arXiv subjects

Yi Cheng

Publications and source records attributed to Yi Cheng.

At least 37 records · Page 2Linked to original sources

arXiv2025

Learning from the Best, Differently: A Diversity-Driven Rethinking on Data Selection

High-quality pre-training data is crutial for large language models, where quality captures factual reliability and semantic value, and diversity ensures broad coverage and distributional heterogeneity. Existing approaches typically rely on single or multiple-dimensional score-based selection. However, directly selecting top-scored data often degrades performance, and sampling from a broader range is required to recover results. The above non-monotonicity between dataset scores and downstream benchmark results reveals a fundamental bias: score-based methods collapse correlated dimensions, causing top-scored data to appear high-quality while systematically overlooking diversity. We argue that ensuring diversity requires decomposing correlated metrics into orthogonal feature dimensions, from which the top-scored data can be directly selected. Therefore, we proposed the Orthogonal Diversity-Aware Selection (ODiS) algorithm, which preserves both quality and diversity during data selection. First, ODiS evaluates data from multiple dimensions, covering language quality, knowledge quality, and comprehension difficulty. The multi-dimensional scores are then decorrelated via Principal Component Analysis (PCA), yielding orthogonal evaluation dimensions. For each dimension, a Roberta-based scorer is trained to regress the data onto PCA-projected scores, enabling scalable inference on large corpora. Finally, ODiS constructs the training dataset by selecting top-scored data within each orthogonal dimension, thereby ensuring both quality and diversity. Empirical results show that ODiS-selected data exhibit less than 2\% inter-dimension overlap, confirming orthogonality between dimensions. More importantly, models trained with ODiS-selected data significantly outperform other baselines on downstream benchmarks, highlighting the necessity of orthogonal, diversity-aware data selection for LLMs.

cs.CL↗

arXiv2025

RAR$^2$: Retrieval-Augmented Medical Reasoning via Thought-Driven Retrieval

Large Language Models (LLMs) have shown promising performance on diverse medical benchmarks, highlighting their potential in supporting real-world clinical tasks. Retrieval-Augmented Generation (RAG) has emerged as a key approach for mitigating knowledge gaps and hallucinations by incorporating external medical information. However, RAG still struggles with complex medical questions that require intensive reasoning, as surface-level input often fails to reflect the true knowledge needs of the task. Existing methods typically focus on refining queries without explicitly modeling the reasoning process, limiting their ability to retrieve and integrate clinically relevant knowledge. In this work, we propose RAR$^2$, a joint learning framework that improves both Reasoning-Augmented Retrieval and Retrieval-Augmented Reasoning. RAR$^2$ constructs a thought process to uncover implicit knowledge requirements and uses it to guide retrieval and answer generation. We build a training dataset of mixed preference pairs and apply Direct Preference Optimization (DPO) to train the model. Moreover, we design two test-time scaling strategies to explore the boundaries of our framework. Experiments demonstrate the effectiveness of RAR$^2$ across several biomedical question answering datasets, outperforming RAG baselines with or without fine-tuning.

cs.CL↗

arXiv2025

CushionCatch: A Compliant Catching Mechanism for Mobile Manipulators via Combined Optimization and Learning

Catching flying objects with a cushioning process is a skill commonly performed by humans, yet it remains a significant challenge for robots. In this paper, we present a framework that combines optimization and learning to achieve compliant catching on mobile manipulators (CCMM). First, we propose a high-level capture planner for mobile manipulators (MM) that calculates the optimal capture point and joint configuration. Next, the pre-catching (PRC) planner ensures the robot reaches the target joint configuration as quickly as possible. To learn compliant catching strategies, we propose a network that leverages the strengths of LSTM for capturing temporal dependencies and positional encoding for spatial context (P-LSTM). This network is designed to effectively learn compliant strategies from human demonstrations. Following this, the post-catching (POC) planner tracks the compliant sequence output by the P-LSTM while avoiding potential collisions due to structural differences between humans and robots. We validate the CCMM framework through both simulated and real-world ball-catching scenarios, achieving a success rate of 98.70% in simulation, 92.59% in real-world tests, and a 28.7% reduction in impact torques. The open source code has be released for the reference of the community.

cs.RO↗

arXiv2025

SSPO: Self-traced Step-wise Preference Optimization for Process Supervision and Reasoning Compression

