Imagen de Portada

BREAD: Branched Rollouts from Expert Anchors Bridge SFT & RL for Reasoning

Fuente: arXiv

Guardado en:

Detalles Bibliográficos
Autores principales:	Zhang, Xuechen, Huang, Zijian, Li, Yingcong, Ni, Chenshun, Chen, Jiasi, Oymak, Samet
Formato:	Preprint
Publicado:	2025
Materias:	Machine Learning
Acceso en línea:	Acceder al recurso
Etiquetas:	Agregar Etiqueta Sin Etiquetas, Sea el primero en etiquetar este registro!

Ejemplares similares

Making Small Language Models Efficient Reasoners: Intervention, Supervision, Reinforcement
por: Zhang, Xuechen, et al.
Publicado: (2025)

VSPO: Vector-Steered Policy Optimization for Behavioral Control
por: Zhang, Xuechen, et al.
Publicado: (2026)

Class-attribute Priors: Adapting Optimization to Heterogeneity and Fairness Objective
por: Zhang, Xuechen, et al.
Publicado: (2024)

Efficient Contextual LLM Cascades through Budget-Constrained Policy Learning
por: Zhang, Xuechen, et al.
Publicado: (2024)

SmartChunk Retrieval: Query-Aware Chunk Compression with Planning for Efficient Document RAG
por: Zhang, Xuechen, et al.
Publicado: (2025)

Selective Attention: Enhancing Transformer through Principled Context Control
por: Zhang, Xuechen, et al.
Publicado: (2024)

On the Power of Convolution Augmented Transformer
por: Li, Mingchen, et al.
Publicado: (2024)

Fine-grained Analysis of In-context Linear Estimation: Data, Architecture, and Beyond
por: Li, Yingcong, et al.
Publicado: (2024)

Transformers as Support Vector Machines
por: Tarzanagh, Davoud Ataee, et al.
Publicado: (2023)

Bridging SFT and RL: Dynamic Policy Optimization for Robust Reasoning
por: Zhu, Taojie, et al.
Publicado: (2026)

Continuous Chain of Thought Enables Parallel Exploration and Reasoning
por: Gozeten, Halil Alperen, et al.
Publicado: (2025)

From Self-Attention to Markov Models: Unveiling the Dynamics of Generative Transformers
por: Ildiz, M. Emrullah, et al.
Publicado: (2024)

Latent Chain-of-Thought Improves Structured-Data Transformers
por: Dudley, Carson, et al.
Publicado: (2026)

Mechanics of Next Token Prediction with Self-Attention
por: Li, Yingcong, et al.
Publicado: (2024)

Gating is Weighting: Understanding Gated Linear Attention through In-context Learning
por: Li, Yingcong, et al.
Publicado: (2025)

Test-Time Training Provably Improves Transformers as In-context Learners
por: Gozeten, Halil Alperen, et al.
Publicado: (2025)

Evolutionary Multi-Task Optimization for LLM-Guided Program Discovery
por: Gozeten, Halil Alperen, et al.
Publicado: (2026)

When and How Unlabeled Data Provably Improve In-Context Learning
por: Li, Yingcong, et al.
Publicado: (2025)

L3GS: Layered 3D Gaussian Splats for Efficient 3D Scene Delivery
por: Tsai, Yi-Zhen, et al.
Publicado: (2025)

Learning to Bet for Horizon-Aware Anytime-Valid Testing
por: Taga, Ege Onur, et al.
Publicado: (2026)

RL Fine-Tuning Heals OOD Forgetting in SFT
por: Jin, Hangzhan, et al.
Publicado: (2025)

Covariance-Aware Transformers for Quadratic Programming and Decision Making
por: Tire, Kutay, et al.
Publicado: (2026)

SFT-then-RL Outperforms Mixed-Policy Methods for LLM Reasoning
por: Limozin, Alexis, et al.
Publicado: (2026)

QuRL: Efficient Reinforcement Learning with Quantized Rollout
por: Li, Yuhang, et al.
Publicado: (2026)

Attention with Trained Embeddings Provably Selects Important Tokens
por: Wu, Diyuan, et al.
Publicado: (2025)

TimePFN: Effective Multivariate Time Series Forecasting with Synthetic Data
por: Taga, Ege Onur, et al.
Publicado: (2025)

In-Context Learning Under Regime Change
por: Dudley, Carson, et al.
Publicado: (2026)

Stable and Efficient Single-Rollout RL for Multimodal Reasoning
por: Liu, Rui, et al.
Publicado: (2025)

AceReason-Nemotron 1.1: Advancing Math and Code Reasoning through SFT and RL Synergy
por: Liu, Zihan, et al.
Publicado: (2025)

Heddle: A Distributed Orchestration System for Agentic RL Rollout
por: Zhang, Zili, et al.
Publicado: (2026)

Can Transformers Learn Optimal Filtering for Unknown Systems?
por: Balim, Haldun, et al.
Publicado: (2023)

EchoRL: Reinforcement Learning via Rollout Echoing
por: Bi, Jinhe, et al.
Publicado: (2026)

Metis-RISE: RL Incentivizes and SFT Enhances Multimodal Reasoning Model Learning
por: Qiu, Haibo, et al.
Publicado: (2025)

Retrieval Augmented Time Series Forecasting
por: Tire, Kutay, et al.
Publicado: (2024)

RL makes MLLMs see better than SFT
por: Song, Junha, et al.
Publicado: (2025)

Quagmires in SFT-RL Post-Training: When High SFT Scores Mislead and What to Use Instead
por: Kang, Feiyang, et al.
Publicado: (2025)

SPEC-RL: Accelerating On-Policy Reinforcement Learning with Speculative Rollouts
por: Liu, Bingshuai, et al.
Publicado: (2025)

Selective Rollout: Mid-Trajectory Termination for Multi-Sample Agent RL
por: Zhai, Zhiyuan, et al.
Publicado: (2026)

Patch the Distribution Mismatch: RL Rewriting Agent for Stable Off-Policy SFT
por: Wang, Jiacheng, et al.
Publicado: (2026)

GAC: Noise-Aware Adaptive Mixing for Hybrid SFT-RL Post-Training
por: Hu, Yuelin, et al.
Publicado: (2026)