Faguo Wu

Guojian Wang, Jianxiang Liu, Xinyuan Li et al.

In this paper, we investigate preference-based reinforcement learning (PbRL), which enables reinforcement learning (RL) agents to learn from human feedback. This is particularly valuable when defining a fine-grain reward function is not feasible. However, this approach is inefficient and impractical for promoting deep exploration in hard-exploration tasks with long horizons and sparse rewards. To tackle this issue, we introduce LOPE: \textbf{L}earning \textbf{O}nline with trajectory \textbf{P}reference guidanc\textbf{E}, an end-to-end preference-guided RL framework that enhances exploration efficiency in hard-exploration tasks. Our intuition is that LOPE directly adjusts the focus of online exploration by considering human feedback as guidance, thereby avoiding the need to learn a separate reward model from preferences. Specifically, LOPE includes a two-step sequential policy optimization technique consisting of trust-region-based policy improvement and preference guidance steps. We reformulate preference guidance as a trajectory-wise state marginal matching problem that minimizes the maximum mean discrepancy distance between the preferred trajectories and the learned policy. Furthermore, we provide a theoretical analysis to characterize the performance improvement bound and evaluate the effectiveness of the LOPE. When assessed in various challenging hard-exploration environments, LOPE outperforms several state-of-the-art methods in terms of convergence rate and overall performance.The code used in this study is available at https://github.com/buaawgj/LOPE.

Feng Liu, Faguo Wu, Xiao Zhang

The normalization constraint on probability density poses a significant challenge for solving the Fokker-Planck equation. Normalizing Flow, an invertible generative model leverages the change of variables formula to ensure probability density conservation and enable the learning of complex data distributions. In this paper, we introduce Physics-Informed Normalizing Flows (PINF), a novel extension of continuous normalizing flows, incorporating diffusion through the method of characteristics. Our method, which is mesh-free and causality-free, can efficiently solve high dimensional time-dependent and steady-state Fokker-Planck equations.

Guojian Wang, Faguo Wu, Xiao Zhang et al.

Deep reinforcement learning (DRL) faces significant challenges in addressing the hard-exploration problems in tasks with sparse or deceptive rewards and large state spaces. These challenges severely limit the practical application of DRL. Most previous exploration methods relied on complex architectures to estimate state novelty or introduced sensitive hyperparameters, resulting in instability. To mitigate these issues, we propose an efficient adaptive trajectory-constrained exploration strategy for DRL. The proposed method guides the policy of the agent away from suboptimal solutions by leveraging incomplete offline demonstrations as references. This approach gradually expands the exploration scope of the agent and strives for optimality in a constrained optimization manner. Additionally, we introduce a novel policy-gradient-based optimization algorithm that utilizes adaptively clipped trajectory-distance rewards for both single- and multi-agent reinforcement learning. We provide a theoretical analysis of our method, including a deduction of the worst-case approximation error bounds, highlighting the validity of our approach for enhancing exploration. To evaluate the effectiveness of the proposed method, we conducted experiments on two large 2D grid world mazes and several MuJoCo tasks. The extensive experimental results demonstrate the significant advantages of our method in achieving temporally extended exploration and avoiding myopic and suboptimal behaviors in both single- and multi-agent settings. Notably, the specific metrics and quantifiable results further support these findings. The code used in the study is available at \url{https://github.com/buaawgj/TACE}.

Nanxiang Jiang, Zhaoxin Fan, Baisen Wang et al.

Concept erasure serves as a vital safety mechanism for removing unwanted concepts from text-to-image (T2I) models. While extensively studied in U-Net and dual-stream architectures (e.g., Flux), this task remains under-explored in the recent emerging paradigm of single-stream diffusion transformers (e.g., Z-Image). In this new paradigm, text and image tokens are processed as a single unified sequence via shared parameters. Consequently, directly applying prior erasure methods typically leads to generation collapse. To bridge this gap, we introduce Z-Erase, the first concept erasure method tailored for single-stream T2I models. To guarantee stable image generation, Z-Erase first proposes a Stream Disentangled Concept Erasure Framework that decouples updates and enables existing methods on single-stream models. Subsequently, within this framework, we introduce Lagrangian-Guided Adaptive Erasure Modulation, a constrained algorithm that further balances the sensitive erasure-preservation trade-off. Moreover, we provide a rigorous convergence analysis proving that Z-Erase can converge to a Pareto stationary point. Experiments demonstrate that Z-Erase successfully overcomes the generation collapse issue, achieving state-of-the-art performance across a wide range of tasks.

Guojian Wang, Faguo Wu, Xiao Zhang et al.

