paper-with-me

홈 › Papers

Towards Flash Thinking via Decoupled Advantage Policy Optimization

2025-10-17 · Zezhong Tan, Hang Gao, Xinhong Ma, Feng Zhang, Ziqiang Dong arxiv

Recent Large Reasoning Models (LRMs) have achieved remarkable performance in solving complex problems via supervised fine-tuning (SFT) and reinforcement learning (RL). Although existing RL algorithms significantly enhance model accuracy, they still suffer from excessively lengthy responses and overthinking issues, resulting in increased inference latency and computational consumption, especially for simple tasks that require minimal reasoning. To address this, we propose a novel RL framework, DEPO, to reduce inefficient reasoning for models. Our method mainly consists of three core components: (1) an innovative advantage decoupled algorithm to guide model reduction of inefficient tokens; (2) a difficulty-aware length penalty to lower the overall length of model responses; (3) an advantage clipping method to prevent bias in policy optimization. In our experiments, applied to DeepSeek-Distill-Qwen-7B and DeepSeek-Distill-Qwen-1.5B as base models, DEPO achieves a significant reduction in sequence length by 39% and reduces excessive reasoning paths in inefficient tokens, while outperforming the base model in overall accuracy.

📄 PDF Abstract BibTeX arXiv:2510.15374

Code (0)

등록된 구현이 없습니다.

Tasks

Reinforcement Learning

Similar Papers 제목 키워드 기반

Yuan3.0 Flash: An Open Multimodal Large Language Model for Enterprise Applications

2026-01-05 · YuanLab. ai, :, Shawn Wu, Sean Wang 외 arxiv

We introduce Yuan3.0 Flash, an open-source Mixture-of-Experts (MoE) MultiModal Large Language Model featuring 3.7B activated parameters and 40B total parameters, specifically designed to enhance performance on enterprise…

JoyAI-LLM Flash: Advancing Mid-Scale LLMs with Token Efficiency

2026-04-03 · Aichen Cai, Anmeng Zhang, Anyu Li, Bo Zhang 외 arxiv

We introduce JoyAI-LLM Flash, an efficient Mixture-of-Experts (MoE) language model designed to redefine the trade-off between strong performance and token efficiency in the sub-50B parameter regime. JoyAI-LLM Flash is pr…

Reinforcement Learning

IterSynth: Rethinking Deep Search Agents via Role-Decoupled Iterative Synthesis

2026-09-24 · Xingyu Wu, Yuchen Yan, Zhengxi Lu, Siqi Chen 외 hf

Deep search requires LLM agents to decompose complex queries, search for evidence, and synthesize grounded answers, yet existing ReAct-style agents suffer from two limitations: role coupling, where one policy must handle…

Reinforcement Learning

BAPO: Stabilizing Off-Policy Reinforcement Learning for LLMs via Balanced Policy Optimization with Adaptive Clipping

2025-10-21 · Zhiheng Xi, Xin Guo, Yang Nan, Enyu Zhou 외 arxiv

Reinforcement learning (RL) has recently become the core paradigm for aligning and strengthening large language models (LLMs). Yet, applying RL in off-policy settings--where stale data from past policies are used for tra…

Reinforcement Learning

DRPO: Efficient Reasoning via Decoupled Reward Policy Optimization

2025-10-06 · Gang Li, Yan Chen, Ming Lin, Tianbao Yang arxiv

Recent large reasoning models (LRMs) driven by reinforcement learning algorithms (e.g., GRPO) have achieved remarkable performance on challenging reasoning tasks. However, these models suffer from overthinking, generatin…

Reinforcement LearningMathematical Reasoning