12 papers
Causal Improvement Graph for Agentic Harness Optimization
Junjie Zhang, Shunyu Liu, Haoyu Wang +3
Agentic Harness is the runtime that constructs task context and controls execution flow, thereby shaping overall agent performance. Given a fixed model and external evaluation, aut…
Beyond Isolation: Unlocking Reinforcement Learning Component Synergy for Sample-Efficient Continuous Control
Qi Zhao, Guozheng Ma, Yilun Kong +9
Reinforcement learning systems are significantly more complex than other machine learning paradigms due to inherent properties, causing RL system design to jointly account for many…
Beyond Zooming: Learning Multi-Tool Visual Reasoning for Ultra-High-Resolution Remote Sensing
Fengxiang Wang, Jiangnan Huang, Mingshuo Chen +8
Ultra-high-resolution (UHR) remote-sensing (RS) imagery provides fine-grained Earth-observation evidence over city-scale scenes, but poses a fundamental challenge for multimodal la…
ExToken: Structured Exploration for Efficient Vision-Language-Action Reinforcement Fine-tuning
Yilun Kong, Yunpeng Qing, Guozheng Ma +4
Reinforcement Learning (RL) has demonstrated significant potential for improving Vision-Language-Action (VLA) models on complex manipulation tasks. However, its practical scalabili…
Fisher Decorator: Refining Flow Policy via a Local Transport Map
Xiaoyuan Cheng, Haoyu Wang, Wenxuan Yuan +4
Recent advances in flow-based offline reinforcement learning (RL) have achieved strong performance by parameterizing policies via flow matching. However, they still face critical t…
GEM: Guided Expectation-Maximization for Behavior-Normalized Candidate Action Selection in Offline RL
Haoyu Wang, Jingcheng Wang, Shunyu Wu +1
Offline reinforcement learning (RL) can fit strong value functions from fixed datasets, yet reliable deployment still hinges on the action selection interface used to query them. W…
Text Before Vision: Staged Knowledge Injection Matters for Agentic RLVR in Ultra-High-Resolution Remote Sensing Understanding
Fengxiang Wang, Mingshuo Chen, Yueying Li +13
Multimodal reasoning for ultra-high-resolution (UHR) remote sensing (RS) is usually bottlenecked by visual evidence acquisition: the model necessitates localizing tiny task-relevan…
What Makes Value Learning Efficient in Residual Reinforcement Learning?
Guozheng Ma, Lu Li, Haoyu Wang +3
Residual reinforcement learning (RL) enables stable online refinement of expressive pretrained policies by freezing the base and learning only bounded corrections. However, value l…
Language-based Trial and Error Falls Behind in the Era of Experience
Haoyu Wang, Guozheng Ma, Shugang Cui +7
While Large Language Models (LLMs) excel in language-based agentic tasks, their applicability to unseen, nonlinguistic environments (e.g., symbolic or spatial tasks) remains limite…
Rethinking the Role of Dynamic Sparse Training for Scalable Deep Reinforcement Learning
Guozheng Ma, Lu Li, Zilin Wang +4
Scaling neural networks has driven breakthrough advances in machine learning, yet this paradigm fails in deep reinforcement learning (DRL), where larger models often degrade perfor…
A Simple "Motivation" Can Enhance Reinforcement Finetuning of Large Reasoning Models
Junjie Zhang, Guozheng Ma, Shunyu Liu +6
Reinforcement Learning with Verifiable Rewards~(RLVR) has emerged as a powerful learn-to-reason paradigm for large reasoning models to tackle complex tasks. However, the current RL…
Pangu Embedded: An Efficient Dual-system LLM Reasoner with Metacognition
Hanting Chen, Yasheng Wang, Kai Han +21
This work presents Pangu Embedded, an efficient Large Language Model (LLM) reasoner developed on Ascend Neural Processing Units (NPUs), featuring flexible fast and slow thinking ca…