Search arXiv⌕ Search

arXiv subjects

Ming Yin

Publications and source records attributed to Ming Yin.

At least 37 records · Page 2Linked to original sources
arXiv2026

Consistent Diffusion Language Models

Diffusion language models (DLMs) are an attractive alternative to autoregressive models because they promise sublinear-time, parallel generation, yet practical gains remain elusive as high-quality samples still demand hundreds of refinement steps. In continuous domains, consistency training along the probability-flow ODE is a popular recipe to accelerate diffusion. For discrete diffusion, no analogous sample-space ODE exists, making direct adaptation ill-defined. We argue that the right discrete substitute is the exact posterior bridge, the closed-form conditional law linking any two noise levels, which is available for broad corruptions including masked and uniform diffusion. Building on this observation, we introduce Multi-Path Discrete Consistency (MPDC), a new principle that trains a denoiser to be path-invariant in expectation across these stochastic bridges, and instantiate it as the Consistent Diffusion Language Model (CDLM), a single-stage training framework that does not require an already trained teacher model. Our CDLM objective recovers masked diffusion, continuous consistency models, and progressive or discrete distillation as analytic limits or empirical approximations of one common view. Empirically, CDLM establishes a new state of the art on both conditional and unconditional text-generation, consistently outperforming strong base discrete diffusion models and often even multi-stage distilled baselines across sampling budgets, with the largest gains in the few-step regime. Together, these results position CDLM as a principled and scalable foundation for the next generation of fast, high-fidelity discrete generative modeling.

cs.LG↗
arXiv2026

Escaping the Mode Lottery: Multi-Response Training Improves Language Model Generalization

Modern language-model fine-tuning typically pairs each prompt with a single response, even though many prompts admit multiple valid completions. This effectively reduces a multi-modal conditional distribution to a one-sample view, a phenomenon we call the "mode lottery," where training emphasizes a subset of plausible modes while leaving others underrepresented. We study multi-response training (MRT), which retains multiple responses per prompt, and develop a principled account of when and why it helps. Our key insight is that prompts and responses are distinct statistical resources: additional prompts reduce uncertainty about the input distribution, while additional responses reduce uncertainty about the conditional output distribution. This yields a variance-budget tradeoff that predicts when retaining multiple responses is worthwhile, shows diminishing returns as prompt-level uncertainty dominates, and explains why large redundant corpora can exhibit an implicit multi-response effect. We further analyze response selection, and show that Random-K-of-N is the unbiased default for distributional fine-tuning, reward-based selection can induce mode collapse, and a submodular quality-diversity objective provides an efficient alternative with theoretical guarantees. Controlled simulations validate the predicted variance and selection effects, including a striking failure mode where reward-only selection produces gradients misaligned with the true objective. Across structured and real-world datasets, including a new multi-prompt, multi-response benchmark, MRT consistently improves distributional generalization, with the largest gains in high response-diversity, low prompt-redundancy regimes. MRT reframes response multiplicity as a data-allocation problem with clear guidance: when responses are cheap and diverse, keeping more than one is not a heuristic, but a statistically grounded choice.

cs.LG↗
arXiv2026

Relative Translation Invariant Wasserstein Distance

Motivated by the Bures distance, we introduce a new family of distances, \emph{relative translation invariant Wasserstein distances}, denoted by $RW_p$, as an extension of the classical Wasserstein distances $W_p$ for $p \in [1, +\infty)$. We establish that $RW_p$ defines a valid metric and demonstrate that this type of metric is more intrinsic than the classical Wasserstein distance. A bi-level algorithm is designed to compute the general $RW_p$ distance between arbitrary discrete distributions. Moreover, when $p = 2$, we show that the optimal coupling matrix is invariant under distributional translation in the discrete setting, and we further propose two algorithms, the $\mathrm{RW}_2$-LP algorithm and the $\mathrm{RW}_2$-Sinkhorn algorithm, to improve the numerical stability of computing $W_2$ distance and the optimal coupling matrix solutions. Finally, we conduct three experiments to validate our theoretical results and algorithms. The first two experiments report that the $\mathrm{RW}_2$-LP algorithm and the $\mathrm{RW}_2$-Sinkhorn algorithm, both with and without normalization, can significantly reduce the numerical errors compared to standard algorithms. The third experiment shows that $RW_p$ algorithms are computationally scalable and applicable to the retrieval of similar thunderstorm patterns in practical applications.

cs.LG↗
arXiv2026

LiveMCP-101: Stress Testing and Diagnosing MCP-enabled Agents on Challenging Queries

