arXiv subjects

Junyi Wei

Publications and source records attributed to Junyi Wei.

9 recordsLinked to original sources

arXiv2026

TDD-Agent: Test-Driven Reasoning for Code Generation

Large Language Models (LLMs) have achieved remarkable progress in code generation, yet ensuring correctness in complex, repository-level tasks remains challenging. Existing approaches often use generated tests as static post-hoc validators, which limits their ability to guide implementation and may introduce misleading feedback when the tests themselves are incomplete or incorrect. In this paper, we introduce TDD-Agent, which operationalizes the test-driven development paradigm for code generation. TDD-Agent first prompts the model to generate executable tests, encouraging it to clarify expected behaviors before implementation, and then performs iterative dual-track refinement over both the generated code and tests using execution feedback. We first isolate the effect of test-first reasoning through a prompt variant TDD-prompt on LiveCodeBench, where it consistently improves upon reasoning-based prompting baselines. Building on this finding, we evaluate the full TDD-Agent framework on RepoEval, a repository-level benchmark, and show that it consistently outperforms retrieval-based and agent-based baselines. Additional analyses show that iterative refinement improves not only code correctness but also the effectiveness of the generated tests, yielding higher pass rates, coverage, and mutation scores, suggesting that tests can serve as evolving reasoning artifacts rather than fixed validators. Our source code is available at https://anonymous.4open.science/r/TDD-Agent-Framework-6370/.

cs.SE↗

arXiv2026

The non-topological $Z^\prime$ string in the 331 model and its classical stability

We study the classical stability of a non-topological $Z^\prime$ string in the minimal 331 model, which arises from the maximal symmetry breaking pattern of an ${\rm SU}(6)$ toy model. Two Higgs triplets are introduced according to the emergent global symmetries in the fermionic sector of the ${\rm SU}(6)$ toy model, which will achieve the sequential symmetry breaking of ${\rm SU}(3)_c\otimes {\rm SU}(3)_W \otimes {\rm U}(1)_X\to {\rm SU}(3)_c\otimes {\rm SU}(2)_W \otimes {\rm U}(1)_Y$. By analyzing small perturbations around the string background and solving the coupled Helmholtz equations numerically, we find that the string is stable only near the semilocal limit of $\vartheta_S \approx \frac{\pi}{2}$, even when Higgs self-couplings are tuned to minimize instabilities. This suggests that such non-topological strings are unlikely to exist in unified theories based on ${\rm SU}(N>5)$ Lie algebras.

hep-ph↗

arXiv2024

Global Tests for Smoothed Functions in Mean Field Variational Additive Models

Variational regression methods are an increasingly popular tool for their efficient estimation of complex. Given the mixed model representation of penalized effects, additive regression models with smoothed effects and scalar-on-function regression models can be fit relatively efficiently in a variational framework. However, inferential procedures for smoothed and functional effects in such a context is limited. We demonstrate that by using the Mean Field Variational Bayesian (MFVB) approximation to the additive model and the subsequent Coordinate Ascent Variational Inference (CAVI) algorithm, we can obtain a form of the estimated effects required of a Frequentist test for semiparametric curves. We establish MFVB approximations and CAVI algorithms for both Gaussian and binary additive models with an arbitrary number of smoothed and functional effects. We then derive a global testing framework for smoothed and functional effects. Our empirical study demonstrates that the test maintains good Frequentist properties in the variational framework and can be used to directly test results from a converged, MFVB approximation and CAVI algorithm. We illustrate the applicability of this approach in a wide range of data illustrations.

stat.ME↗

arXiv2024

Why Larger Language Models Do In-context Learning Differently?

Large language models (LLM) have emerged as a powerful tool for AI, with the key ability of in-context learning (ICL), where they can perform well on unseen tasks based on a brief series of task examples without necessitating any adjustments to the model parameters. One recent interesting mysterious observation is that models of different scales may have different ICL behaviors: larger models tend to be more sensitive to noise in the test context. This work studies this observation theoretically aiming to improve the understanding of LLM and ICL. We analyze two stylized settings: (1) linear regression with one-layer single-head linear transformers and (2) parity classification with two-layer multiple attention heads transformers (non-linear data and non-linear model). In both settings, we give closed-form optimal solutions and find that smaller models emphasize important hidden features while larger ones cover more hidden features; thus, smaller models are more robust to noise while larger ones are more easily distracted, leading to different ICL behaviors. This sheds light on where transformers pay attention to and how that affects ICL. Preliminary experimental results on large base and chat models provide positive support for our analysis.

