arXiv subjects

Xiaochen Zhu

Publications and source records attributed to Xiaochen Zhu.

16 recordsLinked to original sources

arXiv2026

DeliChess: A Multi-party Dialogue Dataset for Deliberation in Chess Puzzle Solving

Multi-party dialogue is a critical setting for studying collaborative reasoning and decision-making, yet existing datasets rarely focus on structured, reasoning-intensive tasks. We introduce DeliChess, a dataset of group deliberation dialogues in which participants collaboratively solve multiple-choice chess puzzles. Participants first answer independently, then engage in multi-party deliberation and revise their individual answers. The dataset comprises 107 dialogues with full transcripts, pre- and post-deliberation choices, and utterance-level annotations of communicative function, epistemic stance, and usefulness for supporting deliberation. Our analyses show that greater diversity in initial solution quality is associated with larger gains, while answer trajectories reveal how deliberation can recover, discover, or lose strong answers. Cases in which groups surpass every independent answer are associated with sustained reasoning and epistemic openness. We further propose a diagnostic action-selection task and find that the tested LLMs show only weak agreement with human-attested helpful actions. Together, our dataset provides a testbed for modelling group reasoning, dialogue dynamics, and conditions for effective deliberation.

cs.CL↗

arXiv2026

PACZero: PAC-Private Fine-Tuning of Language Models via Sign Quantization

We introduce PACZero, a family of PAC-private zeroth-order mechanisms for fine-tuning large language models that delivers usable utility at $I(S^*; Y_{1:T})=0$. This privacy regime bounds the membership-inference attack (MIA) posterior success rate at the prior, an MIA-resistance level the DP framework matches only at $\varepsilon=0$ and infinite noise. All DP-ZO comparisons below are matched at the MIA posterior level. The key insight is that PAC Privacy charges mutual information only when the release depends on which candidate subset is the secret. Sign-quantizing subset-aggregated zeroth-order gradients creates frequent unanimity, steps at which every candidate subset agrees on the update direction; at these steps the released sign costs zero conditional mutual information. We propose two variants that span the privacy-utility trade-off: PACZero-MI (budgeted MI via exact calibration on the binary release) and PACZero-ZPL ($I=0$ via a uniform coin flip on disagreement steps). We evaluate on SST-2 and SQuAD with OPT-1.3B and OPT-6.7B in both LoRA and full-parameter tracks. On SST-2 OPT-1.3B full fine-tuning at $I=0$, PACZero-ZPL reaches ${88.99\pm0.91}$, within $2.1$pp of the non-private MeZO baseline ($91.1$ FT). No prior method produces usable utility in the high-privacy regime $\varepsilon<1$, and PACZero-ZPL obtains competitive SST-2 accuracy and nontrivial SQuAD F1 across OPT-1.3B and OPT-6.7B at $I=0$.

cs.LG↗

arXiv2026

SIMD-PAC-DB: Pretty Performant PAC Privacy

This work presents a highly optimized implementation of PAC-DB, a recent and promising database privacy model. We prove that our SIMD-PAC-DB can compute the same privatized answer with just a single query, instead of the 128 stochastic executions against different 50% database sub-samples needed by the original PAC-DB. Our key insight is that every bit of a hashed primary key can be seen to represent membership of such a sub-sample. We present new algorithms for approximate computation of stochastic aggregates based on these hashes, which, thanks to their SIMD-friendliness, run up to 40x faster than scalar equivalents. We release an open-source DuckDB community extension which includes a rewriter that PAC-privatizes arbitrary SQL queries. Our experiments on TPC-H, Clickbench, and SQLStorm evaluate thousands of queries in terms of performance and utility, significantly advancing the ease of use and functionality of privacy-aware data systems in practice.

cs.DB↗

arXiv2026

Private Prediction via PAC Privacy

Machine learning models are increasingly served behind APIs. This renders private prediction, i.e., privatizing a model's outputs rather than its parameters, a natural privacy target: model outputs are lower-dimensional and far more stable to training-data changes than weights. While differential privacy (DP) cannot effectively exploit this as it calibrates noise to worst-case sensitivity that is intractable to bound for non-convex models, we argue that PAC privacy is a natural fit for private prediction. It is instance-based, and calibrates noise to a black-box function's empirical stability to control mutual-information (MI) leakage. The missing ingredient is efficient, adaptive composition. Serving predictions means answering a long stream of adaptively chosen queries from untrusted users; existing composition either fails under adaptivity, grows quadratically, or reverts to input-independent, DP-like noise. We close this gap with a new adversarial composition result via adaptive noise calibration and prove that MI accumulates only linearly under adaptive and adversarial querying. Experiments across modalities show that prediction stability enables high utility even at a tiny per-query budget: on CIFAR-10, we achieve 87.79% accuracy with a per-query MI budget of $2^{-32}$. This enables serving one million queries while provably bounding membership-inference success to 51.08% -- the same guarantee as $(0.04, 10^{-5})$-DP. Further, in the presence of auxiliary public data, the large volume of PAC-private predictions enables us to distill a publishable model that can be queried without limit. Concretely, 210,000 private labels on an ImageNet subset distill into a student reaching 91.86% accuracy on CIFAR-10 with membership inference success bounded by 50.49%, comparable to $(0.02, 10^{-5})$-DP.

