity Protocol for Benchmark-Based LLM Confidence Signals
par: Cacioli, Jon-Paul
Publié: (2026)

Verbal Confidence Saturation in 3-9B Open-Weight Instruction-Tuned LLMs: A Pre-Registered Psychometric Validity Screen
par: Cacioli, Jon-Paul
Publié: (2026)

Domain-level metacognitive monitoring in frontier LLMs: A 33-model atlas
par: Cacioli, Jon-Paul
Publié: (2026)

Concurrent Criterion Validation of a Validity Screen for LLM Confidence Signals via Selective Prediction
par: Cacioli, Jon-Paul
Publié: (2026)

LLMs as Signal Detectors: Sensitivity, Bias, and the Temperature-Criterion Analogy
par: Cacioli, Jon-Paul
Publié: (2026)

Instruction Complexity Induces Positional Collapse in Adversarial LLM Evaluation
par: Cacioli, Jon-Paul
Publié: (2026)

Below-Chance Blindness: Prompted Underperformance in Small LLMs Produces Positional Bias Rather than Answer Avoidance
par: Cacioli, Jon-Paul
Publié: (2026)

Option-Order Randomisation Reveals a Distributional Position Attractor in Prompted Sandbagging
par: Cacioli, Jon-Paul
Publié: (2026)

Cross-Entropy Is Load-Bearing: A Pre-Registered Scope Test of the K-Way Energy Probe on Bidirectional Predictive Coding
par: Cacioli, Jon-Paul
Publié: (2026)

Distilling Self-Consistency into Verbal Confidence: A Pre-Registered Negative Result and Post-Hoc Rescue on Gemma 3 4B
par: Cacioli, Jon-Paul
Publié: (2026)

Exemplar Retrieval Without Overhypothesis Induction: Limits of Distributional Sequence Learning in Early Word Learning
par: Cacioli, Jon-Paul
Publié: (2026)

Context Structure Reshapes the Representational Geometry of Language Models
par: Hosseini, Eghbal A., et autres
Publié: (2026)

Think$^{2}$: Grounded Metacognitive Reasoning in Large Language Models
par: Elenjical, Abraham Paul, et autres
Publié: (2026)

Decomposing and Steering Functional Metacognition in Large Language Models
par: Li, Yanshi, et autres
Publié: (2026)

Metacognitive Myopia in Large Language Models
par: Scholten, Florian, et autres
Publié: (2024)

Improving Metacognition and Uncertainty Communication in Language Models
par: Steyvers, Mark, et autres
Publié: (2025)

Metacognitive Retrieval-Augmented Large Language Models
par: Zhou, Yujia, et autres
Publié: (2024)

REFLEX: Metacognitive Reasoning for Reflective Zero-Shot Robotic Planning with Large Language Models
par: Lin, Wenjie, et autres
Publié: (2025)

Metacognitive Prompting Improves Understanding in Large Language Models
par: Wang, Yuqing, et autres
Publié: (2023)

ResT: Reshaping Token-Level Policy Gradients for Tool-Use Large Language Models
par: Lin, Zihan, et autres
Publié: (2025)

Language Models Are Capable of Metacognitive Monitoring and Control of Their Internal Activations
par: Ji-An, Li, et autres
Publié: (2025)

Small Language Models Reshape Higher Education: Courses, Textbooks, and Teaching
par: Zhang, Jian, et autres
Publié: (2025)

MDToC: Metacognitive Dynamic Tree of Concepts for Boosting Mathematical Problem-Solving of Large Language Models
par: Ta, Tung Duong, et autres
Publié: (2025)

Same Model, Different Weakness: How Language and Modality Reshape the Jailbreak Attack Surface in Frontier MLLMs
par: Ford, Casey, et autres
Publié: (2026)

RIRO: Reshaping Inputs, Refining Outputs Unlocking the Potential of Large Language Models in Data-Scarce Contexts
par: Hamdi, Ali, et autres
Publié: (2024)

MetaMind: Modeling Human Social Thoughts with Metacognitive Multi-Agent Systems
par: Zhang, Xuanming, et autres
Publié: (2025)

Disentangling Geometry, Performance, and Training in Language Models
par: Kulkarni, Atharva, et autres
Publié: (2026)

Self-Anchoring Calibration Drift in Large Language Models: How Multi-Turn Conversations Reshape Model Confidence
par: Harshavardhan
Publié: (2026)

Hallucinations Undermine Trust; Metacognition is a Way Forward
par: Yona, Gal, et autres
Publié: (2026)

How Value Induction Reshapes LLM Behaviour
par: Arora, Arnav, et autres
Publié: (2026)

Mapping the Schedule x Bit-Width Boundary in Sub-100M Quantisation-Aware Training
par: Thomassen, Christian Brandt
Publié: (2026)