We propose EvoAlpha, a framework that leverages large language models to automatically discover formulaic alpha factors for quantitative trading through an evolutionary search process. Our approach combines the code generation capabilities of LLMs with evolutionary algorithms to iteratively create, evaluate, and refine trading signals, enabling the discovery of novel alpha factors that outperform traditional hand-crafted approaches.

Existing machine unlearning (MU) approaches exhibit significant sensitivity to hyperparameters, requiring meticulous tuning that limits practical deployment. In this work, we first empirically demonstrate the instability and suboptimal performance of existing popular MU methods when deployed in different scenarios. To address this issue, we propose Dual Optimizer (DualOptim), which incorporates adaptive learning rate and decoupled momentum factors. Empirical and theoretical evidence demonstrates that DualOptim contributes to effective and stable unlearning. Through extensive experiments, we show that DualOptim can significantly boost MU efficacy and stability across diverse tasks, including image classification, image generation, and large language models, making it a versatile approach to empower existing MU algorithms.

We introduce AlphaBench, a comprehensive benchmark designed to evaluate large language models on the task of formulaic alpha factor mining for quantitative finance. Our benchmark provides standardized evaluation protocols, diverse task scenarios, and rigorous metrics to assess the capability of LLMs in generating effective trading signals across different market conditions and asset classes.

We introduce DP-DT, a framework that separates representation learning from privacy protection mechanisms, restricting noise injection to weight aggregation while using auxiliary models for feature extraction on private data. We establish convergence guarantees for non-convex objectives and show that privacy loss can converge to a constant bound rather than accumulating with iterations under the Hidden State Assumption. Empirical validation across vision and language tasks shows improved privacy-utility trade-offs compared to existing methods.