Image de couverture de livre

From Signal Degradation to Computation Collapse: Uncovering the Two Failure Modes of LLM Quantization

Fuente: arXiv

Enregistré dans:

Détails bibliographiques
Auteurs principaux:	Zhou, Chenxi, Cao, Pengfei, Li, Jiang, Yu, Bohan, Ye, Jinyu, Zhao, Jun, Liu, Kang
Format:	Preprint
Publié:	2026
Sujets:	Computation and Language Artificial Intelligence Machine Learning
Accès en ligne:	Acceder al recurso
Tags:	Ajouter un tag Pas de tags, Soyez le premier à ajouter un tag!

Documents similaires

Task-Stratified Knowledge Scaling Laws for Post-Training Quantized Large Language Models
par: Zhou, Chenxi, et autres
Publié: (2025)

Knowledge in Superposition: Unveiling the Failures of Lifelong Knowledge Editing for Large Language Models
par: Hu, Chenhui, et autres
Publié: (2024)

Judging Against the Reference: Uncovering Knowledge-Driven Failures in LLM-Judges on QA Evaluation
par: Lee, Dongryeol, et autres
Publié: (2026)

Catastrophic Failure of LLM Unlearning via Quantization
par: Zhang, Zhiwei, et autres
Publié: (2024)

The Knowledge Microscope: Features as Better Analytical Lenses than Neurons
par: Chen, Yuheng, et autres
Publié: (2025)

Lost in Diffusion: Uncovering Hallucination Patterns and Failure Modes in Diffusion Large Language Models
par: Guo, Zhengnan, et autres
Publié: (2026)

FlatQuant: Flatness Matters for LLM Quantization
par: Sun, Yuxuan, et autres
Publié: (2024)

Verbalized Sampling: How to Mitigate Mode Collapse and Unlock LLM Diversity
par: Zhang, Jiayi, et autres
Publié: (2025)

Revealing the Deceptiveness of Knowledge Editing: A Mechanistic Analysis of Superficial Editing
par: Xie, Jiakuan, et autres
Publié: (2025)

EvoEdit: Lifelong Free-Text Knowledge Editing through Latent Perturbation Augmentation and Knowledge-driven Parameter Fusion
par: Cao, Pengfei, et autres
Publié: (2025)

Revisiting On-Policy Distillation: Empirical Failure Modes and Simple Fixes
par: Fu, Yuqian, et autres
Publié: (2026)

ProbeLLM: Automating Principled Diagnosis of LLM Failures
par: Huang, Yue, et autres
Publié: (2026)

Quantization Meets Reasoning: Exploring LLM Low-Bit Quantization Degradation for Mathematical Reasoning
par: Li, Zhen, et autres
Publié: (2025)

Reasoning Beyond Chain-of-Thought: A Latent Computational Mode in Large Language Models
par: He, Zhenghao, et autres
Publié: (2026)

WilKE: Wise-Layer Knowledge Editor for Lifelong Knowledge Editing
par: Hu, Chenhui, et autres
Publié: (2024)

Towards Atoms of Large Language Models
par: Hu, Chenhui, et autres
Publié: (2025)

Knowledge Localization: Mission Not Accomplished? Enter Query Localization!
par: Chen, Yuheng, et autres
Publié: (2024)

DTELS: Towards Dynamic Granularity of Timeline Summarization
par: Zhang, Chenlong, et autres
Publié: (2024)

From Confidence to Collapse in LLM Factual Robustness
par: Fastowski, Alina, et autres
Publié: (2025)

MEMLA: Enhancing Multilingual Knowledge Editing with Neuron-Masked Low-Rank Adaptation
par: Xie, Jiakuan, et autres
Publié: (2024)

A Troublemaker with Contagious Jailbreak Makes Chaos in Honest Towns
par: Men, Tianyi, et autres
Publié: (2024)

Unlocking the Future: Exploring Look-Ahead Planning Mechanistic Interpretability in Large Language Models
par: Men, Tianyi, et autres
Publié: (2024)

One Mind, Many Tongues: A Deep Dive into Language-Agnostic Knowledge Neurons in Large Language Models
par: Cao, Pengfei, et autres
Publié: (2024)

From Documents to Database: Failure Modes for Industrial Assets
par: Kabakci-Zorlu, Duygu, et autres
Publié: (2025)

SR-KI: Scalable and Real-Time Knowledge Integration into LLMs via Supervised Attention
par: Yu, Bohan, et autres
Publié: (2025)

Sample-efficient LLM Optimization with Reset Replay
par: Liu, Zichuan, et autres
Publié: (2025)

Annotations Mitigate Post-Training Mode Collapse
par: Springer, Jacob Mitchell, et autres
Publié: (2026)

Escaping Mode Collapse in LLM Generation via Geometric Regulation
par: Du, Xin, et autres
Publié: (2026)

Adaptive Stopping for Multi-Turn LLM Reasoning
par: Zhou, Xiaofan, et autres
Publié: (2026)

MIRAGE: Evaluating and Explaining Inductive Reasoning Process in Language Models
par: Li, Jiachun, et autres
Publié: (2024)

Beyond Under-Alignment: Atomic Preference Enhanced Factuality Tuning for Large Language Models
par: Yuan, Hongbang, et autres
Publié: (2024)

Agent-RewardBench: Towards a Unified Benchmark for Reward Modeling across Perception, Planning, and Safety in Real-World Multimodal Agents
par: Men, Tianyi, et autres
Publié: (2025)

The Zero-Step Thinking: An Empirical Study of Mode Selection as Harder Early Exit in Reasoning Models
par: Tan, Yuqiao, et autres
Publié: (2025)

Do Self-Evolving Agents Forget? Capability Degradation and Preservation in Lifelong LLM Agent Adaptation
par: Yu, Ye, et autres
Publié: (2026)

MotivGraph-SoIQ: Integrating Motivational Knowledge Graphs and Socratic Dialogue for Enhanced LLM Ideation
par: Lei, Xinping, et autres
Publié: (2025)

From Generation to Judgment: Opportunities and Challenges of LLM-as-a-judge
par: Li, Dawei, et autres
Publié: (2024)

CITI: Enhancing Tool Utilizing Ability in Large Language Models without Sacrificing General Performance
par: Hao, Yupu, et autres
Publié: (2024)

Continual Few-shot Event Detection via Hierarchical Augmentation Networks
par: Zhang, Chenlong, et autres
Publié: (2024)

Evaluating Personalized Tool-Augmented LLMs from the Perspectives of Personalization and Proactivity
par: Hao, Yupu, et autres
Publié: (2025)

EvolKV: Evolutionary KV Cache Compression for LLM Inference
par: Yu, Bohan, et autres
Publié: (2025)