Test-time scaling has proven effective in further enhancing the performance of pretrained Large Language Models (LLMs). However, mainstream post-training methods (i.e., reinforcement learning (RL) with chain-of-thought (CoT) reasoning) often incur substantial computational overhead due to auxiliary models and overthinking. In this paper, we empirically reveal that the incorrect answers partially stem from verbose reasoning processes lacking correct self-fix, where errors accumulate across multiple reasoning steps. To this end, we propose Self-traced Step-wise Preference Optimization (SSPO), a pluggable RL process supervision framework that enables fine-grained optimization of each reasoning step. Specifically, SSPO requires neither auxiliary models nor stepwise manual annotations. Instead, it leverages step-wise preference signals generated by the model itself to guide the optimization process for reasoning compression. Experiments demonstrate that the generated reasoning sequences from SSPO are both accurate and succinct, effectively mitigating overthinking behaviors without compromising model performance across diverse domains and languages.

cs.LG↗

arXiv2025

Training oscillator Ising machines to assign the dynamic stability of their equilibrium points

We propose a neural network model, which, with appropriate assignment of the stability of its equilibrium points (EPs), achieves Hopfield-like associative memory. The oscillator Ising machine (OIM) is an ideal candidates for such a model, as all its $0/π$ binary EPs are structurally stable with their dynamic stability tunable by the coupling weights. Traditional Hopfield-based models store the desired patterns by designing the coupling weights between neurons. The design of coupling weights should simultaneously take into account both the existence and the dynamic stability of the EPs for the storage of the desired patterns. For OIMs, since all $0/π$ binary EPs are structurally stable, the design of the coupling weights needs only to focus on assigning appropriate stability for the $0/π$ binary EPs according to the desired patterns. In this paper, we establish a connection between the stability and the Hamiltonian energy of EPs for OIMs, and, based on this connection, provide a Hamiltonian-Regularized Eigenvalue Contrastive Method (HRECM) to train the coupling weights of OIMs for assigning appropriate stability to their EPs. Finally, numerical experiments are performed to validate the effectiveness of the proposed method.

cs.NE↗

arXiv2025

Optical trapping based on microring resonators with transverse slot structure

Over the past few decades, optical manipulation has emerged as a highly successful tool in various fields, such as biology, micro/nanorobotics, and physics. Among the different techniques, the transverse slot optical waveguide has shown remarkable potential in enhancing the field and significantly improving optical trapping capabilities. Additionally, microring resonators have demonstrated the ability to enhance the field at specific resonance wavelengths, enabling the manipulation and capture of particles. In this study, we investigated the impact of the structure on nanoparticle capture by introducing a 50 nm transverse slot in a 5 μm microring resonator. Through the integration of a transverse slot in the microring resonator, we observed a substantial increase in the maximum bound optical power for a nanosphere with a refractive index of 1.6 and a diameter of 50 nm, reaching 3988.8 pN/W. This value is 2292 times higher than the maximum optical force in a straight waveguide and 2.266 times higher than the maximum optical force in a microring resonator. The proposed structure significantly enhances the optical trapping capabilities for nanoscale particles, thus paving the way for the development of advanced micro/nanomanipulation techniques.

physics.optics↗

arXiv2025

Towards Dynamic Theory of Mind: Evaluating LLM Adaptation to Temporal Evolution of Human States

As Large Language Models (LLMs) increasingly participate in human-AI interactions, evaluating their Theory of Mind (ToM) capabilities - particularly their ability to track dynamic mental states - becomes crucial. While existing benchmarks assess basic ToM abilities, they predominantly focus on static snapshots of mental states, overlooking the temporal evolution that characterizes real-world social interactions. We present \textsc{DynToM}, a novel benchmark specifically designed to evaluate LLMs' ability to understand and track the temporal progression of mental states across interconnected scenarios. Through a systematic four-step framework, we generate 1,100 social contexts encompassing 5,500 scenarios and 78,100 questions, each validated for realism and quality. Our comprehensive evaluation of ten state-of-the-art LLMs reveals that their average performance underperforms humans by 44.7\%, with performance degrading significantly when tracking and reasoning about the shift of mental states. This performance gap highlights fundamental limitations in current LLMs' ability to model the dynamic nature of human mental states.

