# Entropy and Attention Dynamics in Small Language Models: A Trace-Level Structural Analysis on the TruthfulQA Benchmark

arXiv: https://arxiv.org/abs/2604.03589 · HTML: https://mlsanity.com/paper/2604.03589/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 26 of its 47 references that are in the map.
- First posted on arXiv 4 Apr 2026.

## Builds on (most cited first, 26 of 26)

- [Analyzing Multi-Head Self-Attention: Specialized Heads Do the Heavy Lifting, the Rest Can Be Pruned](https://mlsanity.com/paper/1905.09418/) (2019)
- [Measuring and Improving Consistency in Pretrained Language Models](https://mlsanity.com/paper/2102.01017/) (2021)
- [Layer by Layer: Uncovering Hidden Representations in Language Models](https://mlsanity.com/paper/2502.02013/) (2025)
- [A Survey of Confidence Estimation and Calibration in Large Language Models](https://mlsanity.com/paper/2311.08298/) (2023)
- [Analyzing the Structure of Attention in a Transformer Language Model](https://mlsanity.com/paper/1906.04284/) (2019)
- [Domain Specialization as the Key to Make Large Language Models Disruptive: A Comprehensive Survey](https://mlsanity.com/paper/2305.18703/) (2023)
- [Mobile Edge Intelligence for Large Language Models: A Contemporary Survey](https://mlsanity.com/paper/2407.18921/) (2024)
- [Factuality Challenges in the Era of Large Language Models](https://mlsanity.com/paper/2310.05189/) (2023)
- [A Survey on Uncertainty Quantification of Large Language Models: Taxonomy, Open Research Challenges, and Future Directions](https://mlsanity.com/paper/2412.05563/) (2024)
- [Uncertainty Quantification and Confidence Calibration in Large Language Models: A Survey](https://mlsanity.com/paper/2503.15850/) (2025)
- [Calibrating Language Models with Adaptive Temperature Scaling](https://mlsanity.com/paper/2409.19817/) (2024)
- [Calibration, Entropy Rates, and Memory in Language Models](https://mlsanity.com/paper/1906.05664/) (2019)
- [Vision-Language Models for Edge Networks: A Comprehensive Survey](https://mlsanity.com/paper/2502.07855/) (2025)
- [Attend First, Consolidate Later: On the Importance of Attention in Different LLM Layers](https://mlsanity.com/paper/2409.03621/) (2024)
- [Do Language Models Mirror Human Confidence? Exploring Psychological Insights to Address Overconfidence in LLMs](https://mlsanity.com/paper/2506.00582/) (2025)
- [Influences on LLM Calibration: A Study of Response Agreement, Loss Functions, and Prompt Styles](https://mlsanity.com/paper/2501.03991/) (2025)
- [Are the Hidden States Hiding Something? Testing the Limits of Factuality-Encoding Capabilities in LLMs](https://mlsanity.com/paper/2505.16520/) (2025)
- [Multi-Head Attention with Diversity for Learning Grounded Multilingual Multimodal Representations](https://mlsanity.com/paper/1910.00058/) (2019)
- [Large Language-Geometry Model: When LLM meets Equivariance](https://mlsanity.com/paper/2502.11149/) (2025)
- [EDIS: Diagnosing LLM Reasoning via Entropy Dynamics](https://mlsanity.com/paper/2602.01288/) (2026)
- [Simple Yet Effective: An Information-Theoretic Approach to Multi-LLM Uncertainty Quantification](https://mlsanity.com/paper/2507.07236/) (2025)
- [Cognitive Mirrors: Exploring the Diverse Functional Roles of Attention Heads in LLM Reasoning](https://mlsanity.com/paper/2512.10978/) (2025)
- [Normal Transformer: Extracting Surface Geometry from LiDAR Points Enhanced by Visual Semantics](https://mlsanity.com/paper/2211.10580/) (2022)
- [Detecting Errors through Ensembling Prompts (DEEP): An End-to-End LLM Framework for Detecting Factual Errors](https://mlsanity.com/paper/2406.13009/) (2024)
- [Exploring specialization and sensitivity of convolutional neural networks in the context of simultaneous image augmentations](https://mlsanity.com/paper/2503.03283/) (2025)
- [Stream: Scaling up Mechanistic Interpretability to Long Context in LLMs via Sparse Attention](https://mlsanity.com/paper/2510.19875/) (2025)

## Abstract (arXiv)

Small language models (SLMs) have been increasingly deployed in edge devices and other resource-constrained settings. However, these models make confident mispredictions and produce unstable output, making them risky for factual and decision-critical tasks. Current evaluation methodology relies on final accuracy or hallucination rates without explaining how internal model behavior affects outputs. Specifically, how entropy evolves during decoding, how attention is distributed across layers, and how hidden representations contribute to uncertainty, logical inconsistencies, and misinformation propagation are often overlooked. Consequently, this study introduces a trace-level analysis of entropy and attention dynamics in SLMs evaluated with the TruthfulQA dataset. Four models with parameter ranges of 1B-1.7B parameters were examined via token-level output entropy, attention entropy, head dispersion, and hidden-state representation. The results reflect three model classifications by entropy patterns. Deterministic models (DeepSeek-1.5B and LLaMA-1B): output entropy decreases over time. Exploratory models (Gemma-1B): with increasing entropy, and balanced models (Qwen-1.7B): have moderate and stable entropy. Also, each group has distinctively different hidden-state movement and attention dispersion patterns. The analysis demonstrates that truthfulness in SLMs emerges from structured entropy and attention dynamics. Monitoring and optimizing these internal uncertainty patterns can guide the design of a more reliable, hallucination-aware, and application-specific edge SLMs.