Image de couverture de livre

A Task-Centric Theory for Iterative Self-Improvement with Easy-to-Hard Curricula

Fuente: arXiv

Enregistré dans:

Détails bibliographiques
Auteurs principaux:	Liu, Chenruo, Dong, Yijun, Shen, Yiqiu, Lei, Qi
Format:	Preprint
Publié:	2026
Sujets:	Machine Learning
Accès en ligne:	Acceder al recurso
Tags:	Ajouter un tag Pas de tags, Soyez le premier à ajouter un tag!

Documents similaires

Does Weak-to-strong Generalization Happen under Spurious Correlations?
par: Liu, Chenruo, et autres
Publié: (2025)

Superclass-Guided Representation Disentanglement for Spurious Correlation Mitigation
par: Liu, Chenruo, et autres
Publié: (2025)

Bridging Distribution Shift and AI Safety: Conceptual and Methodological Synergies
par: Liu, Chenruo, et autres
Publié: (2025)

Teaching by Failure: Counter-Example-Driven Curricula for Transformer Self-Improvement
par: Vejendla, Harshil
Publié: (2025)

The Unreasonable Effectiveness of Easy Training Data for Hard Tasks
par: Hase, Peter, et autres
Publié: (2024)

Language Models Do Hard Arithmetic Tasks Easily and Hardly Do Easy Arithmetic Tasks
par: Gambardella, Andrew, et autres
Publié: (2024)

When does Chain-of-Thought Help: A Markovian Perspective
par: Wang, Zihan, et autres
Publié: (2026)

Self-Improving Transformers Overcome Easy-to-Hard and Length Generalization Challenges
par: Lee, Nayoung, et autres
Publié: (2025)

Easy-to-Hard Generalization: Scalable Alignment Beyond Human Supervision
par: Sun, Zhiqing, et autres
Publié: (2024)

WIST: Web-Grounded Iterative Self-Play Tree for Domain-Targeted Reasoning Improvement
par: Li, Fangyuan, et autres
Publié: (2026)

Curriculum Reinforcement Learning from Easy to Hard Tasks Improves LLM Reasoning
par: Parashar, Shubham, et autres
Publié: (2025)

Bootstrapping Task Spaces for Self-Improvement
par: Jiang, Minqi, et autres
Publié: (2025)

Learning Compositional Functions with Transformers from Easy-to-Hard Data
par: Wang, Zixuan, et autres
Publié: (2025)

Multi-Agent Imitation Learning: Value is Easy, Regret is Hard
par: Tang, Jingwu, et autres
Publié: (2024)

Compress the Easy, Explore the Hard: Difficulty-Aware Entropy Regularization for Efficient LLM Reasoning
par: Luo, Qin-Wen, et autres
Publié: (2026)

Dissecting Sample Hardness: A Fine-Grained Analysis of Hardness Characterization Methods for Data-Centric AI
par: Seedat, Nabeel, et autres
Publié: (2024)

MetaLint: Easy-to-Hard Generalization for Code Linting
par: Naik, Atharva, et autres
Publié: (2025)

Sparsity via Hyperpriors: A Theoretical and Algorithmic Study under Empirical Bayes Framework
par: Li, Zhitao, et autres
Publié: (2025)

Sketchy Moment Matching: Toward Fast and Provable Data Selection for Finetuning
par: Dong, Yijun, et autres
Publié: (2024)

More Consistent Accuracy PINN via Alternating Easy-Hard Training
par: Gao, Zhaoqian, et autres
Publié: (2025)

The Easy Path to Robustness: Coreset Selection using Sample Hardness
par: Ramesh, Pranav, et autres
Publié: (2025)

Frustratingly Easy Task-aware Pruning for Large Language Models
par: Tian, Yuanhe, et autres
Publié: (2025)

SC2Arena and StarEvolve: Benchmark and Self-Improvement Framework for LLMs in Complex Decision-Making Tasks
par: Shen, Pengbo, et autres
Publié: (2025)

Sample-Centric Multi-Task Learning for Detection and Segmentation of Industrial Surface Defects
par: Dong, Hang-Cheng, et autres
Publié: (2025)

Probabilistic Iterative Hard Thresholding for Sparse Learning
par: Bergamaschi, Matteo, et autres
Publié: (2024)

Normalized Iterative Hard Thresholding for Tensor Recovery
par: Li, Li, et autres
Publié: (2025)

Towards Initialization-Agnostic Clustering with Iterative Adaptive Resonance Theory
par: Qu, Xiaozheng, et autres
Publié: (2025)

EasyTime: Time Series Forecasting Made Easy
par: Qiu, Xiangfei, et autres
Publié: (2024)

EH-MAM: Easy-to-Hard Masked Acoustic Modeling for Self-Supervised Speech Representation Learning
par: Seth, Ashish, et autres
Publié: (2024)

Easy Adaptation: An Efficient Task-Specific Knowledge Injection Method for Large Models in Resource-Constrained Environments
par: Chen, Dong, et autres
Publié: (2025)

Enhancing Sample Selection Against Label Noise by Cutting Mislabeled Easy Examples
par: Yuan, Suqin, et autres
Publié: (2025)

Thermodynamics of Reinforcement Learning Curricula
par: Adamczyk, Jacob, et autres
Publié: (2026)

Easy to Learn, Yet Hard to Forget: Towards Robust Unlearning Under Bias
par: Kwon, JuneHyoung, et autres
Publié: (2026)

Bottom-Up Synthesis of Knowledge-Grounded Task-Oriented Dialogues with Iteratively Self-Refined Prompts
par: Qian, Kun, et autres
Publié: (2025)

Detecting is Easy, Adapting is Hard: Local Expert Growth for Visual Model-Based Reinforcement Learning under Distribution Shift
par: Zhao, Haiyang
Publié: (2026)

Reward-Free Curricula for Training Robust World Models
par: Rigter, Marc, et autres
Publié: (2023)

Discrepancies are Virtue: Weak-to-Strong Generalization through Lens of Intrinsic Dimension
par: Dong, Yijun, et autres
Publié: (2025)

Easy2Hard-Bench: Standardized Difficulty Labels for Profiling LLM Performance and Generalization
par: Ding, Mucong, et autres
Publié: (2024)

Adaptive Auto-Harness: Sustained Self-Improvement for Agentic System Deployment on Open-Ended Task Streams
par: Liu, Zewen, et autres
Publié: (2026)

Self-Improvement for Neural Combinatorial Optimization: Sample without Replacement, but Improvement
par: Pirnay, Jonathan, et autres
Publié: (2024)