cs.CL↗

arXiv2025

AutoPal: Autonomous Adaptation to Users for Personal AI Companionship

Previous research has demonstrated the potential of AI agents to act as companions that can provide constant emotional support for humans. In this paper, we emphasize the necessity of autonomous adaptation in personal AI companionship, an underexplored yet promising direction. Such adaptability is crucial as it can facilitate more tailored interactions with users and allow the agent to evolve in response to users' changing needs. However, imbuing agents with autonomous adaptability presents unique challenges, including identifying optimal adaptations to meet users' expectations and ensuring a smooth transition during the adaptation process. To address them, we devise a hierarchical framework, AutoPal, that enables controllable and authentic adjustments to the agent's persona based on user interactions. A personamatching dataset is constructed to facilitate the learning of optimal persona adaptations. Extensive experiments demonstrate the effectiveness of AutoPal and highlight the importance of autonomous adaptability in AI companionship.

cs.CL↗

arXiv2025

RADAR: Enhancing Radiology Report Generation with Supplementary Knowledge Injection

Large language models (LLMs) have demonstrated remarkable capabilities in various domains, including radiology report generation. Previous approaches have attempted to utilize multimodal LLMs for this task, enhancing their performance through the integration of domain-specific knowledge retrieval. However, these approaches often overlook the knowledge already embedded within the LLMs, leading to redundant information integration. To address this limitation, we propose Radar, a framework for enhancing radiology report generation with supplementary knowledge injection. Radar improves report generation by systematically leveraging both the internal knowledge of an LLM and externally retrieved information. Specifically, it first extracts the model's acquired knowledge that aligns with expert image-based classification outputs. It then retrieves relevant supplementary knowledge to further enrich this information. Finally, by aggregating both sources, Radar generates more accurate and informative radiology reports. Extensive experiments on MIMIC-CXR, CheXpert-Plus, and IU X-ray demonstrate that our model outperforms state-of-the-art LLMs in both language quality and clinical accuracy.

cs.CV↗

arXiv2025

Subtle Errors in Reasoning: Preference Learning via Error-injected Self-editing

Large Language Models (LLMs) have exhibited strong mathematical reasoning prowess, tackling tasks ranging from basic arithmetic to advanced competition-level problems. However, frequently occurring subtle yet critical errors, such as miscalculations or incorrect substitutions, limit the LLMs' full potential. Existing studies to improve mathematical ability typically involve applying preference learning to step-wise solution pairs. Although these methods leverage samples of varying granularity to mitigate reasoning errors, they overlook critical subtle errors. In this work, we propose a novel preference learning framework called eRror-Injected Self-Editing (RISE), which injects predefined subtle errors into pivotal tokens in reasoning or computation steps to construct hard pairs for error mitigation. In detail, RISE uses the LLM itself to edit a small number of tokens in the solution, injecting designed subtle errors. Then, pairs composed of self-edited solutions and their corresponding correct ones, along with pairs of correct and incorrect solutions obtained through sampling, are used together for subtle error-aware DPO training. Compared with other preference learning methods, RISE further refines the training objective without requiring fine-grained sampling or preference annotation. Extensive experiments validate the effectiveness of RISE, with preference learning on Qwen2-7B-Instruct yielding notable improvements of 3.0% on GSM8K and 7.9% on MATH with only 4.5K training samples. Moreover, the effect of error mitigation extends from mathematical reasoning to logical reasoning and code generation.

cs.CL↗

arXiv2025

Learning to Align Multi-Faceted Evaluation: A Unified and Robust Framework

Large Language Models (LLMs) are being used more and more extensively for automated evaluation in various scenarios. Previous studies have attempted to fine-tune open-source LLMs to replicate the evaluation explanations and judgments of powerful proprietary models, such as GPT-4. However, these methods are largely limited to text-based analyses under predefined general criteria, resulting in reduced adaptability for unseen instructions and demonstrating instability in evaluating adherence to quantitative and structural constraints. To address these limitations, we propose a novel evaluation framework, ARJudge, that adaptively formulates evaluation criteria and synthesizes both text-based and code-driven analyses to evaluate LLM responses. ARJudge consists of two components: a fine-tuned Analyzer that generates multi-faceted evaluation analyses and a tuning-free Refiner that combines and refines all analyses to make the final judgment. We construct a Composite Analysis Corpus that integrates tasks for evaluation criteria generation alongside text-based and code-driven analysis generation to train the Analyzer. Our results demonstrate that ARJudge outperforms existing fine-tuned evaluators in effectiveness and robustness. Furthermore, it demonstrates the importance of multi-faceted evaluation and code-driven analyses in enhancing evaluation capabilities.

