Posts by Collection

publications

Differentially Private Neural Network Training Under the Hidden State Assumption

Published in arXiv preprint arXiv:2407.08233, 2024

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.

Paper

Paper

EvoAlpha: Evolutionary Alpha Factor Discovery with Large Language Models

Published in NeurIPS 2025 Workshop on Generative AI in Finance, 2025

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.

Paper

Paper

DualOptim: Enhancing Efficacy and Stability in Machine Unlearning with Dual Optimizers

Published in Neural Information Processing Systems (NeurIPS) 2025, 2025

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.

Paper

Paper Code Poster

AlphaBench: Benchmarking Large Language Models in Formulaic Alpha Factor Mining

Published in International Conference on Learning Representations (ICLR) 2026, 2026

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.

Website

CommuteProp: Decoupled Training for Communication Bound Split LLM Fine-Tuning

Published in arXiv preprint arXiv:2610.05105, 2026

CommuteProp accelerates split learning for large language model fine-tuning by running cross-block gradient propagation and local weight updates concurrently. This asynchronous schedule overlaps network transfers with computation, reducing the idle time imposed by synchronous client-server training. The work analyzes convergence and asynchronous errors, and introduces an NS preconditioner to address noise from stale information. Experiments demonstrate higher training throughput with comparable accuracy and compatibility with existing privacy-enhancing techniques.

Paper

Paper

Arbitrage-Free Implied Volatility Surface Reconstruction via Additive Martingales

Published in arXiv preprint arXiv:2610.25671, 2026

AddMart reconstructs implied volatility surfaces from sparse option quotes using an exponential additive martingale representation. Its architecture encodes static no-arbitrage constraints through non-negative increment densities, allowing risk-neutral distributions and implied volatilities to be recovered without relying on penalty terms to enforce validity. Evaluations across ten real-option datasets show competitive reconstruction performance with negligible static arbitrage, alongside estimates of value at risk, conditional value at risk, and crash probabilities.

Paper

Paper