10 papers
Hybrid Advantage Estimation with Unified Critic for VLM Agentic Reinforcement Learning
Wenxuan Zhang, Yuhui Wang, Donggang Jia +5
Large Vision-Language Models (VLMs) now act as agents in interactive environments, where success requires coherent reasoning and decision-making across turns. Although end-to-end t…
Efficient Morphology-Control Co-Design via Stackelberg Proximal Policy Optimization
Yanning Dai, Yuhui Wang, Dylan R. Ashley +1
Morphology-control co-design concerns the coupled optimization of an agent's body structure and control policy. This problem exhibits a bi-level structure, where the control dynami…
A Unified Framework for Rethinking Policy Divergence Measures in GRPO
Qingyuan Wu, Yuhui Wang, Simon Sinong Zhan +6
Reinforcement Learning with Verified Reward (RLVR) has emerged as a critical paradigm for advancing the reasoning capabilities of Large Language Models (LLMs). Most existing RLVR m…
RoadSceneBench: A Lightweight Benchmark for Mid-Level Road Scene Understanding
Xiyan Liu, Han Wang, Yuhu Wang +4
Understanding mid-level road semantics, which capture the structural and contextual cues that link low-level perception to high-level planning, is essential for reliable autonomous…
Directly Forecasting Belief for Reinforcement Learning with Delays
Qingyuan Wu, Yuhui Wang, Simon Sinong Zhan +6
Reinforcement learning (RL) with delays is challenging as sensory perceptions lag behind the actual events: the RL agent needs to estimate the real state of its environment based o…
Highway Value Iteration Networks
Yuhui Wang, Weida Li, Francesco Faccio +2
Value iteration networks (VINs) enable end-to-end learning for planning tasks by employing a differentiable "planning module" that approximates the value iteration algorithm. Howev…
Scaling Value Iteration Networks to 5000 Layers for Extreme Long-Term Planning
Yuhui Wang, Qingyuan Wu, Dylan R. Ashley +4
The Value Iteration Network (VIN) is an end-to-end differentiable neural network architecture for planning. It exhibits strong generalization to unseen domains by incorporating a d…
Highway Reinforcement Learning
Yuhui Wang, Miroslav Strupl, Francesco Faccio +5
Learning from multi-step off-policy data collected by a set of policies is a core problem of reinforcement learning (RL). Approaches based on importance sampling (IS) often suffer…
Variational Delayed Policy Optimization
Qingyuan Wu, Simon Sinong Zhan, Yixuan Wang +5
In environments with delayed observation, state augmentation by including actions within the delay window is adopted to retrieve Markovian property to enable reinforcement learning…
Boosting Reinforcement Learning with Strongly Delayed Feedback Through Auxiliary Short Delays
Qingyuan Wu, Simon Sinong Zhan, Yixuan Wang +6
Reinforcement learning (RL) is challenging in the common case of delays between events and their sensory perceptions. State-of-the-art (SOTA) state augmentation techniques either s…