Tool calling has emerged as a critical capability for AI agents. In contrast to conventional tool calling frameworks that rely on static, provider-specific tool definitions, the Model Context Protocol (MCP) offers a unified interface to discover and invoke tools dynamically. However, there is a significant gap in benchmarking multi-step tasks using diverse MCP tools in realistic, dynamic scenarios. In this work, we present LiveMCP-101, a benchmark of 101 real-world queries that require coordinated use of multiple MCP tools. To address temporal variability in real-world tool responses, we introduce a parallel evaluation framework where a reference agent executes a validated plan simultaneously to produce real-time reference outputs. Experiments show that even frontier LLMs achieve a success rate below 60\%, highlighting challenges in multi-step tool use. Comprehensive error analysis identifies seven failure modes spanning tool planning, parameterization, and output handling, pointing to concrete directions for improving current models. LiveMCP-101 sets a rigorous standard for evaluating real-world agent capabilities, advancing toward autonomous agent systems that reliably execute complex tasks through MCP tool orchestration.

cs.CL↗
arXiv2026

Qumus: Realization of An Embodied AI Quantum Material Experimentalist

While modern Large Language Models (LLMs) and agentic artificial intelligence (AI) have demonstrated transformative capabilities in digital domains, the realization of embodied AI capable of real-world scientific discovery remains a difficult frontier. The advancements are hindered by the inherent complexity of integrating high-level reasoning, multimodal information processing and real-time physical execution. Here we introduce Qumus, the first AI quantum materials experimentalist. Physically embodied within a robotic mini-laboratory, Qumus is an intelligent, multimodal, and multi-agent system designed for the creation and nano-processing of atomically thin two-dimensional (2D) materials and stacked van der Waals (vdW) structures. Qumus autonomously navigates the full scientific cycle, from hypothesis generation and protocol planning to multi-step experimental execution, result analysis and reporting, acting as an experimentalist. Markedly, the system has achieved, for the first time, the AI-creation of graphene, as well as the first AI-fabrication of complex nanodevices including atomically thin field-effect transistors via vdW stacking. Qumus excels at these tasks by demonstrating autonomous error correction and closed-loop experimentation. Our results establish a generalizable framework for self-improving embodied AI systems that learn directly from the quantum world, opening a pathway toward accelerated discovery in quantum materials, electronics and beyond.

cond-mat.mes-hall↗
arXiv2026

From Fallback to Frontline: When Can LLMs be Superior Annotators of Human Perspectives?

Although large language models (LLMs) are increasingly used as annotators at scale, they are typically treated as a pragmatic fallback rather than a faithful estimator of human perspectives. This work challenges that presumption. By framing perspective-taking as the estimation of a latent group-level judgment, we characterize the conditions under which modern LLMs can outperform human annotators, including in-group humans, when predicting aggregate subgroup opinions on subjective tasks, and show that these conditions are common in practice. This advantage arises from structural properties of LLMs as estimators, including low variance and reduced coupling between representation and processing biases, rather than any claim of lived experience. Our analysis identifies clear regimes where LLMs act as statistically superior frontline estimators, as well as principled limits where human judgment remains essential. These findings reposition LLMs from a cost-saving compromise to a principled tool for estimating collective human perspectives.

cs.AI↗
arXiv2026

NumCoKE: Ordinal-Aware Numerical Reasoning over Knowledge Graphs with Mixture-of-Experts and Contrastive Learning

Knowledge graphs (KGs) serve as a vital backbone for a wide range of AI applications, including natural language understanding and recommendation. A promising yet underexplored direction is numerical reasoning over KGs, which involves inferring new facts by leveraging not only symbolic triples but also numerical attribute values (e.g., length, weight). However, existing methods fall short in two key aspects: (1) Incomplete semantic integration: Most models struggle to jointly encode entities, relations, and numerical attributes in a unified representation space, limiting their ability to extract relation-aware semantics from numeric information. (2) Ordinal indistinguishability: Due to subtle differences between close values and sampling imbalance, models often fail to capture fine-grained ordinal relationships (e.g., longer, heavier), especially in the presence of hard negatives. To address these challenges, we propose NumCoKE, a numerical reasoning framework for KGs based on Mixture-of-Experts and Ordinal Contrastive Embedding. To overcome (C1), we introduce a Mixture-of-Experts Knowledge-Aware (MoEKA) encoder that jointly aligns symbolic and numeric components into a shared semantic space, while dynamically routing attribute features to relation-specific experts. To handle (C2), we propose Ordinal Knowledge Contrastive Learning (OKCL), which constructs ordinal-aware positive and negative samples using prior knowledge, enabling the model to better discriminate subtle semantic shifts. Extensive experiments on three public KG benchmarks demonstrate that NumCoKE consistently outperforms competitive baselines across diverse attribute distributions, validating its superiority in both semantic integration and ordinal reasoning.

