Zishi Zhang

Qinghao Guan, Yuchen Pan, Donghao Li et al.

In religion and theology studies, spirituality has garnered significant research attention for the reason that it not only transcends culture but offers unique experience to each individual. However, social scientists often rely on limited datasets, which are basically unavailable online. In this study, we collaborated with social scientists to develop a high-quality multimedia multi-modal datasets, \textbf{SACRED}, in which the faithfulness of classification is guaranteed. Using \textbf{SACRED}, we evaluated the performance of 13 popular LLMs as well as traditional rule-based and fine-tuned approaches. The result suggests DeepSeek-V3 model performs well in classifying such abstract concepts (i.e., 79.19\% accuracy in the Quora test set), and the GPT-4o-mini model surpassed the other models in the vision tasks (63.99\% F1 score). Purportedly, this is the first annotated multi-modal dataset from online spirituality communication. Our study also found a new type of connectedness which is valuable for communication science studies.

Ruihan Zhou, Zishi Zhang, Jinhui Han et al.

Forecasting the life-cycle trajectory of a newly launched product is important for launch planning, resource allocation, and early risk assessment. This task is especially difficult in the pre-launch and early post-launch phases, when product-specific outcome history is limited or unavailable, creating a cold-start problem. In these phases, firms must make decisions before demand patterns become reliably observable, while early signals are often sparse, noisy, and unstable We propose the Conditional Diffusion Life-cycle Forecaster (CDLF), a conditional generative framework for forecasting new-product life-cycle trajectories under cold start. CDLF combines three sources of information: static descriptors, reference trajectories from similar products, and newly arriving observations when available. Here, static descriptors refer to structured pre-launch characteristics of the product, such as category, price tier, brand or organization identity, scale, and access conditions. This structure allows the model to condition forecasts on relevant product context and to update them adaptively over time without retraining, yielding flexible multi-modal predictive distributions under extreme data scarcity. The method satisfies consistency with a horizon-uniform distributional error bound for recursive generation. Across studies on Intel microprocessor stock keeping unit (SKU) life cycles and the platform-mediated adoption of open large language model repositories, CDLF delivers more accurate point forecasts and higher-quality probabilistic forecasts than classical diffusion models, Bayesian updating approaches, and other state-of-the-art machine-learning baselines.

Zehao Li, Tao Ren, Zishi Zhang et al.

Large language model (LLM) training is often bottlenecked by memory constraints and stochastic gradient noise in extremely high-dimensional parameter spaces. Motivated by empirical evidence that many LLM gradient matrices are effectively low-rank during training, we present an unbiased, memory-efficient, low-rank matrix estimator with the lowest variance that is applicable across common stochastic gradient estimation paradigms. The core idea is to project a high-dimensional stochastic gradient estimator onto a random low-dimensional subspace and lift it back, reducing memory while keeping the estimator unbiased and controlling mean-squared error via an optimally designed projection distribution, including Haar--Stiefel projections. The projection distribution is derived by solving a constrained functional optimization problem, yielding an optimal random projector that guides algorithm design. Empirically, the resulting low-rank gradient estimators deliver both practical memory savings and improved training behavior. In RoBERTa-large fine-tuning, our method attains the lowest peak GPU memory among compared methods (e.g., 3.83GB versus 16.7GB for full BP) while remaining competitive in accuracy; in autoregressive LLM pretraining (LLaMA-20M/60M/100M), our method outperforms the traditional methods, supporting the benefit of the proposed optimal projection strategy.

Zishi Zhang, Jinhui Han, Ming Hu et al.

