Hanwei Xu

Hanwei Xu, Yujun Chen, Yulun Du et al.

Prompt-based techniques have demostrated great potential for improving the few-shot generalization of pretrained language models. However, their performance heavily relies on the manual design of prompts and thus requires a lot of human efforts. In this paper, we introduce Genetic Prompt Search (GPS) to improve few-shot learning with prompts, which utilizes a genetic algorithm to automatically search for high-performing prompts. GPS is gradient-free and requires no update of model parameters but only a small validation set. Experiments on diverse datasets proved the effectiveness of GPS, which outperforms manual prompts by a large margin of 2.6 points. Our method is also better than other parameter-efficient tuning methods such as prompt tuning.

DeepSeek-AI, Aixin Liu, Aoxue Mei et al.

We introduce DeepSeek-V3.2, a model that harmonizes high computational efficiency with superior reasoning and agent performance. The key technical breakthroughs of DeepSeek-V3.2 are as follows: (1) DeepSeek Sparse Attention (DSA): We introduce DSA, an efficient attention mechanism that substantially reduces computational complexity while preserving model performance in long-context scenarios. (2) Scalable Reinforcement Learning Framework: By implementing a robust reinforcement learning protocol and scaling post-training compute, DeepSeek-V3.2 performs comparably to GPT-5. Notably, our high-compute variant, DeepSeek-V3.2-Speciale, surpasses GPT-5 and exhibits reasoning proficiency on par with Gemini-3.0-Pro, achieving gold-medal performance in both the 2025 International Mathematical Olympiad (IMO) and the International Olympiad in Informatics (IOI). (3) Large-Scale Agentic Task Synthesis Pipeline: To integrate reasoning into tool-use scenarios, we developed a novel synthesis pipeline that systematically generates training data at scale. This methodology facilitates scalable agentic post-training, yielding substantial improvements in generalization and instruction-following robustness within complex, interactive environments.

Kedi Zheng, Hanwei Xu, Zeyang Long et al.

The growing penetration of electric vehicles (EVs) significantly changes typical load curves in smart grids. With the development of fast charging technology, the volatility of EV charging demand is increasing, which requires additional flexibility for real-time power balance. The forecasting of EV charging demand involves probabilistic modeling of high dimensional time series dynamics across diverse electric vehicle charging stations (EVCSs). This paper studies the forecasting problem of multiple EVCS in a hierarchical probabilistic manner. For each charging station, a deep learning model based on a partial input convex neural network (PICNN) is trained to predict the day-ahead charging demand's conditional distribution, preventing the common quantile crossing problem in traditional quantile regression models. Then, differentiable convex optimization layers (DCLs) are used to reconcile the scenarios sampled from the distributions to yield coherent scenarios that satisfy the hierarchical constraint. It learns a better weight matrix for adjusting the forecasting results of different targets in a machine-learning approach compared to traditional optimization-based hierarchical reconciling methods. Numerical experiments based on real-world EV charging data are conducted to demonstrate the efficacy of the proposed method.

Hanwei Xu, Yujun Chen, Yulun Du et al.

We propose a multitask pretraining approach ZeroPrompt for zero-shot generalization, focusing on task scaling and zero-shot prompting. While previous models are trained on only a few dozen tasks, we scale to 1,000 tasks for the first time using real-world data. This leads to a crucial discovery that task scaling can be an efficient alternative to model scaling; i.e., the model size has little impact on performance with an extremely large number of tasks. Our results show that task scaling can substantially improve training efficiency by 30 times in FLOPs. Moreover, we present a prompting method that incorporates a genetic algorithm to automatically search for the best prompt for unseen tasks, along with a few other improvements. Empirically, ZeroPrompt substantially improves both the efficiency and the performance of zero-shot learning across a variety of academic and production datasets.