cs.AI↗
arXiv2026

OilSAM2: Memory-Augmented SAM2 for Scalable SAR Oil Spill Detection

Segmenting oil spills from Synthetic Aperture Radar (SAR) imagery remains challenging due to severe appearance variability, scale heterogeneity, and the absence of temporal continuity in real world monitoring scenarios. While foundation models such as Segment Anything (SAM) enable prompt driven segmentation, existing SAM based approaches operate on single images and cannot effectively reuse information across scenes. Memory augmented variants (e.g., SAM2) further assume temporal coherence, making them prone to semantic drift when applied to unordered SAR image collections. We propose OilSAM2, a memory augmented segmentation framework tailored for unordered SAR oil spill monitoring. OilSAM2 introduces a hierarchical feature aware multi scale memory bank that explicitly models texture, structure, and semantic level representations, enabling robust cross image information reuse. To mitigate memory drift, we further propose a structure semantic consistent memory update strategy that selectively refreshes memory based on semantic discrepancy and structural variation.Experiments on two public SAR oil spill datasets demonstrate that OilSAM2 achieves state of the art segmentation performance, delivering stable and accurate results under noisy SAR monitoring scenarios. The source code is available at https://github.com/Chenshuaiyu1120/OILSAM2.

cs.CV↗
arXiv2026

XAgen: An Explainability Tool for Identifying and Correcting Failures in Multi-Agent Workflows

As multi-agent systems powered by Large Language Models (LLMs) are increasingly adopted in real-world workflows, users with diverse technical backgrounds are now building and refining their own agentic processes. However, these systems can fail in opaque ways, making it difficult for users to observe, understand, and correct errors. We conducted formative interviews with 12 practitioners to identify mismatches between existing debugging tools and users' needs. Based on these insights, we designed XAgen, an explainability tool that supports users with varying AI expertise through three core capabilities: log visualization for glanceable workflow understanding, human-in-the-loop feedback to capture expert judgment, and automatic error detection via an LLM-as-a-judge. In a user study with 8 participants, XAgen helped users locate failures more easily, attribute to specific agents or steps, and iteratively improve configurations. Our findings surface human-centered design guidelines for explainable agentic AI development and highlight opportunities for more context-aware interactive debugging.

cs.HC↗
arXiv2026

MMTok: Multimodal Coverage Maximization for Efficient Inference of VLMs

Vision-Language Models (VLMs) demonstrate impressive performance in understanding visual content with language instruction by converting visual inputs to vision tokens. However, redundancy in vision tokens results in the degraded inference efficiency of VLMs. While many algorithms have been proposed to reduce the number of vision tokens, most of them apply only unimodal information (i.e., vision/text) for pruning and ignore the inherent multimodal property of vision-language tasks. Moreover, it lacks a generic criterion that can be applied to different modalities. To mitigate this limitation, in this work, we propose to leverage both vision and text tokens to select informative vision tokens by the coverage criterion. We first formulate the subset selection problem as a maximum coverage problem. Afterwards, a subset of vision tokens is optimized to cover the text tokens and the original set of vision tokens, simultaneously. The proposed method MMTok is extensively evaluated on benchmark datasets with different VLMs. The comparison illustrates that vision and text information are complementary, and combining multimodal information can surpass the unimodal baseline with a clear margin. Moreover, under the maximum coverage criterion on the POPE dataset, our method achieves a 1.87x speedup while maintaining 98.7% of the original performance on LLaVA-NeXT-13B. Finally, with only four vision tokens, 87.7% of the original performance is still preserved on LLaVA-1.5-7B. These results highlight the effectiveness of coverage in token selection. The code is available at https://github.com/Ironieser/mmtok

cs.CV↗
arXiv2026

Understanding the Effects of AI-Assisted Critical Thinking on Human-AI Decision Making