We consider small-data, large-scale decision problems in which a firm must make many operational decisions simultaneously (e.g., across a large product portfolio) while observing only a few, potentially noisy, data points per instance. Inspired by the success of large language models (LLMs), we propose a pretrain-then-finetune approach built on a designed Transformer model to address this challenge. The model is first pretrained on large-scale, domain-informed synthetic data that encode managerial knowledge and structural features of the decision environment, and is then fine-tuned on real observations. This new pipeline offers two complementary advantages: pretraining injects domain knowledge into the learning process and enables the training of high-capacity models using abundant synthetic data, while finetuning adapts the pretrained model to the operational environment and improves alignment with the true data-generating regime. While we have leveraged the Transformer's state-of-the-art representational capacity, particularly its attention mechanism, to efficiently extract cross-task structure, our approach is not an off-the-shelf application. Instead, it relies on problem-specific architectural design and a tailored training procedure to match the decision setting. Theoretically, we develop the first comprehensive error analysis regarding Transformer learning in relevant contexts, establishing nonasymptotic guarantees that validate the method's effectiveness. Critically, our analysis reveals how pretraining and fine-tuning jointly determine performance, with the dominant contribution governed by whichever is more favorable. In particular, finetuning exhibits an economies-of-scale effect, whereby transfer learning becomes increasingly effective as the number of instances grows.

Zishi Zhang, Yijie Peng

This work aims to improve the sample efficiency of parallel large-scale ranking and selection (R&S) problems by leveraging correlation information. We modify the commonly used "divide and conquer" framework in parallel computing by adding a correlation-based clustering step, transforming it into "clustering and conquer". Analytical results under a symmetric benchmark scenario show that this seemingly simple modification yields an $\mathcal{O}(p)$ reduction in sample complexity for a widely used class of sample-optimal R&S procedures. Our approach enjoys two key advantages: 1) it does not require highly accurate correlation estimation or precise clustering, and 2) it allows for seamless integration with various existing R&S procedures, while achieving optimal sample complexity. Theoretically, we develop a novel gradient analysis framework to analyze sample efficiency and guide the design of large-scale R&S procedures. We also introduce a new parallel clustering algorithm tailored for large-scale scenarios. Finally, in large-scale AI applications such as neural architecture search, our methods demonstrate superior performance.

Tao Ren, Zishi Zhang, Jingyang Jiang et al.

The probabilistic diffusion model (DM), generating content by inferencing through a recursive chain structure, has emerged as a powerful framework for visual generation. After pre-training on enormous data, the model needs to be properly aligned to meet requirements for downstream applications. How to efficiently align the foundation DM is a crucial task. Contemporary methods are either based on Reinforcement Learning (RL) or truncated Backpropagation (BP). However, RL and truncated BP suffer from low sample efficiency and biased gradient estimation, respectively, resulting in limited improvement or, even worse, complete training failure. To overcome the challenges, we propose the Recursive Likelihood Ratio (RLR) optimizer, a Half-Order (HO) fine-tuning paradigm for DM. The HO gradient estimator enables the computation graph rearrangement within the recursive diffusive chain, making the RLR's gradient estimator an unbiased one with lower variance than other methods. We theoretically investigate the bias, variance, and convergence of our method. Extensive experiments are conducted on image and video generation to validate the superiority of the RLR. Furthermore, we propose a novel prompt technique that is natural for the RLR to achieve a synergistic effect.

Tao Ren, Jinyang Jiang, Hui Yang et al. · pku

Reinforcement learning with verifiable reward has recently emerged as a central paradigm for post-training large language models (LLMs); however, prevailing mean-based methods, such as Group Relative Policy Optimization (GRPO), suffer from entropy collapse and limited reasoning gains. We argue that these issues stem from overemphasizing high-probability output sequences while neglecting rare but informative reasoning paths. To address these challenges, we propose Risk-based Policy Optimization (RiskPO), which substitutes classical mean-based objectives with principled risk measures. Specifically, we introduce a Mixed Value-at-Risk objective that integrates weighted attention over multiple regions of the reward distribution, thereby amplifying gradient signals on challenging instances and preventing overconfident convergence. We further design a bundling scheme that aggregates multiple questions into bundles, thus enriching the feedback signal and yielding more stable and informative training dynamics. Theoretically, we prove that the risk-averse update alleviates entropy collapse and promotes exploration. Numerically, RiskPO achieves consistent and significant improvements in mathematical reasoning, multi-modal reasoning, and code generation benchmarks, surpassing GRPO and its variants on both Pass@1 and Pass@k metrics. Our results demonstrate that risk-based optimization provides a rigorous and effective paradigm for enhancing LLM reasoning capabilities.