paper-with-me

홈 › Papers

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

2026-04-04 · Adeyemi Adeseye, Aisvarya Adeseye, Hannu Tenhunen, Jouni Isoaho 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.

📄 PDF Abstract BibTeX arXiv:2604.03589

Code (0)

등록된 구현이 없습니다.

Similar Papers 제목 키워드 기반

Permutation, Multiscale and Modified Multiscale Entropies a Natural Complexity for Low-High Infection Level Intracellular Viral Reaction Kinetics

2017-03-28

Viral infectious diseases, such as HIV virus growth, cause an important health concern. Study of intracellular viral processes can provide us to develop drug and understanding the drug dose to decrease the HIV virus in d…

The Stepwise Informativeness Assumption: Why are Entropy Dynamics and Reasoning Correlated in LLMs?

2026-03-11 · Mar Gonzàlez I Català, Haitz Sáez de Ocáriz Borde, George D. Montañez, Pietro Liò arxiv

Recent work uses entropy-based signals at multiple representation levels to study reasoning in large language models, but the field remains largely empirical. A central unresolved puzzle is why internal entropy dynamics,…

Self-attention Networks Localize When QK-eigenspectrum Concentrates

2024-02-03 · Han Bao, Ryuichiro Hataya, Ryo Karakida

The self-attention mechanism prevails in modern machine learning. It has an interesting functionality of adaptively selecting tokens from an input sequence by modulating the degree of attention localization, which many r…

DiffAdapt: Difficulty-Adaptive Reasoning for Token-Efficient LLM Inference

2025-10-22 · Xiang Liu, Xuming Hu, Xiaowen Chu, Eunsol Choi arxiv

Recent reasoning Large Language Models (LLMs) demonstrate remarkable problem-solving abilities but often generate long thinking traces whose utility is unclear. Our work aims to improve their efficiency, enabling them to…

R-Stitch: Dynamic Trajectory Stitching for Efficient Reasoning

2025-07-23 · Zhuokun Chen, Zeren Chen, Jiahao He, Lu Sheng 외 arxiv

Chain-of-thought (CoT) enhances the problem-solving ability of large language models (LLMs) but incurs substantial inference cost due to long autoregressive trajectories. Existing acceleration strategies either shorten t…