Despite the growing prevalence of human-AI decision making, the human-AI team's decision performance often remains suboptimal, partially due to insufficient examination of humans' own reasoning. In this paper, we explore designing AI systems that directly analyze humans' decision rationales and encourage critical reflection of their own decisions. We introduce the AI-Assisted Critical Thinking (AACT) framework, which leverages a domain-specific AI model's counterfactual analysis of human decision to help decision-makers identify potential flaws in their decision argument and support the correction of them. Through a case study on house price prediction, we find that AACT outperforms traditional AI-based decision-support in reducing over-reliance on AI, though also triggering higher cognitive load. Subgroup analysis reveals AACT can be particularly beneficial for some decision-makers such as those very familiar with AI technologies. We conclude by discussing the practical implications of our findings, use cases and design choices of AACT, and considerations for using AI to facilitate critical thinking.

cs.HC↗
arXiv2026

Align When They Want, Complement When They Need! Human-Centered Ensembles for Adaptive Human-AI Collaboration

In human-AI decision making, designing AI that complements human expertise has been a natural strategy to enhance human-AI collaboration, yet it often comes at the cost of decreased AI performance in areas of human strengths. This can inadvertently erode human trust and cause them to ignore AI advice precisely when it is most needed. Conversely, an aligned AI fosters trust yet risks reinforcing suboptimal human behavior and lowering human-AI team performance. In this paper, we start by identifying this fundamental tension between performance-boosting (i.e., complementarity) and trust-building (i.e., alignment) as an inherent limitation of the traditional approach for training a single AI model to assist human decision making. To overcome this, we introduce a novel human-centered adaptive AI ensemble that strategically toggles between two specialist AI models - the aligned model and the complementary model - based on contextual cues, using an elegantly simple yet provably near-optimal Rational Routing Shortcut mechanism. Comprehensive theoretical analyses elucidate why the adaptive AI ensemble is effective and when it yields maximum benefits. Moreover, experiments on both simulated and real-world data show that when humans are assisted by the adaptive AI ensemble in decision making, they can achieve significantly higher performance than when they are assisted by single AI models that are trained to either optimize for their independent performance or even the human-AI team performance.

cs.AI↗
arXiv2026

Human-LLM Collaborative Feature Engineering for Tabular Data

Large language models (LLMs) are increasingly used to automate feature engineering in tabular learning. Given task-specific information, LLMs can propose diverse feature transformation operations to enhance downstream model performance. However, current approaches typically assign the LLM as a black-box optimizer, responsible for both proposing and selecting operations based solely on its internal heuristics, which often lack calibrated estimations of operation utility and consequently lead to repeated exploration of low-yield operations without a principled strategy for prioritizing promising directions. In this paper, we propose a human-LLM collaborative feature engineering framework for tabular learning. We begin by decoupling the transformation operation proposal and selection processes, where LLMs are used solely to generate operation candidates, while the selection is guided by explicitly modeling the utility and uncertainty of each proposed operation. Since accurate utility estimation can be difficult especially in the early rounds of feature engineering, we design a mechanism within the framework that selectively elicits and incorporates human expert preference feedback, comparing which operations are more promising, into the selection process to help identify more effective operations. Our evaluations on both the synthetic study and the real user study demonstrate that the proposed framework improves feature engineering performance across a variety of tabular datasets and reduces users' cognitive load during the feature engineering process.

cs.LG↗
arXiv2025

Assessing Automated Fact-Checking for Medical LLM Responses with Knowledge Graphs

The recent proliferation of large language models (LLMs) holds the potential to revolutionize healthcare, with strong capabilities in diverse medical tasks. Yet, deploying LLMs in high-stakes healthcare settings requires rigorous verification and validation to understand any potential harm. This paper investigates the reliability and viability of using medical knowledge graphs (KGs) for the automated factuality evaluation of LLM-generated responses. To ground this investigation, we introduce FAITH, a framework designed to systematically probe the strengths and limitations of this KG-based approach. FAITH operates without reference answers by decomposing responses into atomic claims, linking them to a medical KG, and scoring them based on evidence paths. Experiments on diverse medical tasks with human subjective evaluations demonstrate that KG-grounded evaluation achieves considerably higher correlations with clinician judgments and can effectively distinguish LLMs with varying capabilities. It is also robust to textual variances. The inherent explainability of its scoring can further help users understand and mitigate the limitations of current LLMs. We conclude that while limitations exist, leveraging KGs is a prominent direction for automated factuality assessment in healthcare.

