arXiv subjects

Hua Cai

Publications and source records attributed to Hua Cai.

8 recordsLinked to original sources

arXiv2026

Reconciling Contradictory Views on the Effectiveness of SFT in LLMs: An Interaction Perspective

This paper explores a scientific question in supervised fine-tuning (SFT): why SFT is broadly effective for small-scale deep neural networks, yet can produce inconsistent or even detrimental effects when applied to large language models (LLMs). Recent advances in interaction-based explanations suggest that interactions between words/tokens provide a faithful metric for quantifying the inference patterns encoded by LLMs. We find that the evolution of interactions during SFT can effectively explain the inconsistent effectiveness of SFT for LLMs. Specifically, we find that (1) SFT primarily removes noise-like interactions, while rarely acquiring reliable new interactions. (2) This denoising stage is extremely brief, after which continued fine-tuning tends to introduce overfitted interactions. We validate these findings across multiple LLMs and datasets. Our findings provide new insights into early stopping and offer practical guidance for LLM training.

cs.AI↗

arXiv2025

Unilaw-R1: A Large Language Model for Legal Reasoning with Reinforcement Learning and Iterative Inference

Reasoning-focused large language models (LLMs) are rapidly evolving across various domains, yet their capabilities in handling complex legal problems remains underexplored. In this paper, we introduce Unilaw-R1, a large language model tailored for legal reasoning. With a lightweight 7-billion parameter scale, Unilaw-R1 significantly reduces deployment cost while effectively tackling three core challenges in the legal domain: insufficient legal knowledge, unreliable reasoning logic, and weak business generalization. To address these issues, we first construct Unilaw-R1-Data, a high-quality dataset containing 17K distilled and screened chain-of-thought (CoT) samples. Based on this, we adopt a two-stage training strategy combining Supervised Fine-Tuning (SFT) and Reinforcement Learning (RL), which significantly boosts the performance on complex legal reasoning tasks and supports interpretable decision-making in legal AI applications. To assess legal reasoning ability, we also introduce Unilaw-R1-Eval, a dedicated benchmark designed to evaluate models across single- and multi-choice legal tasks. Unilaw-R1 demonstrates strong results on authoritative benchmarks, outperforming all models of similar scale and achieving performance on par with the much larger DeepSeek-R1-Distill-Qwen-32B (54.9%). Following domain-specific training, it also showed significant gains on LawBench and LexEval, exceeding Qwen-2.5-7B-Instruct (46.6%) by an average margin of 6.6%.

cs.CL↗

arXiv2025

ViTSP: A Vision Language Models Guided Framework for Solving Large-Scale Traveling Salesman Problems

Solving the Traveling Salesman Problem (TSP) is NP-hard yet fundamental for a wide range of real-world applications. Classical exact methods face challenges in scaling, and heuristic methods often require domain-specific parameter calibration. While learning-based approaches have shown promise, they suffer from poor generalization and limited scalability due to fixed training data. This work proposes ViTSP, a novel framework that leverages pre-trained vision language models (VLMs) to visually guide the solution process for large-scale TSPs. The VLMs function to identify promising small-scale subproblems from a visualized TSP instance, which are then efficiently optimized using an off-the-shelf solver to improve the global solution. ViTSP bypasses the dedicated model training at the user end while maintaining effectiveness across diverse instances. Experiments on real-world TSP instances ranging from 1k to 88k nodes demonstrate that ViTSP consistently achieves solutions with average optimality gaps of 0.24%, outperforming existing learning-based methods. Under the same runtime budget, it surpasses the best-performing heuristic solver, LKH-3, by reducing its gaps by 3.57% to 100%, particularly on very-large-scale instances with more than 10k nodes. Our framework offers a new perspective in hybridizing pre-trained generative models and operations research solvers in solving combinatorial optimization problems. The framework holds potential for integration into more complex real-world logistics systems. The code is available at https://github.itap.purdue.edu/uSMART/ViTSP_ICLR2026.

