Copertina

When Correct Demonstrations Hurt: Rethinking the Role of Exemplars in In-Context Learning

Fuente: arXiv

Salvato in:

Dettagli Bibliografici
Autori principali:	Qiu, Chenghao, Peng, Chunli, Yang, Yufeng, Huang, Kuan-Hao, Zhou, Yi
Natura:	Preprint
Pubblicazione:	2026
Soggetti:	Machine Learning Artificial Intelligence
Accesso online:	Acceder al recurso
Tags:	Aggiungi Tag Nessun Tag, puoi essere il primo ad aggiungerne!!

Documenti analoghi

The Pitfalls of Memorization: When Memorization Hurts Generalization
di: Bayat, Reza, et al.
Pubblicazione: (2024)

KITE: Kernelized and Information Theoretic Exemplars for In-Context Learning
di: Singh, Vaibhav, et al.
Pubblicazione: (2025)

Difficult Examples Hurt Unsupervised Contrastive Learning: A Theoretical Perspective
di: Zhang, Yi-Ge, et al.
Pubblicazione: (2025)

Learning Quadruped Walking from Seconds of Demonstration
di: Zhang, Ruipeng, et al.
Pubblicazione: (2026)

Learning to Answer from Correct Demonstrations
di: Joshi, Nirmit, et al.
Pubblicazione: (2025)

Understanding the Dynamics of Demonstration Conflict in In-Context Learning
di: Jiao, Difan, et al.
Pubblicazione: (2026)

Exemplar-condensed Federated Class-incremental Learning
di: Sun, Rui, et al.
Pubblicazione: (2024)

Unraveling the Mechanics of Learning-Based Demonstration Selection for In-Context Learning
di: Liu, Hui, et al.
Pubblicazione: (2024)

Rethinking Refinement: Correcting Generative Bias without Noise Injection
di: Peng, Xin, et al.
Pubblicazione: (2026)

When Drafts Evolve: Speculative Decoding Meets Online Learning
di: Qian, Yu-Yang, et al.
Pubblicazione: (2026)

Privacy-Preserving Personalized Federated Learning for Distributed Photovoltaic Disaggregation under Statistical Heterogeneity
di: Chen, Xiaolu, et al.
Pubblicazione: (2025)

FedCoSR: Personalized Federated Learning with Contrastive Shareable Representations for Label Heterogeneity in Non-IID Data
di: Huang, Chenghao, et al.
Pubblicazione: (2024)

Learning to Select In-Context Demonstration Preferred by Large Language Model
di: Zhang, Zheng, et al.
Pubblicazione: (2025)

Analytic Drift Resister for Non-Exemplar Continual Graph Learning
di: Song, Lei, et al.
Pubblicazione: (2026)

When Shared Knowledge Hurts: Spectral Over-Accumulation in Model Merging
di: Li, Yayuan, et al.
Pubblicazione: (2026)

Rethinking the Unsolvable: When In-Context Search Meets Test-Time Scaling
di: Xia, Fanzeng, et al.
Pubblicazione: (2025)

Rethink the Role of Neural Decoders in Quantum Error Correction
di: Yan, Ge, et al.
Pubblicazione: (2026)

When Generative Augmentation Hurts: A Benchmark Study of GAN and Diffusion Models for Bias Correction in AI Classification Systems
di: Gupta, Shesh Narayan, et al.
Pubblicazione: (2026)

LMAct: A Benchmark for In-Context Imitation Learning with Long Multimodal Demonstrations
di: Ruoss, Anian, et al.
Pubblicazione: (2024)

When Reasoning Hurts: Source-Aware Evaluation of Frontier LLMs for Clinical SOAP Note Generation
di: Faisal, Faizan
Pubblicazione: (2026)

DemoShapley: Valuation of Demonstrations for In-Context Learning
di: Xie, Shan, et al.
Pubblicazione: (2024)

When a Robot is More Capable than a Human: Learning from Constrained Demonstrators
di: Li, Xinhu, et al.
Pubblicazione: (2025)

Mimic In-Context Learning for Multimodal Tasks
di: Jiang, Yuchu, et al.
Pubblicazione: (2025)

Process In-Context Learning: Enhancing Mathematical Reasoning via Dynamic Demonstration Insertion
di: Gao, Ang, et al.
Pubblicazione: (2026)

Data Shifts Hurt CoT: A Theoretical Study
di: Yin, Lang, et al.
Pubblicazione: (2025)

Prototype-Guided Non-Exemplar Continual Learning for Cross-subject EEG Decoding
di: Li, Dan, et al.
Pubblicazione: (2025)

Expandable and Differentiable Dual Memories with Orthogonal Regularization for Exemplar-free Continual Learning
di: Moon, Hyung-Jun, et al.
Pubblicazione: (2025)

Not All Queries Need Rewriting: When Prompt-Only LLM Refinement Helps and Hurts Dense Retrieval
di: Kotte, Varun
Pubblicazione: (2026)

Batch Calibration: Rethinking Calibration for In-Context Learning and Prompt Engineering
di: Zhou, Han, et al.
Pubblicazione: (2023)

Scaling Context Requires Rethinking Attention
di: Gelada, Carles, et al.
Pubblicazione: (2025)

Unmasking On-Policy Distillation: Where It Helps, Where It Hurts, and Why
di: Armandpour, Mohammadreza, et al.
Pubblicazione: (2026)

Smart Energy Guardian: A Hybrid Deep Learning Model for Detecting Fraudulent PV Generation
di: Chen, Xiaolu, et al.
Pubblicazione: (2025)

Rethinking LoRA for Data Heterogeneous Federated Learning: Subspace and State Alignment
di: Peng, Hongyi, et al.
Pubblicazione: (2026)

Rethinking Hallucinations: Correctness, Consistency, and Prompt Multiplicity
di: Ganesh, Prakhar, et al.
Pubblicazione: (2026)

When Demonstrations Meet Generative World Models: A Maximum Likelihood Framework for Offline Inverse Reinforcement Learning
di: Zeng, Siliang, et al.
Pubblicazione: (2023)

Comparative Analysis of Demonstration Selection Algorithms for LLM In-Context Learning
di: Shu, Dong, et al.
Pubblicazione: (2024)

VL Norm: Rethink Loss Aggregation in RLVR
di: He, Zhiyuan, et al.
Pubblicazione: (2025)

Feature-Based vs. GAN-Based Learning from Demonstrations: When and Why
di: Li, Chenhao, et al.
Pubblicazione: (2025)

The Role of Diversity in In-Context Learning for Large Language Models
di: Xiao, Wenyang, et al.
Pubblicazione: (2025)

TaskVAE: Task-Specific Variational Autoencoders for Exemplar Generation in Continual Learning for Human Activity Recognition
di: Kann, Bonpagna, et al.
Pubblicazione: (2025)