Simon Sinong Zhan

Yixuan Wang, Simon Sinong Zhan, Ruochen Jiao et al.

It is quite challenging to ensure the safety of reinforcement learning (RL) agents in an unknown and stochastic environment under hard constraints that require the system state not to reach certain specified unsafe regions. Many popular safe RL methods such as those based on the Constrained Markov Decision Process (CMDP) paradigm formulate safety violations in a cost function and try to constrain the expectation of cumulative cost under a threshold. However, it is often difficult to effectively capture and enforce hard reachability-based safety constraints indirectly with such constraints on safety violation costs. In this work, we leverage the notion of barrier function to explicitly encode the hard safety constraints, and given that the environment is unknown, relax them to our design of \emph{generative-model-based soft barrier functions}. Based on such soft barriers, we propose a safe RL approach that can jointly learn the environment and optimize the control policy, while effectively avoiding unsafe regions with safety probability optimization. Experiments on a set of examples demonstrate that our approach can effectively enforce hard safety constraints and significantly outperform CMDP-based baseline methods in system safe rate measured via simulations.

Simon Sinong Zhan, Yixuan Wang, Qingyuan Wu et al.

In the context of safe exploration, Reinforcement Learning (RL) has long grappled with the challenges of balancing the tradeoff between maximizing rewards and minimizing safety violations, particularly in complex environments with contact-rich or non-smooth dynamics, and when dealing with high-dimensional pixel observations. Furthermore, incorporating state-wise safety constraints in the exploration and learning process, where the agent must avoid unsafe regions without prior knowledge, adds another layer of complexity. In this paper, we propose a novel pixel-observation safe RL algorithm that efficiently encodes state-wise safety constraints with unknown hazard regions through a newly introduced latent barrier-like function learning mechanism. As a joint learning framework, our approach begins by constructing a latent dynamics model with low-dimensional latent spaces derived from pixel observations. We then build and learn a latent barrier-like function on top of the latent dynamics and conduct policy optimization simultaneously, thereby improving both safety and the total expected return. Experimental evaluations on the safety-gym benchmark suite demonstrate that our proposed method significantly reduces safety violations throughout the training process, and demonstrates faster safety convergence compared to existing methods while achieving competitive results in reward return.

Qingyuan Wu, Simon Sinong Zhan, Yixuan Wang et al.

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 suffer from state space explosion or performance degeneration in stochastic environments. To address these challenges, we present a novel Auxiliary-Delayed Reinforcement Learning (AD-RL) method that leverages auxiliary tasks involving short delays to accelerate RL with long delays, without compromising performance in stochastic environments. Specifically, AD-RL learns a value function for short delays and uses bootstrapping and policy improvement techniques to adjust it for long delays. We theoretically show that this can greatly reduce the sample complexity. On deterministic and stochastic benchmarks, our method significantly outperforms the SOTAs in both sample efficiency and policy performance. Code is available at https://github.com/QingyuanWuNothing/AD-RL.

Qingyuan Wu, Yuhui Wang, Simon Sinong Zhan et al.

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 methods, such as GRPO and its variants, ensure stable updates by constraining policy divergence through clipping likelihood ratios. This paper introduces a unified clipping framework that characterizes existing methods via a general notion of policy divergence, encompassing both likelihood ratios and Kullback-Leibler (KL) divergences and extending to alternative measures. The framework provides a principled foundation for systematically analyzing how different policy divergence measures affect exploration and performance. We further identify the KL3 estimator, a variance-reduced Monte Carlo estimator of the KL divergence, as a key policy divergence constraint. We theoretically demonstrate that the KL3-based constraint is mathematically equivalent to an asymmetric ratio-based clipping that reallocates probability mass toward high-confidence actions, promoting stronger exploration while retaining the simplicity of GRPO-style methods. Empirical results on mathematical reasoning benchmarks demonstrate that incorporating the KL3 estimator into GRPO improves both training stability and final performance, highlighting the importance of principled policy divergence constraints in policy optimization.

Zhenyang Ni, Yijiang Li, Ruochen Jiao et al.

Video generation models trained on heterogeneous data with likelihood-surrogate objectives can produce visually plausible rollouts that violate physical constraints in embodied manipulation. Although reinforcement-learning post-training offers a natural route to adapting VGMs, existing video-RL rewards often reduce each rollout to a low-level visual metric, whereas manipulation video evaluation requires logic-based verification of whether the rollout satisfies a compositional task specification. To fill this gap, we introduce a compositional constraint-based reward model for post-training embodied video generation models, which automatically formulates task requirements as a composition of Linear Temporal Logic constraints, providing faithful rewards and localized error information in generated videos. To achieve effective improvement in high-dimensional video generation using these reward signals, we further propose CreFlow, a novel online RL framework with two key designs: i) a credit-aware NFT loss that confines the RL update to reward-relevant regions, preventing perturbations to unrelated regions during post-training; and ii) a corrective reflow loss that leverages within-group positive samples as an explicit estimate of the correction direction, stabilizing and accelerating training. Experiments show that CreFlow yields reward judgments better aligned with human and simulator success labels than existing methods and improves downstream execution success by 23.8 percentage points across eight bimanual manipulation tasks.

Qingyuan Wu, Yuhui Wang, Simon Sinong Zhan et al.

