Cover Image

The Readout Shortcut: Positional Number Copying Dominates Arithmetic CoT Readout in Small Language Models

Fuente: arXiv

Saved in:

Bibliographic Details
Main Author:	Liu, Ming
Format:	Preprint
Published:	2026
Subjects:	Machine Learning Artificial Intelligence Computation and Language I.2.7; I.2.6
Online Access:	Acceder al recurso
Tags:	Add Tag No Tags, Be the first to tag this record!

Similar Items

Entropy-Based Measurement of Value Drift and Alignment Work in Large Language Models
by: Fadli, Samih
Published: (2025)

Self-Training Doesn't Flatten Language -- It Restructures It: Surface Markers Amplify While Deep Syntax Dies
by: Liu, Ming
Published: (2026)

Shattered Compositionality: Counterintuitive Learning Dynamics of Transformers for Arithmetic
by: Zhao, Xingyu, et al.
Published: (2026)

In-Context Fixation: When Demonstrated Labels Override Semantics in Few-Shot Classification
by: Liu, Ming
Published: (2026)

Model Collapse as Cultural Evolution
by: Guo, Dongxin, et al.
Published: (2026)

Prototype Transformer: Towards Language Model Architectures Interpretable by Design
by: Yordanov, Yordan, et al.
Published: (2026)

Turning the TIDE: Cross-Architecture Distillation for Diffusion Large Language Models
by: Zhang, Gongbo, et al.
Published: (2026)

Cognitive Load Limits in Large Language Models: Benchmarking Multi-Hop Reasoning
by: Adapala, Sai Teja Reddy
Published: (2025)

Domain-Specific Pretraining of Language Models: A Comparative Study in the Medical Field
by: Kerner, Tobias
Published: (2024)

Painless Activation Steering: An Automated, Lightweight Approach for Post-Training Large Language Models
by: Cui, Sasha, et al.
Published: (2025)

Beyond Hallucinations: A Composite Score for Measuring Reliability in Open-Source Large Language Models
by: Salla, Rohit Kumar, et al.
Published: (2025)

The Deterministic Horizon: When Extended Reasoning Fails and Tool Delegation Becomes Necessary
by: Guo, Dongxin, et al.
Published: (2026)

The Geometry of Thought: How Scale Restructures Reasoning In Large Language Models
by: Anderson, Samuel Cyrenius
Published: (2026)

Rethinking Addressing in Language Models via Contexualized Equivariant Positional Encoding
by: Zhu, Jiajun, et al.
Published: (2025)

Synthius-Mem: Brain-Inspired Hallucination-Resistant Persona Memory Achieving 94.4% Memory Accuracy and 99.6% Adversarial Robustness on LoCoMo
by: Gadzhiev, Artem, et al.
Published: (2026)

Reasoning Large Language Model Errors Arise from Hallucinating Critical Problem Features
by: Heyman, Alex, et al.
Published: (2025)

Language as a Wave Phenomenon: Semantic Phase Locking and Interference in Neural Networks
by: Yıldırım, Alper, et al.
Published: (2025)

The Anti-Ouroboros Effect: Emergent Resilience in Large Language Models from Recursive Selective Feedback
by: Adapala, Sai Teja Reddy
Published: (2025)

Alif: Advancing Urdu Large Language Models via Multilingual Synthetic Data Distillation
by: Shafique, Muhammad Ali, et al.
Published: (2025)

The Drill-Down and Fabricate Test (DDFT): A Protocol for Measuring Epistemic Robustness in Language Models
by: Baxi, Rahul
Published: (2025)

Decodable but Not Corrected by Fixed Residual-Stream Linear Steering: Evidence from Medical LLM Failure Regimes
by: Liu, Ming
Published: (2026)

ImmigrationQA: A Source-Grounded Dataset and Small-Model Adaptation for U.S. Immigration Law
by: Shportun, Nazarii
Published: (2026)

On Semantic Loss Fine-Tuning Approach for Preventing Model Collapse in Causal Reasoning
by: Deshmukh, Pratik, et al.
Published: (2026)

Quantization Undoes Alignment: Bias Emergence in Compressed LLMs Across Models and Precision Levels
by: Rath, Plawan Kumar, et al.
Published: (2026)

Control Reinforcement Learning: Interpretable Token-Level Steering of LLMs via Sparse Autoencoder Features
by: Cho, Seonglae, et al.
Published: (2026)

No Free Swap: Protocol-Dependent Layer Redundancy in Transformers
by: Garcia, Gabriel
Published: (2026)

Beyond Pass@k: Breadth-Depth Metrics for Reasoning Boundaries
by: Dragoi, Marius, et al.
Published: (2025)

Revisiting Intermediate-Layer Matching in Knowledge Distillation: Layer-Selection Strategy Doesn't Matter (Much)
by: Yu, Zony, et al.
Published: (2025)

The Last Word Often Wins: A Format Confound in Chain-of-Thought Corruption Studies
by: Garcia, Gabriel
Published: (2026)

Counterfactual Likelihood Tests for Indirect Influence in Private Reasoning Channels
by: Lorup, Alexander Boesgaard
Published: (2026)

TIAR: Trajectory-Informed Advantage Reweighting for LLM Abstention Learning
by: Pan, Muyu, et al.
Published: (2026)

Pressure-Testing Deception Probes in LLMs: Scaling, Robustness, and the Geometry of Deceptive Representations
by: Kumar, Sachin
Published: (2026)

Structured Prompt Optimization Meets Reinforcement Learning for Global and Local Interpretability over Complex Text
by: Zhou, Tianyang, et al.
Published: (2026)

AMEL: Accumulated Message Effects on LLM Judgments
by: Temkit, Sid-Ali
Published: (2026)

Forget Attention: Importance-Aware Attention Is All You Need
by: Shin, Soohyeong, et al.
Published: (2026)

Enhancing Burmese News Classification with Kolmogorov-Arnold Network Head Fine-tuning
by: Aung, Thura, et al.
Published: (2025)

MaPPO: Maximum a Posteriori Preference Optimization with Prior Knowledge
by: Lan, Guangchen, et al.
Published: (2025)

Dodo: Dynamic Contextual Compression for Decoder-only LMs
by: Qin, Guanghui, et al.
Published: (2023)

CorrSteer: Generation-Time LLM Steering via Correlated Sparse Autoencoder Features
by: Cho, Seonglae, et al.
Published: (2025)

Adapting While Learning: Grounding LLMs for Scientific Problems with Intelligent Tool Usage Adaptation
by: Lyu, Bohan, et al.
Published: (2024)