Reinforcement learning (RL) with sparse and deceptive rewards is challenging because non-zero rewards are rarely obtained. Hence, the gradient calculated by the agent can be stochastic and without valid information. Recent studies that utilize memory buffers of previous experiences can lead to a more efficient learning process. However, existing methods often require these experiences to be successful and may overly exploit them, which can cause the agent to adopt suboptimal behaviors. This paper develops an approach that uses diverse past trajectories for faster and more efficient online RL, even if these trajectories are suboptimal or not highly rewarded. The proposed algorithm combines a policy improvement step with an additional exploration step using offline demonstration data. The main contribution of this paper is that by regarding diverse past trajectories as guidance, instead of imitating them, our method directs its policy to follow and expand past trajectories while still being able to learn without rewards and approach optimality. Furthermore, a novel diversity measurement is introduced to maintain the team's diversity and regulate exploration. The proposed algorithm is evaluated on discrete and continuous control tasks with sparse and deceptive rewards. Compared with the existing RL methods, the experimental results indicate that our proposed algorithm is significantly better than the baseline methods regarding diverse exploration and avoiding local optima.

Qingmao Yao, Zhichao Lei, Tianyuan Chen et al.

Offline Reinforcement Learning (RL) struggles with distributional shifts, leading to the $Q$-value overestimation for out-of-distribution (OOD) actions. Existing methods address this issue by imposing constraints; however, they often become overly conservative when evaluating OOD regions, which constrains the $Q$-function generalization. This over-constraint issue results in poor $Q$-value estimation and hinders policy improvement. In this paper, we introduce a novel approach to achieve better $Q$-value estimation by enhancing $Q$-function generalization in OOD regions within Convex Hull and its Neighborhood (CHN). Under the safety generalization guarantees of the CHN, we propose the Smooth Bellman Operator (SBO), which updates OOD $Q$-values by smoothing them with neighboring in-sample $Q$-values. We theoretically show that SBO approximates true $Q$-values for both in-sample and OOD actions within the CHN. Our practical algorithm, Smooth Q-function OOD Generalization (SQOG), empirically alleviates the over-constraint issue, achieving near-accurate $Q$-value estimation. On the D4RL benchmarks, SQOG outperforms existing state-of-the-art methods in both performance and computational efficiency.

Guojian Wang, Faguo Wu, Xiao Zhang

Mastering deep reinforcement learning (DRL) proves challenging in tasks featuring scant rewards. These limited rewards merely signify whether the task is partially or entirely accomplished, necessitating various exploration actions before the agent garners meaningful feedback. Consequently, the majority of existing DRL exploration algorithms struggle to acquire practical policies within a reasonable timeframe. To address this challenge, we introduce an approach leveraging offline demonstration trajectories for swifter and more efficient online RL in environments with sparse rewards. Our pivotal insight involves treating offline demonstration trajectories as guidance, rather than mere imitation, allowing our method to learn a policy whose distribution of state-action visitation marginally matches that of offline demonstrations. We specifically introduce a novel trajectory distance relying on maximum mean discrepancy (MMD) and cast policy optimization as a distance-constrained optimization problem. We then illustrate that this optimization problem can be streamlined into a policy-gradient algorithm, integrating rewards shaped by insights from offline demonstrations. The proposed algorithm undergoes evaluation across extensive discrete and continuous control tasks with sparse and misleading rewards. The experimental findings demonstrate the significant superiority of our proposed algorithm over baseline methods concerning diverse exploration and the acquisition of an optimal policy.

Guojian Wang, Faguo Wu, Xiao Zhang et al.

The sparsity of reward feedback remains a challenging problem in online deep reinforcement learning (DRL). Previous approaches have utilized offline demonstrations to achieve impressive results in multiple hard tasks. However, these approaches place high demands on demonstration quality, and obtaining expert-like actions is often costly and unrealistic. To tackle these problems, we propose a simple and efficient algorithm called Policy Optimization with Smooth Guidance (POSG), which leverages a small set of state-only demonstrations (where expert action information is not included in demonstrations) to indirectly make approximate and feasible long-term credit assignments and facilitate exploration. Specifically, we first design a trajectory-importance evaluation mechanism to determine the quality of the current trajectory against demonstrations. Then, we introduce a guidance reward computation technology based on trajectory importance to measure the impact of each state-action pair, fusing the demonstrator's state distribution with reward information into the guidance reward. We theoretically analyze the performance improvement caused by smooth guidance rewards and derive a new worst-case lower bound on the performance improvement. Extensive results demonstrate POSG's significant advantages in control performance and convergence speed in four sparse-reward environments, including the grid-world maze, Hopper-v4, HalfCheetah-v4, and Ant maze. Notably, the specific metrics and quantifiable results are investigated to demonstrate the superiority of POSG.