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 on past observations. State-of-the-art (SOTA) methods typically employ recursive, step-by-step forecasting of states. This can cause the accumulation of compounding errors. To tackle this problem, our novel belief estimation method, named Directly Forecasting Belief Transformer (DFBT), directly forecasts states from observations without incrementally estimating intermediate states step-by-step. We theoretically demonstrate that DFBT greatly reduces compounding errors of existing recursively forecasting methods, yielding stronger performance guarantees. In experiments with D4RL offline datasets, DFBT reduces compounding errors with remarkable prediction accuracy. DFBT's capability to forecast state sequences also facilitates multi-step bootstrapping, thus greatly improving learning efficiency. On the MuJoCo benchmark, our DFBT-based method substantially outperforms SOTA baselines. Code is available at https://github.com/QingyuanWuNothing/DFBT.

Jiacheng Lin, Kun Qian, Arvind Srinivasan et al.

Personalization today is fundamentally platform-centric: services build user representations from the behavioral fragments they observe. Yet no platform can construct a complete picture of the user, as competitive incentives, legal constraints, user privacy concerns, and epistemic limits create persistent data barriers. This paper argues for a shift from platform-centric personalization to user-governed personalization, where only the user can integrate fragmented contexts across platforms and the offline world. The key asymmetry lies in data access: only users can aggregate their own cross-platform and offline information. Large language model (LLM) agents make such integration practically feasible for the first time by enabling reasoning over heterogeneous personal data and transforming users' cross-context information into actionable personalization capabilities. We provide proof-of-concept evidence that users equipped with cross-platform data exports and an off-the-shelf LLM agent can outperform single-platform personalization baselines. We conclude by outlining a research agenda for building scalable user-governed personalization systems.

Qingyuan Wu, Simon Sinong Zhan, Yixuan Wang et al.

In environments with delayed observation, state augmentation by including actions within the delay window is adopted to retrieve Markovian property to enable reinforcement learning (RL). However, state-of-the-art (SOTA) RL techniques with Temporal-Difference (TD) learning frameworks often suffer from learning inefficiency, due to the significant expansion of the augmented state space with the delay. To improve learning efficiency without sacrificing performance, this work introduces a novel framework called Variational Delayed Policy Optimization (VDPO), which reformulates delayed RL as a variational inference problem. This problem is further modelled as a two-step iterative optimization problem, where the first step is TD learning in the delay-free environment with a small state space, and the second step is behaviour cloning which can be addressed much more efficiently than TD learning. We not only provide a theoretical analysis of VDPO in terms of sample complexity and performance, but also empirically demonstrate that VDPO can achieve consistent performance with SOTA methods, with a significant enhancement of sample efficiency (approximately 50\% less amount of samples) in the MuJoCo benchmark.

Simon Sinong Zhan, Yao Liu, Philip Wang et al.

We present Sentinel, the first framework for formally evaluating the physical safety of Large Language Model(LLM-based) embodied agents across the semantic, plan, and trajectory levels. Unlike prior methods that rely on heuristic rules or subjective LLM judgments, Sentinel grounds practical safety requirements in formal temporal logic (TL) semantics that can precisely specify state invariants, temporal dependencies, and timing constraints. It then employs a multi-level verification pipeline where (i) at the semantic level, intuitive natural language safety requirements are formalized into TL formulas and the LLM agent's understanding of these requirements is probed for alignment with the TL formulas; (ii) at the plan level, high-level action plans and subgoals generated by the LLM agent are verified against the TL formulas to detect unsafe plans before execution; and (iii) at the trajectory level, multiple execution trajectories are merged into a computation tree and efficiently verified against physically-detailed TL specifications for a final safety check. We apply Sentinel in VirtualHome and ALFRED, and formally evaluate multiple LLM-based embodied agents against diverse safety requirements. Our experiments show that by grounding physical safety in temporal logic and applying verification methods across multiple levels, Sentinel provides a rigorous foundation for systematically evaluating LLM-based embodied agents in physical environments, exposing safety violations overlooked by previous methods and offering insights into their failure modes.

Simon Sinong Zhan, Qingyuan Wu, Frank Yang et al.

Offline-to-online deployment of reinforcement-learning (RL) agents must bridge two gaps: (1) the sim-to-real gap, where real systems add latency and other imperfections not present in simulation, and (2) the interaction gap, where policies trained purely offline face out-of-distribution states during online execution because gathering new interaction data is costly or risky. Agents therefore have to generalize from static, delay-free datasets to dynamic, delay-prone environments. Standard offline RL learns from delay-free logs yet must act under delays that break the Markov assumption and hurt performance. We introduce DT-CORL (Delay-Transformer belief policy Constrained Offline RL), an offline-RL framework built to cope with delayed dynamics at deployment. DT-CORL (i) produces delay-robust actions with a transformer-based belief predictor even though it never sees delayed observations during training, and (ii) is markedly more sample-efficient than naïve history-augmentation baselines. Experiments on D4RL benchmarks with several delay settings show that DT-CORL consistently outperforms both history-augmentation and vanilla belief-based methods, narrowing the sim-to-real latency gap while preserving data efficiency.

Simon Sinong Zhan, Qingyuan Wu, Zhian Ruan et al.

Inverse Reinforcement Learning (IRL) has demonstrated effectiveness in a variety of imitation tasks. In this paper, we introduce an IRL framework designed to extract rewarding features from expert trajectories affected by delayed disturbances. Instead of relying on direct observations, our approach employs an efficient off-policy adversarial training framework to derive expert features and recover optimal policies from augmented delayed observations. Empirical evaluations in the MuJoCo environment under diverse delay settings validate the effectiveness of our method. Furthermore, we provide a theoretical analysis showing that recovering expert policies from augmented delayed observations outperforms using direct delayed observations.