cs.LG↗
arXiv2025

CRISPR-GPT for Agentic Automation of Gene-editing Experiments

The introduction of genome engineering technology has transformed biomedical research, making it possible to make precise changes to genetic information. However, creating an efficient gene-editing system requires a deep understanding of CRISPR technology, and the complex experimental systems under investigation. While Large Language Models (LLMs) have shown promise in various tasks, they often lack specific knowledge and struggle to accurately solve biological design problems. In this work, we introduce CRISPR-GPT, an LLM agent augmented with domain knowledge and external tools to automate and enhance the design process of CRISPR-based gene-editing experiments. CRISPR-GPT leverages the reasoning ability of LLMs to facilitate the process of selecting CRISPR systems, designing guide RNAs, recommending cellular delivery methods, drafting protocols, and designing validation experiments to confirm editing outcomes. We showcase the potential of CRISPR-GPT for assisting non-expert researchers with gene-editing experiments from scratch and validate the agent's effectiveness in a real-world use case. Furthermore, we explore the ethical and regulatory considerations associated with automated gene-editing design, highlighting the need for responsible and transparent use of these tools. Our work aims to bridge the gap between beginner biological researchers and CRISPR genome engineering techniques, and demonstrate the potential of LLM agents in facilitating complex biological discovery tasks. The published version of this draft is available at https://www.nature.com/articles/s41551-025-01463-z.

cs.AI↗
arXiv2025

A Classification Program for Nonlocality Paradoxes of Three Qubits

Nonlocality is a quintessential signature of nonclassical behaviour and a resource for quantum advantages in communication and computation. The paradoxical correlations witnessed by strong nonlocality undergird the standard probabilistic form of nonlocality and provide optimal advantages in numerous informational tasks. Three-qubit systems are the simplest ones that admit strong nonlocality. Abramsky et al. (TQC, 2017) established the existence of an infinite family of three-qubit paradoxes, beyond the well-known GHZ paradox, which exhibited a novel conditional structure. In this work, we introduce several new infinite families of three-qubit paradoxes and articulate a detailed roadmap towards the complete classification of all three-qubit nonlocality paradoxes. In particular, we prove that our paradoxes exhaust all those satisfying reasonable regularity conditions. We give an example of a highly exotic paradox and place constraints on the search for new exotic paradoxes. We conjecture that all paradoxes must involve states from a one-parameter family and provide significant evidence in support of this conjecture.

quant-ph↗
arXiv2025

Memory-Augmented SAM2 for Training-Free Surgical Video Segmentation

Surgical video segmentation is a critical task in computer-assisted surgery, essential for enhancing surgical quality and patient outcomes. Recently, the Segment Anything Model 2 (SAM2) framework has demonstrated remarkable advancements in both image and video segmentation. However, the inherent limitations of SAM2's greedy selection memory design are amplified by the unique properties of surgical videos-rapid instrument movement, frequent occlusion, and complex instrument-tissue interaction-resulting in diminished performance in the segmentation of complex, long videos. To address these challenges, we introduce Memory Augmented (MA)-SAM2, a training-free video object segmentation strategy, featuring novel context-aware and occlusion-resilient memory models. MA-SAM2 exhibits strong robustness against occlusions and interactions arising from complex instrument movements while maintaining accuracy in segmenting objects throughout videos. Employing a multi-target, single-loop, one-prompt inference further enhances the efficiency of the tracking process in multi-instrument videos. Without introducing any additional parameters or requiring further training, MA-SAM2 achieved performance improvements of 4.36% and 6.1% over SAM2 on the EndoVis2017 and EndoVis2018 datasets, respectively, demonstrating its potential for practical surgical applications.

cs.CV↗
arXiv2025

Which Agent Causes Task Failures and When? On Automated Failure Attribution of LLM Multi-Agent Systems

Failure attribution in LLM multi-agent systems-identifying the agent and step responsible for task failures-provides crucial clues for systems debugging but remains underexplored and labor-intensive. In this paper, we propose and formulate a new research area: automated failure attribution for LLM multi-agent systems. To support this initiative, we introduce the Who&When dataset, comprising extensive failure logs from 127 LLM multi-agent systems with fine-grained annotations linking failures to specific agents and decisive error steps. Using the Who&When, we develop and evaluate three automated failure attribution methods, summarizing their corresponding pros and cons. The best method achieves 53.5% accuracy in identifying failure-responsible agents but only 14.2% in pinpointing failure steps, with some methods performing below random. Even SOTA reasoning models, such as OpenAI o1 and DeepSeek R1, fail to achieve practical usability. These results highlight the task's complexity and the need for further research in this area. Code and dataset are available at https://github.com/mingyin1/Agents_Failure_Attribution

cs.MA↗