cs.LG↗

arXiv2026

Demystifying Multi-Agent Debate: The Role of Confidence and Diversity

Multi-agent debate (MAD) is widely used to improve large language model (LLM) performance through test-time scaling, yet recent work shows that vanilla MAD often underperforms simple majority vote despite higher computational cost. Studies show that, under homogeneous agents and uniform belief updates, debate preserves expected correctness and therefore cannot reliably improve outcomes. Drawing on findings from human deliberation and collective decision-making, we identify two key mechanisms missing from vanilla MAD: (i) diversity of initial viewpoints and (ii) explicit, calibrated confidence communication. We propose two lightweight interventions. First, a diversity-aware initialisation that selects a more diverse pool of candidate answers, increasing the likelihood that a correct hypothesis is present at the start of debate. Second, a confidence-modulated debate protocol in which agents express calibrated confidence and condition their updates on others' confidence. We show theoretically that diversity-aware initialisation improves the prior probability of MAD success without changing the underlying update dynamics, while confidence-modulated updates enable debate to systematically drift to the correct hypothesis. Empirically, across six reasoning-oriented QA benchmarks, our methods consistently outperform vanilla MAD and majority vote. Our results connect human deliberation with LLM-based debate and demonstrate that simple, principled modifications can substantially enhance debate effectiveness.

cs.CL↗

arXiv2026

Confidence Estimation for LLMs in Multi-turn Interactions

While confidence estimation is a promising direction for mitigating hallucinations in Large Language Models (LLMs), current research overwhelmingly focuses on single-turn settings. The dynamics of model confidence in multi-turn conversations, where context accumulates and ambiguity is progressively resolved, remain largely unexplored. This work presents the first systematic study of confidence estimation in multi-turn interactions, establishing a formal evaluation framework grounded in two key desiderata: per-turn calibration and monotonicity of confidence as more information becomes available. To facilitate this, we introduce novel metrics, including a length-normalized Expected Calibration Error (InfoECE), and a new "Hinter-Guesser" paradigm for generating controlled evaluation datasets. Our experiments reveal that widely-used confidence techniques struggle with calibration and monotonicity in multi-turn dialogues. In contrast, a novel logit-based probe we introduce, P(Sufficient), proves comparatively more effective, robustly tracking evidence accumulation and distinguishing it from conversational filler. Our work provides a foundational methodology for developing more reliable and trustworthy conversational agents.

cs.CL↗

arXiv2025

LoVeC: Reinforcement Learning for Better Verbalized Confidence in Long-Form Generations

Hallucination remains a major challenge for the safe and trustworthy deployment of large language models (LLMs) in factual content generation. Prior work has explored confidence estimation as an effective approach to hallucination detection, but often relies on post-hoc self-consistency methods that require computationally expensive sampling. Verbalized confidence offers a more efficient alternative, but existing approaches are largely limited to short-form question answering (QA) tasks and do not generalize well to open-ended generation. In this paper, we propose LoVeC (Long-form Verbalized Confidence), a novel reinforcement learning based method that trains LLMs to append an on-the-fly numerical confidence score to each generated statement during long-form generation. The confidence score serves as a direct and interpretable signal of the factuality of generation. We introduce two evaluation settings, free-form tagging and iterative tagging, to assess different verbalized confidence estimation methods. Experiments on three long-form QA datasets show that our RL-trained models achieve better calibration and generalize robustly across domains. Also, our method is highly efficient, being 20 times faster than traditional self-consistency methods while achieving better calibration.

cs.CL↗

arXiv2025

Collaborative Evaluation of Deepfake Text with Deliberation-Enhancing Dialogue Systems