cs.LG↗

arXiv2024

Towards Few-Shot Adaptation of Foundation Models via Multitask Finetuning

Foundation models have emerged as a powerful tool for many AI problems. Despite the tremendous success of foundation models, effective adaptation to new tasks, particularly those with limited labels, remains an open question and lacks theoretical understanding. An emerging solution with recent success in vision and NLP involves finetuning a foundation model on a selection of relevant tasks, before its adaptation to a target task with limited labeled samples. In this paper, we study the theoretical justification of this multitask finetuning approach. Our theoretical analysis reveals that with a diverse set of related tasks, this multitask finetuning leads to reduced error in the target task, in comparison to directly adapting the same pretrained model. We quantify the relationship between finetuning tasks and target tasks by diversity and consistency metrics, and further propose a practical task selection algorithm. We substantiate our theoretical claims with extensive empirical evidence. Further, we present results affirming our task selection algorithm adeptly chooses related finetuning tasks, providing advantages to the model performance on target tasks. We believe our study shed new light on the effective adaptation of foundation models to new tasks that lack abundant labels. Our code is available at https://github.com/OliverXUZY/Foudation-Model_Multitask.

cs.LG↗

arXiv2023

Interfacing Foundation Models' Embeddings

Foundation models possess strong capabilities in reasoning and memorizing across modalities. To further unleash the power of foundation models, we present FIND, a generalized interface for aligning foundation models' embeddings with unified image and dataset-level understanding spanning modality and granularity. As shown in the teaser figure, a lightweight transformer interface without tuning any foundation model weights is enough for segmentation, grounding, and retrieval in an interleaved manner. The proposed interface has the following favorable attributes: (1) Generalizable. It applies to various tasks spanning retrieval, segmentation, etc., under the same architecture and weights. (2) Interleavable. With the benefit of multi-task multi-modal training, the proposed interface creates an interleaved shared embedding space. (3) Extendable. The proposed interface is adaptive to new tasks, and new models. In light of the interleaved embedding space, we introduce FIND-Bench, which introduces new training and evaluation annotations to the COCO dataset for interleaved segmentation and retrieval. We are the first work aligning foundations models' embeddings for interleave understanding. Meanwhile, our approach achieves state-of-the-art performance on FIND-Bench and competitive performance on standard retrieval and segmentation settings.

cs.CV↗

arXiv2023

Provable Guarantees for Neural Networks via Gradient Feature Learning

Neural networks have achieved remarkable empirical performance, while the current theoretical analysis is not adequate for understanding their success, e.g., the Neural Tangent Kernel approach fails to capture their key feature learning ability, while recent analyses on feature learning are typically problem-specific. This work proposes a unified analysis framework for two-layer networks trained by gradient descent. The framework is centered around the principle of feature learning from gradients, and its effectiveness is demonstrated by applications in several prototypical problems, such as mixtures of Gaussians and parity functions. The framework also sheds light on interesting network learning phenomena such as feature learning beyond kernels and the lottery ticket hypothesis.

cs.LG↗

arXiv2022

A Theoretical Analysis on Feature Learning in Neural Networks: Emergence from Inputs and Advantage over Fixed Features

An important characteristic of neural networks is their ability to learn representations of the input data with effective features for prediction, which is believed to be a key factor to their superior empirical performance. To better understand the source and benefit of feature learning in neural networks, we consider learning problems motivated by practical data, where the labels are determined by a set of class relevant patterns and the inputs are generated from these along with some background patterns. We prove that neural networks trained by gradient descent can succeed on these problems. The success relies on the emergence and improvement of effective features, which are learned among exponentially many candidates efficiently by exploiting the data (in particular, the structure of the input distribution). In contrast, no linear models on data-independent features of polynomial sizes can learn to as good errors. Furthermore, if the specific input structure is removed, then no polynomial algorithm in the Statistical Query model can learn even weakly. These results provide theoretical evidence showing that feature learning in neural networks depends strongly on the input structure and leads to the superior performance. Our preliminary experimental results on synthetic and real data also provide positive support.

cs.LG↗

arXiv2015

How Market Structure Drives Commodity Prices

We introduce an agent-based model, in which agents set their prices to maximize profit. At steady state the market self-organizes into three groups: excess producers, consumers and balanced agents, with prices determined by their own resource level and a couple of macroscopic parameters that emerge naturally from the analysis, akin to mean-field parameters in statistical mechanics. When resources are scarce prices rise sharply below a turning point that marks the disappearance of excess producers. To compare the model with real empirical data, we study the relations between commodity prices and stock-to-use ratios of a range of commodities such as agricultural products and metals. By introducing an elasticity parameter to mitigate noise and long-term changes in commodities data, we confirm the trend of rising prices, provide evidence for turning points, and indicate yield points for less essential commodities.

q-fin.GN↗