cs.CL↗

arXiv2025

Symmetrical bipolar electrobending deformation in acceptor-doped piezoceramics

Since 2022, large apparent strains (>1%) with highly asymmetrical strain-electric field (S-E) curves have been reported in various thin piezoceramic materials, attributed to a bidirectional electric-field-induced bending (electrobending) deformation, which consistently produces convex bending along the negative electric field direction. In this study, we report a novel unidirectional electrobending behavior in acceptor-doped K0.5Na0.5NbO3 ceramics, where convex bending always occurs along the pre-poling direction regardless of the direction of the applied electric field. This unique deformation is related to the reorientation of the defect dipoles in one surface layer during the pre-poling process, resulting in an asymmetrical distribution of defect dipoles in the two surface layers. The synergistic interaction between ferroelectric domains and defect dipoles in the surface layers induces this unidirectional electrobending, as evidenced by a butterfly-like symmetrical bipolar S-E curve with a giant apparent strain of 3.2%. These findings provide new insights into defect engineering strategies for developing advanced piezoelectric materials with large electroinduced displacements.

cond-mat.mtrl-sci↗

arXiv2025

Sigma: Differential Rescaling of Query, Key and Value for Efficient Language Models

We introduce Sigma, an efficient large language model specialized for the system domain, empowered by a novel architecture including DiffQKV attention, and pre-trained on our meticulously collected system domain data. DiffQKV attention significantly enhances the inference efficiency of Sigma by optimizing the Query (Q), Key (K), and Value (V) components in the attention mechanism differentially, based on their varying impacts on the model performance and efficiency indicators. Specifically, we (1) conduct extensive experiments that demonstrate the model's varying sensitivity to the compression of K and V components, leading to the development of differentially compressed KV, and (2) propose augmented Q to expand the Q head dimension, which enhances the model's representation capacity with minimal impacts on the inference speed. Rigorous theoretical and empirical analyses reveal that DiffQKV attention significantly enhances efficiency, achieving up to a 33.36% improvement in inference speed over the conventional grouped-query attention (GQA) in long-context scenarios. We pre-train Sigma on 6T tokens from various sources, including 19.5B system domain data that we carefully collect and 1T tokens of synthesized and rewritten data. In general domains, Sigma achieves comparable performance to other state-of-arts models. In the system domain, we introduce the first comprehensive benchmark AIMicius, where Sigma demonstrates remarkable performance across all tasks, significantly outperforming GPT-4 with an absolute improvement up to 52.5%.

cs.CL↗

arXiv2025

Training an LLM-as-a-Judge Model: Pipeline, Insights, and Practical Lessons

The rapid advancement of large language models (LLMs) has opened new possibilities for their adoption as evaluative judges. This paper introduces Themis, a fine-tuned LLM judge that delivers sophisticated context-aware evaluations. We provide a comprehensive overview of the development pipeline for Themis, highlighting its scenario-dependent evaluation prompts and two novel methods for controlled instruction generation. These designs enable Themis to effectively distill evaluative skills from teacher models, while retaining flexibility for continuous development. We introduce two human-labeled benchmarks for meta-evaluation, demonstrating that Themis can achieve high alignment with human preferences in an economical manner. Additionally, we explore insights into the LLM-as-a-judge paradigm, revealing nuances in performance and the varied effects of reference answers. Notably, we observe that pure knowledge distillation from strong LLMs, though common, does not guarantee performance improvement through scaling. We propose a mitigation strategy based on instruction-following difficulty. Furthermore, we provide practical guidelines covering data balancing, prompt customization, multi-objective training, and metric aggregation. We aim for our method and findings, along with the fine-tuning data, benchmarks, and model checkpoints, to support future research and development in this area.