cs.AI↗

arXiv2025

EmoHead: Emotional Talking Head via Manipulating Semantic Expression Parameters

Generating emotion-specific talking head videos from audio input is an important and complex challenge for human-machine interaction. However, emotion is highly abstract concept with ambiguous boundaries, and it necessitates disentangled expression parameters to generate emotionally expressive talking head videos. In this work, we present EmoHead to synthesize talking head videos via semantic expression parameters. To predict expression parameter for arbitrary audio input, we apply an audio-expression module that can be specified by an emotion tag. This module aims to enhance correlation from audio input across various emotions. Furthermore, we leverage pre-trained hyperplane to refine facial movements by probing along the vertical direction. Finally, the refined expression parameters regularize neural radiance fields and facilitate the emotion-consistent generation of talking head videos. Experimental results demonstrate that semantic expression parameters lead to better reconstruction quality and controllability.

cs.CV↗

arXiv2024

Muon beamtest results of high-density glass scintillator tiles

To achieve the physics goal of precisely measure the Higgs, Z, W bosons and the top quark, future electron-positron colliders require that their detector system has excellent jet energy resolution. One feasible technical option is the high granular calorimetery based on the particle flow algorithm (PFA). A new high-granularity hadronic calorimeter with glass scintillator tiles (GSHCAL) has been proposed, which focus on the significant improvement of hadronic energy resolution with a notable increase of the energy sampling fraction by using high-density glass scintillator tiles. The minimum ionizing particle (MIP) response of a glass scintillator tile is crucial to the hadronic calorimeter, so a dedicated beamtest setup was developed for testing the first batch of large-size glass scintillators. The maximum MIP response of the first batch of glass scintillator tiles can reach up to 107 p.e./MIP, which essentially meets the design requirements of the CEPC GSHCAL. An optical simulation model of a single glass scintillator tile has been established, and the simulation results are consistent with the beamtest results.

physics.ins-det↗

arXiv2023

Improving Empathetic Dialogue Generation by Dynamically Infusing Commonsense Knowledge

In empathetic conversations, individuals express their empathy towards others. Previous work has mainly focused on generating empathetic responses by utilizing the speaker's emotion. Besides, external commonsense knowledge has been applied to enhance the system's understandings of the speaker's situation. However, given an event, commonsense knowledge base contains various relations, potentially leading to confusion for the dialogue system. Consequently, inconsistencies arise among the emotion, generated response and speaker's contextual information. To this end, we propose a novel approach for empathetic response generation, which incorporates an adaptive module for commonsense knowledge selection to ensure consistency between the generated empathetic responses and the speaker's situation. This selected knowledge is used to refine the commonsense cognition and empathy expression for generated responses. Experimental results show that our approach significantly outperforms baseline models in both automatic and human evaluations, exhibiting the generation of more coherent and empathetic responses. Moreover, case studies highlight the interpretability of knowledge selection in the responses and the effectiveness of adaptive module in our model. Code: https://github.com/Hanscal/DCKS.

cs.CL↗

arXiv2021

A Scalable 256-Elements E-Band Phased-Array Transceiver for Broadband Communication

For E-band wireless communications, a high gain steerable antenna with sub-arrays is desired to reduce the implementation complexity. This paper presents an E-band communication link with 256-elements antennas based on 8-elements sub-arrays and four beam-forming chips in silicon germanium (SiGe) bipolar complementary metal-oxide-semiconductor (BiCMOS), which is packaged on a 19-layer low temperature co-fired ceramic (LTCC) substrate. After the design and manufacture of the 256-elements antenna, a fast near-field calibration method is proposed for calibration, where a single near-field measurement is required. Then near-field to far-field (NFFF) transform and far-field to near-field (FFNF) transform are used for the bore-sight calibration. The comparison with high frequency structure simulator (HFSS) is utilized for the non-bore-sight calibration. Verified on the 256-elements antenna, the beam-forming performance measured in the chamber is in good agreement with the simulations. The communication in the office environment is also realized using a fifth generation (5G) new radio (NR) system, whose bandwidth is 400 megahertz (MHz) and waveform format is orthogonal frequency division multiplexing (OFDM) with 120 kilohertz (kHz) sub-carrier spacing.

eess.SP↗

arXiv2020

Machine Learning in/for Blockchain: Future and Challenges

Machine learning and blockchain are two of the most noticeable technologies in recent years. The first one is the foundation of artificial intelligence and big data, and the second one has significantly disrupted the financial industry. Both technologies are data-driven, and thus there are rapidly growing interests in integrating them for more secure and efficient data sharing and analysis. In this paper, we review the research on combining blockchain and machine learning technologies and demonstrate that they can collaborate efficiently and effectively. In the end, we point out some future directions and expect more researches on deeper integration of the two promising technologies.

cs.CR↗