The proliferation of generative models has presented significant challenges in distinguishing authentic human-authored content from deepfake content. Collaborative human efforts, augmented by AI tools, present a promising solution. In this study, we explore the potential of DeepFakeDeLiBot, a deliberation-enhancing chatbot, to support groups in detecting deepfake text. Our findings reveal that group-based problem-solving significantly improves the accuracy of identifying machine-generated paragraphs compared to individual efforts. While engagement with DeepFakeDeLiBot does not yield substantial performance gains overall, it enhances group dynamics by fostering greater participant engagement, consensus building, and the frequency and diversity of reasoning-based utterances. Additionally, participants with higher perceived effectiveness of group collaboration exhibited performance benefits from DeepFakeDeLiBot. These findings underscore the potential of deliberative chatbots in fostering interactive and productive group dynamics while ensuring accuracy in collaborative deepfake text detection. \textit{Dataset and source code used in this study will be made publicly available upon acceptance of the manuscript.

cs.CL↗

arXiv2024

Segment-Level Diffusion: A Framework for Controllable Long-Form Generation with Diffusion Language Models

Diffusion models have shown promise in text generation, but often struggle with generating long, coherent, and contextually accurate text. Token-level diffusion doesn't model word-order dependencies explicitly and operates on short, fixed output windows, while passage-level diffusion struggles with learning robust representations for long-form text. To address these challenges, we propose Segment-Level Diffusion (SLD), a framework that enhances diffusion-based text generation through text segmentation, robust representation training with adversarial and contrastive learning, and improved latent-space guidance. By segmenting long-form outputs into multiple latent representations and decoding them with an autoregressive decoder, SLD simplifies diffusion predictions and improves scalability. Experiments on four datasets demonstrate that, when compared to other diffusion and autoregressive baselines SLD achieves competitive or superior fluency, coherence, and contextual compatibility in automatic and human evaluations.

cs.CL↗

arXiv2024

Conformity in Large Language Models

The conformity effect describes the tendency of individuals to align their responses with the majority. Studying this bias in large language models (LLMs) is crucial, as LLMs are increasingly used in various information-seeking and decision-making tasks as conversation partners to improve productivity. Thus, conformity to incorrect responses can compromise their effectiveness. In this paper, we adapt psychological experiments to examine the extent of conformity in popular LLMs. Our findings reveal that all tested models exhibit varying levels of conformity toward the majority, regardless of their initial choice or correctness, across different knowledge domains. Notably, we are the first to show that LLMs are more likely to conform when they are more uncertain in their own prediction. We further explore factors that influence conformity, such as training paradigms and input characteristics, finding that instruction-tuned models are less susceptible to conformity, while increasing the naturalness of majority tones amplifies conformity. Finally, we propose two interventions, Devil's Advocate and Question Distillation, to mitigate conformity, providing insights into building more robust language models.

cs.CL↗

arXiv2023

ConvLab-3: A Flexible Dialogue System Toolkit Based on a Unified Data Format

Task-oriented dialogue (TOD) systems function as digital assistants, guiding users through various tasks such as booking flights or finding restaurants. Existing toolkits for building TOD systems often fall short of in delivering comprehensive arrays of data, models, and experimental environments with a user-friendly experience. We introduce ConvLab-3: a multifaceted dialogue system toolkit crafted to bridge this gap. Our unified data format simplifies the integration of diverse datasets and models, significantly reducing complexity and cost for studying generalization and transfer. Enhanced with robust reinforcement learning (RL) tools, featuring a streamlined training process, in-depth evaluation tools, and a selection of user simulators, ConvLab-3 supports the rapid development and evaluation of robust dialogue policies. Through an extensive study, we demonstrate the efficacy of transfer learning and RL and showcase that ConvLab-3 is not only a powerful tool for seasoned researchers but also an accessible platform for newcomers.

cs.CL↗

arXiv2023

Passive Inference Attacks on Split Learning via Adversarial Regularization

Split Learning (SL) has emerged as a practical and efficient alternative to traditional federated learning. While previous attempts to attack SL have often relied on overly strong assumptions or targeted easily exploitable models, we seek to develop more capable attacks. We introduce SDAR, a novel attack framework against SL with an honest-but-curious server. SDAR leverages auxiliary data and adversarial regularization to learn a decodable simulator of the client's private model, which can effectively infer the client's private features under the vanilla SL, and both features and labels under the U-shaped SL. We perform extensive experiments in both configurations to validate the effectiveness of our proposed attacks. Notably, in challenging scenarios where existing passive attacks struggle to reconstruct the client's private data effectively, SDAR consistently achieves significantly superior attack performance, even comparable to active attacks. On CIFAR-10, at the deep split level of 7, SDAR achieves private feature reconstruction with less than 0.025 mean squared error in both the vanilla and the U-shaped SL, and attains a label inference accuracy of over 98% in the U-shaped setting, while existing attacks fail to produce non-trivial results.

cs.CR↗

arXiv2023

Blink: Link Local Differential Privacy in Graph Neural Networks via Bayesian Estimation