cs.CL↗

arXiv2025

Integrative Decoding: Improve Factuality via Implicit Self-consistency

Self-consistency-based approaches, which involve repeatedly sampling multiple outputs and selecting the most consistent one as the final response, prove to be remarkably effective in improving the factual accuracy of large language models. Nonetheless, existing methods usually have strict constraints on the task format, largely limiting their applicability. In this paper, we present Integrative Decoding (ID), to unlock the potential of self-consistency in open-ended generation tasks. ID operates by constructing a set of inputs, each prepended with a previously sampled response, and then processes them concurrently, with the next token being selected by aggregating of all their corresponding predictions at each decoding step. In essence, this simple approach implicitly incorporates self-consistency in the decoding objective. Extensive evaluation shows that ID consistently enhances factuality over a wide range of language models, with substantial improvements on the TruthfulQA (+11.2%), Biographies (+15.4%) and LongFact (+8.5%) benchmarks. The performance gains amplify progressively as the number of sampled responses increases, indicating the potential of ID to scale up with repeated sampling.

cs.CL↗

arXiv2024

Memory-Augmented Multimodal LLMs for Surgical VQA via Self-Contained Inquiry

Comprehensively understanding surgical scenes in Surgical Visual Question Answering (Surgical VQA) requires reasoning over multiple objects. Previous approaches address this task using cross-modal fusion strategies to enhance reasoning ability. However, these methods often struggle with limited scene understanding and question comprehension, and some rely on external resources (e.g., pre-extracted object features), which can introduce errors and generalize poorly across diverse surgical environments. To address these challenges, we propose SCAN, a simple yet effective memory-augmented framework that leverages Multimodal LLMs to improve surgical context comprehension via Self-Contained Inquiry. SCAN operates autonomously, generating two types of memory for context augmentation: Direct Memory (DM), which provides multiple candidates (or hints) to the final answer, and Indirect Memory (IM), which consists of self-contained question-hint pairs to capture broader scene context. DM directly assists in answering the question, while IM enhances understanding of the surgical scene beyond the immediate query. Reasoning over these object-aware memories enables the model to accurately interpret images and respond to questions. Extensive experiments on three publicly available Surgical VQA datasets demonstrate that SCAN achieves state-of-the-art performance, offering improved accuracy and robustness across various surgical scenarios.

cs.CV↗

arXiv2024

ICON: Improving Inter-Report Consistency in Radiology Report Generation via Lesion-aware Mixup Augmentation

Previous research on radiology report generation has made significant progress in terms of increasing the clinical accuracy of generated reports. In this paper, we emphasize another crucial quality that it should possess, i.e., inter-report consistency, which refers to the capability of generating consistent reports for semantically equivalent radiographs. This quality is even of greater significance than the overall report accuracy in terms of ensuring the system's credibility, as a system prone to providing conflicting results would severely erode users' trust. Regrettably, existing approaches struggle to maintain inter-report consistency, exhibiting biases towards common patterns and susceptibility to lesion variants. To address this issue, we propose ICON, which improves the inter-report consistency of radiology report generation. Aiming to enhance the system's ability to capture similarities in semantically equivalent lesions, our approach first involves extracting lesions from input images and examining their characteristics. Then, we introduce a lesion-aware mixup technique to ensure that the representations of the semantically equivalent lesions align with the same attributes, achieved through a linear combination during the training phase. Extensive experiments on three publicly available chest X-ray datasets verify the effectiveness of our approach, both in terms of improving the consistency and accuracy of the generated reports.

cs.CV↗

arXiv2024

MBC: Multi-Brain Collaborative Control for Quadruped Robots

In the field of locomotion task of quadruped robots, Blind Policy and Perceptive Policy each have their own advantages and limitations. The Blind Policy relies on preset sensor information and algorithms, suitable for known and structured environments, but it lacks adaptability in complex or unknown environments. The Perceptive Policy uses visual sensors to obtain detailed environmental information, allowing it to adapt to complex terrains, but its effectiveness is limited under occluded conditions, especially when perception fails. Unlike the Blind Policy, the Perceptive Policy is not as robust under these conditions. To address these challenges, we propose a MBC:Multi-Brain collaborative system that incorporates the concepts of Multi-Agent Reinforcement Learning and introduces collaboration between the Blind Policy and the Perceptive Policy. By applying this multi-policy collaborative model to a quadruped robot, the robot can maintain stable locomotion even when the perceptual system is impaired or observational data is incomplete. Our simulations and real-world experiments demonstrate that this system significantly improves the robot's passability and robustness against perception failures in complex environments, validating the effectiveness of multi-policy collaboration in enhancing robotic motion performance.

cs.RO↗