Imagen de Portada

Beyond Distribution Sharpening: The Importance of Task Rewards

Fuente: arXiv

Guardado en:

Detalles Bibliográficos
Autores principales:	Mittal, Sarthak, Gagnon, Leo, Lajoie, Guillaume
Formato:	Preprint
Publicado:	2026
Materias:	Machine Learning Artificial Intelligence
Acceso en línea:	Acceder al recurso
Etiquetas:	Agregar Etiqueta Sin Etiquetas, Sea el primero en etiquetar este registro!

Ejemplares similares

In-Context Parametric Inference: Point or Distribution Estimators?
por: Mittal, Sarthak, et al.
Publicado: (2025)

Does learning the right latent variables necessarily improve in-context learning?
por: Mittal, Sarthak, et al.
Publicado: (2024)

Next-Token Prediction Should be Ambiguity-Sensitive: A Meta-Learning Perspective
por: Gagnon, Leo, et al.
Publicado: (2025)

In-context learning and Occam's razor
por: Elmoznino, Eric, et al.
Publicado: (2024)

Iterative Amortized Inference: Unifying In-Context Learning and Learned Optimizers
por: Mittal, Sarthak, et al.
Publicado: (2025)

A Compression Perspective on Simplicity Bias
por: Marty, Tom, et al.
Publicado: (2026)

Amortized In-Context Bayesian Posterior Estimation
por: Mittal, Sarthak, et al.
Publicado: (2025)

Scalable Power Sampling: Unlocking Efficient, Training-Free Reasoning for LLMs via Distribution Sharpening
por: Ji, Xiaotong, et al.
Publicado: (2026)

Preference as Reward, Maximum Preference Optimization with Importance Sampling
por: Jiang, Zaifan, et al.
Publicado: (2023)

Omni-Thinker: Scaling Multi-Task RL in LLMs with Hybrid Reward and Task Scheduling
por: Li, Derek, et al.
Publicado: (2025)

A Complexity-Based Theory of Compositionality
por: Elmoznino, Eric, et al.
Publicado: (2024)

Accelerating Training with Neuron Interaction and Nowcasting Networks
por: Knyazev, Boris, et al.
Publicado: (2024)

Rewarding the Unlikely: Lifting GRPO Beyond Distribution Sharpening
por: He, Andre, et al.
Publicado: (2025)

Mitigating Distribution Sharpening in Math RLVR via Distribution-Aligned Hint Synthesis and Backward Hint Annealing
por: Xie, Pei-Xi, et al.
Publicado: (2026)

BoxRL-NNV: Boxed Refinement of Latin Hypercube Samples for Neural Network Verification
por: Das, Sarthak
Publicado: (2025)

Sharpen Your Flow: Sharpness-Aware Sampling for Flow Matching
por: Gupta, Aditi, et al.
Publicado: (2026)

Beyond Reward Hacking: Causal Rewards for Large Language Model Alignment
por: Wang, Chaoqi, et al.
Publicado: (2025)

Discrete, compositional, and symbolic representations through attractor dynamics
por: Nam, Andrew, et al.
Publicado: (2023)

Beyond Importance Sampling: Rejection-Gated Policy Optimization
por: Sun, Ziwu, et al.
Publicado: (2026)

All You Need Is Synthetic Task Augmentation
por: Godin, Guillaume
Publicado: (2025)

Beyond Rewards in Reinforcement Learning for Cyber Defence
por: Bates, Elizabeth, et al.
Publicado: (2026)

On the Interplay Between Stepsize Tuning and Progressive Sharpening
por: Roulet, Vincent, et al.
Publicado: (2023)

Self-Improvement in Language Models: The Sharpening Mechanism
por: Huang, Audrey, et al.
Publicado: (2024)

Beyond Scalar Rewards: Distributional Reinforcement Learning with Preordered Objectives for Safe and Reliable Autonomous Driving
por: Abouelazm, Ahmed, et al.
Publicado: (2026)

JEDI: Jointly Embedded Inference of Neural Dynamics
por: Jamkhandi, Anirudh, et al.
Publicado: (2026)

Beyond Scalar Rewards: An Axiomatic Framework for Lexicographic MDPs
por: Shakerinava, Mehran, et al.
Publicado: (2025)

Multi-Task Reward Learning from Human Ratings
por: Wu, Mingkang, et al.
Publicado: (2025)

On the Importance of Task Complexity in Evaluating LLM-Based Multi-Agent Systems
por: Tang, Bohan, et al.
Publicado: (2025)

On the Sample Efficiency of Abstractions and Potential-Based Reward Shaping in Reinforcement Learning
por: Canonaco, Giuseppe, et al.
Publicado: (2024)

Beyond Semantic Manipulation: Token-Space Attacks on Reward Models
por: Zhang, Yuheng, et al.
Publicado: (2026)

Distributional Process Reward Models: Calibrated Prediction of Future Rewards via Conditional Optimal Transport
por: Ma, Rachel, et al.
Publicado: (2026)

Learning Robust Reward Machines from Noisy Labels
por: Parac, Roko, et al.
Publicado: (2024)

Sharpening the Spear: Adaptive Expert-Guided Adversarial Attack Against DRL-based Autonomous Driving Policies
por: Fan, Junchao, et al.
Publicado: (2025)

Mitigating Long-Tailed Anomaly Score Distributions with Importance-Weighted Loss
por: Lee, Jungi, et al.
Publicado: (2026)

Beyond Correctness: Harmonizing Process and Outcome Rewards through RL Training
por: Ye, Chenlu, et al.
Publicado: (2025)

Goldilocks RL: Tuning Task Difficulty to Escape Sparse Rewards for Reasoning
por: Mahrooghi, Ilia, et al.
Publicado: (2026)

Hybrid Reward Normalization for Process-supervised Non-verifiable Agentic Tasks
por: Xu, Peiran, et al.
Publicado: (2025)

MAESTRO: Multi-Agent Environment Shaping through Task and Reward Optimization
por: Wu, Boyuan
Publicado: (2025)

Scalable Neural Network Training over Distributed Graphs
por: Kolluri, Aashish, et al.
Publicado: (2023)

Beyond Fixed Tasks: Unsupervised Environment Design for Task-Level Pairs
por: Furelos-Blanco, Daniel, et al.
Publicado: (2025)