Xiaopeng Wu

Runjin Chen, Mingxuan Ju, Ngoc Bui et al.

Generative recommendation systems, driven by large language models (LLMs), present an innovative approach to predicting user preferences by modeling items as token sequences and generating recommendations in a generative manner. A critical challenge in this approach is the effective tokenization of items, ensuring that they are represented in a form compatible with LLMs. Current item tokenization methods include using text descriptions, numerical strings, or sequences of discrete tokens. While text-based representations integrate seamlessly with LLM tokenization, they are often too lengthy, leading to inefficiencies and complicating accurate generation. Numerical strings, while concise, lack semantic depth and fail to capture meaningful item relationships. Tokenizing items as sequences of newly defined tokens has gained traction, but it often requires external models or algorithms for token assignment. These external processes may not align with the LLM's internal pretrained tokenization schema, leading to inconsistencies and reduced model performance. To address these limitations, we propose a self-improving item tokenization method that allows the LLM to refine its own item tokenizations during training process. Our approach starts with item tokenizations generated by any external model and periodically adjusts these tokenizations based on the LLM's learned patterns. Such alignment process ensures consistency between the tokenization and the LLM's internal understanding of the items, leading to more accurate recommendations. Furthermore, our method is simple to implement and can be integrated as a plug-and-play enhancement into existing generative recommendation systems. Experimental results on multiple datasets and using various initial tokenization strategies demonstrate the effectiveness of our method, with an average improvement of 8\% in recommendation performance.

Zexing Zhao, Guangsi Shi, Xiaopeng Wu et al.

Molecular property prediction is a key component of AI-driven drug discovery and molecular characterization learning. Despite recent advances, existing methods still face challenges such as limited ability to generalize, and inadequate representation of learning from unlabeled data, especially for tasks specific to molecular structures. To address these limitations, we introduce DIG-Mol, a novel self-supervised graph neural network framework for molecular property prediction. This architecture leverages the power of contrast learning with dual interaction mechanisms and unique molecular graph enhancement strategies. DIG-Mol integrates a momentum distillation network with two interconnected networks to efficiently improve molecular characterization. The framework's ability to extract key information about molecular structure and higher-order semantics is supported by minimizing loss of contrast. We have established DIG-Mol's state-of-the-art performance through extensive experimental evaluation in a variety of molecular property prediction tasks. In addition to demonstrating superior transferability in a small number of learning scenarios, our visualizations highlight DIG-Mol's enhanced interpretability and representation capabilities. These findings confirm the effectiveness of our approach in overcoming challenges faced by traditional methods and mark a significant advance in molecular property prediction.

Jingyu Liu, Xiaopeng Wu, Jingquan Peng et al.

Building autonomous agents capable of solving long-horizon, real-world tasks has garnered significant research interest. But outcome based rewards may cause reward miscalibration which means it might mistakenly allocate positive reward to flawed middle steps which is regarded as the key reason making the bad actions being reinforced during training. However we reveal that outcome based reward ensures expected negative advantage for those flawed middle steps, which means the flawed actions should be punished during training. Even accounting for the ``squeezing effect", the probability mass of good actions should increase and the actor should gradually get rid of harmful actions. This shows that flawed actions should be punished during training. We further identify gradient coupling between similar samples as a key issue in agentic RL, the input prompt is extremely similar and the output action space is limited, therefore during training, gradients from well-performing samples can inadvertently strengthen suboptimal or incorrect actions due to similar input observation and output actions. We show that with gradient coupling, some flawed actions might be enhanced. To address this, we propose training the actor to classify good or bad actions to separate the embedding of good/bad actions and alleviate the gradient interference, extensive experiments shows its effectiveness.