Graph neural networks (GNNs) have gained an increasing amount of popularity due to their superior capability in learning node embeddings for various graph inference tasks, but training them can raise privacy concerns. To address this, we propose using link local differential privacy over decentralized nodes, enabling collaboration with an untrusted server to train GNNs without revealing the existence of any link. Our approach spends the privacy budget separately on links and degrees of the graph for the server to better denoise the graph topology using Bayesian estimation, alleviating the negative impact of LDP on the accuracy of the trained GNNs. We bound the mean absolute error of the inferred link probabilities against the ground truth graph topology. We then propose two variants of our LDP mechanism complementing each other in different privacy settings, one of which estimates fewer links under lower privacy budgets to avoid false positive link estimates when the uncertainty is high, while the other utilizes more information and performs better given relatively higher privacy budgets. Furthermore, we propose a hybrid variant that combines both strategies and is able to perform better across different privacy budgets. Extensive experiments show that our approach outperforms existing methods in terms of accuracy under varying privacy budgets.

cs.LG↗

arXiv2017

Van der Waals Schottky barriers as interface probes of the correlation between chemical potential shifts and charge density wave formation in 1T-TiSe$_2$ and 2H-NbSe$_2$

Layered transition metal dichalcogenide (TMD) materials, i.e. 1T-TiSe$_2$ and 2H-NbSe$_2$, harbor a second order charge density wave (CDW) transition where phonons play a key role for the periodic modulations of conduction electron densities and associated lattice distortions. We systematically study the transport and capacitance characteristics over a wide temperature range of Schottky barriers formed by intimately contacting freshly exfoliated flakes of 1T-TiSe$_2$ and 2H-NbSe$_2$ to \textit{n}-type GaAs semiconductor substrates. The extracted temperature-dependent parameters (zero-bias barrier height, ideality and built-in potential) reflect changes at the TMD/GaAs interface induced by CDW formation for both TMD materials. The measured built-in potential reveals chemical-potential shifts relating to CDW formation. With decreasing temperature a peak in the chemical-potential shifts during CDW evolution indicates a competition between electron energy re-distributions and a combination of lattice strain energies and Coulomb interactions. These modulations of chemical potential in CDW systems, such as 1T-TiSe$_2$ and 2H-NbSe$_2$ harboring second-order phase transitions, reflect a corresponding conversion from short to long-range order.

cond-mat.str-el↗

arXiv2017

Bi-2212/1T-TaS$_2$ Van der Waals junctions: Interplay of proximity induced high-$T_c$ superconductivity and CDW order

Understanding the coexistence, competition and/or cooperation between superconductivity and charge density waves (CDWs) in the transition metal dichalcogenides (TMDs) is an elusive goal which, when realized, promises to reveal fundamental information on this important class of materials. Here, we use four-terminal current-voltage measurements to study the Van der Waals interface between freshly exfoliated flakes of the high-$T_{c}$ superconductor, Bi-2212, and the CDW-dominated TMD layered material, 1T-TaS$_{2}$. For highly transparent barriers, there is a pronounced Andreev reflection feature providing evidence for proximity-induced high-$T_{c}$ superconductivity in 1T-TaS$_{2}$ with a surprisingly large energy gap ($\sim 20\thinspace $meV) equal to half that of intrinsic Bi-2212 ($\sim 40\thinspace $meV). Our systematic study using conductance spectroscopy of junctions with different transparencies also reveals the presence of two separate boson modes, each associated with a "dip-hump" structure. We infer that the proximity-induced high-$T_c$ superconductivity in the 1T-TaS$_2$ is driven by coupling to the metastable metallic phase coexisting within the Mott commensurate CDW (CCDW) phase and associated with a concomitant change of the CCDW order parameter in the interfacial region.

cond-mat.supr-con↗

arXiv2017

Detection of charge density wave phase transitions at 1T-TaS$_2$/GaAs interfaces

The transition metal dichalcogenide 1T-TaS$_2$ is well known to harbor a rich variety of charge density wave (CDW) distortions which are correlated with underlying lattice atom modulations. The long range CDW phases extend throughout the whole crystal and terminate with charge displacements at the crystal boundaries. Here we report on the transport properties and capacitance characteristics of the interface between freshly exfoliated flakes of 1T-TaS$_2$ in intimate van der Waals contact with \textit{n}-type GaAs substrates. The extracted barrier parameters (ideality, barrier height and built-in potential) experience pronounced changes across the Mott-CDW transition in the 1T-TaS$_2$. The CDW-induced changes in barrier properties are well described by a bond polarization model which upon decreasing temperature gives rise to an increased potential drop across the interfacial region due to the localization of carriers and a decreased dielectric constant.

cond-mat.mtrl-sci↗