Imagen de Portada

From Context Shift to Stylistic Collapse: Why Training Objectives Matter More Than Scale

Fuente: arXiv

Guardado en:

Detalles Bibliográficos
Autor principal:	Mahapatra, Rohan
Formato:	Preprint
Publicado:	2026
Materias:	Computation and Language
Acceso en línea:	Acceder al recurso
Etiquetas:	Agregar Etiqueta Sin Etiquetas, Sea el primero en etiquetar este registro!

Ejemplares similares

Attention Reveals More Than Tokens: Training-Free Long-Context Reasoning with Attention-guided Retrieval
por: Zhang, Yuwei, et al.
Publicado: (2025)

Long Is More Important Than Difficult for Training Reasoning Models
por: Shen, Si, et al.
Publicado: (2025)

Learning Through Dialogue: Engagement and Efficacy Matter More Than Explanations
por: Furniturewala, Shaz, et al.
Publicado: (2026)

The Algebra of Meaning: Why Machines Need Montague More Than Moore's Law
por: Jeong, Cheonkam, et al.
Publicado: (2025)

Why Are Web AI Agents More Vulnerable Than Standalone LLMs? A Security Analysis
por: Chiang, Jeffrey Yang Fan, et al.
Publicado: (2025)

Controllable Stylistic Text Generation with Train-Time Attribute-Regularized Diffusion
por: Zhou, Fan, et al.
Publicado: (2025)

More Than Sum of Its Parts: Deciphering Intent Shifts in Multimodal Hate Speech Detection
por: Sun, Runze, et al.
Publicado: (2026)

MoreHopQA: More Than Multi-hop Reasoning
por: Schnitzler, Julian, et al.
Publicado: (2024)

More Rounds, More Noise: Why Multi-Turn Review Fails to Improve Cross-Context Verification
por: Tae-Eun, Song
Publicado: (2026)

Why Low-Resource NLP Needs More Than Cross-Lingual Transfer: Lessons Learned from Luxembourgish
por: Philippy, Fred, et al.
Publicado: (2026)

Where Matters More Than What: Decoding-aligned KV Cache Compression via Position-aware Pseudo Queries
por: Tian, Zhenxu, et al.
Publicado: (2026)

Rethinking Layer Redundancy: Calibration Matters More Than Search in LLM Depth Pruning
por: Kim, Minkyu, et al.
Publicado: (2026)

Narrative Flattening: How Post-Training Compresses Thematic, Affective, and Stylistic Variation in LLM Fiction
por: Li, Zehan, et al.
Publicado: (2026)

Why LVLMs Are More Prone to Hallucinations in Longer Responses: The Role of Context
por: Zheng, Ge, et al.
Publicado: (2025)

Confidence Should Be Calibrated More Than One Turn Deep
por: Zhang, Zhaohan, et al.
Publicado: (2026)

Should LLM Safety Be More Than Refusing Harmful Instructions?
por: Maskey, Utsav, et al.
Publicado: (2025)

Stylistic Evolution and LLM Neutrality in Singlish Language
por: Foo, Linus Tze En, et al.
Publicado: (2026)

Detecting Stylistic Fingerprints of Large Language Models
por: Bitton, Yehonatan, et al.
Publicado: (2025)

Neurobiber: Fast and Interpretable Stylistic Feature Extraction
por: Alkiek, Kenan, et al.
Publicado: (2025)

Why Are Linear RNNs More Parallelizable?
por: Merrill, William, et al.
Publicado: (2026)

Scaling with Collapse: Efficient and Predictable Training of LLM Families
por: Bergsma, Shane, et al.
Publicado: (2025)

IDIOLEX: Unified and Continuous Representations for Idiolectal and Stylistic Variation
por: Kantharuban, Anjali, et al.
Publicado: (2026)

When Large Language Models are More PersuasiveThan Incentivized Humans, and Why
por: Schoenegger, Philipp, et al.
Publicado: (2025)

More Than Meets the Eye? Uncovering the Reasoning-Planning Disconnect in Training Vision-Language Driving Models
por: Song, Xurui, et al.
Publicado: (2025)

Tokenization Is More Than Compression
por: Schmidt, Craig W., et al.
Publicado: (2024)

More is not always better? Enhancing Many-Shot In-Context Learning with Differentiated and Reweighting Objectives
por: Zhang, Xiaoqing, et al.
Publicado: (2025)

Are Any-to-Any Models More Consistent Across Modality Transfers Than Specialists?
por: Chung, Jiwan, et al.
Publicado: (2025)

More Than Catastrophic Forgetting: Integrating General Capabilities For Domain-Specific LLMs
por: Liu, Chengyuan, et al.
Publicado: (2024)

Why Slop Matters
por: Kommers, Cody, et al.
Publicado: (2025)

LLMs Struggle with Abstract Meaning Comprehension More Than Expected
por: Alhazmi, Hamoud, et al.
Publicado: (2026)

CLASE: A Hybrid Method for Chinese Legalese Stylistic Evaluation
por: Ma, Yiran Rex, et al.
Publicado: (2026)

Latent Space Interpretation for Stylistic Analysis and Explainable Authorship Attribution
por: Alshomary, Milad, et al.
Publicado: (2024)

ShortGPT: Layers in Large Language Models are More Redundant Than You Expect
por: Men, Xin, et al.
Publicado: (2024)

More Than a Score: Probing the Impact of Prompt Specificity on LLM Code Generation
por: Zi, Yangtian, et al.
Publicado: (2025)

Why Mean Pooling Works: Quantifying Second-Order Collapse in Text Embeddings
por: Hara, Tomomasa, et al.
Publicado: (2026)

Annotations Mitigate Post-Training Mode Collapse
por: Springer, Jacob Mitchell, et al.
Publicado: (2026)

Training-Free Long-Context Scaling of Large Language Models
por: An, Chenxin, et al.
Publicado: (2024)

Real World Conversational Entity Linking Requires More Than Zeroshots
por: Hoveyda, Mohanna, et al.
Publicado: (2024)

Towards a Linguistic Evaluation of Narratives: A Quantitative Stylistic Framework
por: Maisto, Alessandro
Publicado: (2026)

StylusAI: Stylistic Adaptation for Robust German Handwritten Text Generation
por: Riaz, Nauman, et al.
Publicado: (2024)