Buchumschlag

Self-Training Doesn't Flatten Language -- It Restructures It: Surface Markers Amplify While Deep Syntax Dies

Fuente: arXiv

Gespeichert in:

Bibliographische Detailangaben
1. Verfasser:	Liu, Ming
Format:	Preprint
Veröffentlicht:	2026
Schlagworte:	Computation and Language Artificial Intelligence Machine Learning I.2.7; I.2.6
Online-Zugang:	Acceder al recurso
Tags:	Tag hinzufügen Keine Tags, Fügen Sie den ersten Tag hinzu!

Ähnliche Einträge

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

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

Truth as a Compression Artifact in Language Model Training
von: Krestnikov, Konstantin
Veröffentlicht: (2026)

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

The Readout Shortcut: Positional Number Copying Dominates Arithmetic CoT Readout in Small Language Models
von: Liu, Ming
Veröffentlicht: (2026)

Machine Unlearning for Masked Diffusion Language Models
von: Lee, Georu, et al.
Veröffentlicht: (2026)

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

Intention Collapse: Intention-Level Metrics for Reasoning in Language Models
von: Vera, Patricio
Veröffentlicht: (2026)

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

Assessing Large Language Models on Islamic Legal Reasoning: Evidence from Inheritance Law Evaluation
von: Bouchekif, Abdessalam, et al.
Veröffentlicht: (2025)

SECURA: Sigmoid-Enhanced CUR Decomposition with Uninterrupted Retention and Low-Rank Adaptation in Large Language Models
von: Zhang, Yuxuan
Veröffentlicht: (2025)

A Hierarchical Error Framework for Reliable Automated Coding in Communication Research: Applications to Health and Political Communication
von: Zhao, Zhilong, et al.
Veröffentlicht: (2025)

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

Automated CAD Modeling Sequence Generation from Text Descriptions via Transformer-Based Large Language Models
von: Liao, Jianxing, et al.
Veröffentlicht: (2025)

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

Align and Shine: Building High-Quality Sentence-Aligned Corpora for Multilingual Text Simplification
von: Hilasaca, Kenji, et al.
Veröffentlicht: (2026)

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

Whether, Not Which: Mechanistic Interpretability Reveals Dissociable Affect Reception and Emotion Categorization in LLMs
von: Keeman, Michael
Veröffentlicht: (2026)

KAConvText: Novel Approach to Burmese Sentence Classification using Kolmogorov-Arnold Convolution
von: Thu, Ye Kyaw, et al.
Veröffentlicht: (2025)

When Persuasion Overrides Truth in Multi-Agent LLM Debates: Introducing a Confidence-Weighted Persuasion Override Rate (CW-POR)
von: Agarwal, Mahak, et al.
Veröffentlicht: (2025)

Why Models Know But Don't Say: Chain-of-Thought Faithfulness Divergence Between Thinking Tokens and Answers in Open-Weight Reasoning Models
von: Young, Richard J.
Veröffentlicht: (2026)

The Pragmatic Persona: Discovering LLM Persona through Bridging Inference
von: Yang, Jisoo, et al.
Veröffentlicht: (2026)

RMGAP: Benchmarking the Generalization of Reward Models across Diverse Preferences
von: Zhou, Yangyang, et al.
Veröffentlicht: (2026)

UrduBench: An Urdu Reasoning Benchmark using Contextually Ensembled Translations with Human-in-the-Loop
von: Shafique, Muhammad Ali, et al.
Veröffentlicht: (2026)

Can AI Read Between The Lines? Benchmarking LLMs On Financial Nuance
von: Kubica, Dominick, et al.
Veröffentlicht: (2025)

Eyla: Toward an Identity-Anchored LLM Architecture with Integrated Biological Priors -- Vision, Implementation Attempt, and Lessons from AI-Assisted Development
von: Aditto, Arif
Veröffentlicht: (2026)

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

JURY-RL: Votes Propose, Proofs Dispose for Label-Free RLVR
von: Chen, Xinjie, et al.
Veröffentlicht: (2026)

MeMo: Towards Language Models with Associative Memory Mechanisms
von: Zanzotto, Fabio Massimo, et al.
Veröffentlicht: (2025)

OSCToM: RL-Guided Adversarial Generation for High-Order Theory of Mind
von: Srishty, Sharmin Sultana, et al.
Veröffentlicht: (2026)

BitCal-TTS: Bit-Calibrated Test-Time Scaling for Quantized Reasoning Models
von: Patarlapalli, Sai Babu, et al.
Veröffentlicht: (2026)

Planning vs Reasoning: Ablations to Test Capabilities of LoRA layers
von: Redkar, Neel
Veröffentlicht: (2024)

CoE: Collaborative Entropy for Uncertainty Quantification in Agentic Multi-LLM Systems
von: Sun, Kangkang, et al.
Veröffentlicht: (2026)

Curveball Steering: The Right Direction To Steer Isn't Always Linear
von: Raval, Shivam, et al.
Veröffentlicht: (2026)

Instruction-Level Weight Shaping: A Framework for Self-Improving AI Agents
von: Costa, Rimom
Veröffentlicht: (2025)

DPO Unchained: Your Training Algorithm is Secretly Disentangled in Human Choice Theory
von: Zhou, Wenxuan, et al.
Veröffentlicht: (2025)

Shallow Robustness, Deep Vulnerabilities: Multi-Turn Evaluation of Medical LLMs
von: Manczak, Blazej, et al.
Veröffentlicht: (2025)

Model Collapse as Cultural Evolution
von: Guo, Dongxin, et al.
Veröffentlicht: (2026)

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

Generalizing Numerical Reasoning in Table Data through Operation Sketches and Self-Supervised Learning
von: Cho, Hanjun, et al.
Veröffentlicht: (2026)