arXiv subjects

Yucheng Wang

Publications and source records attributed to Yucheng Wang.

At least 55 records · Page 3Linked to original sources

arXiv2026

Attention Residuals

Residual connections with PreNorm are standard in modern LLMs, yet they accumulate all layer outputs with fixed unit weights. This uniform aggregation causes uncontrolled hidden-state growth with depth, progressively diluting each layer's contribution. We propose Attention Residuals (AttnRes), which replaces this fixed accumulation with softmax attention over preceding layer outputs, allowing each layer to selectively aggregate earlier representations with learned, input-dependent weights. To address the memory and communication overhead of attending over all preceding layer outputs for large-scale model training, we introduce Block AttnRes, which partitions layers into blocks and attends over block-level representations, reducing the memory footprint while preserving most of the gains of full AttnRes. Combined with cache-based pipeline communication and a two-phase computation strategy, Block AttnRes becomes a practical drop-in replacement for standard residual connections with minimal overhead. Scaling law experiments confirm that the improvement is consistent across model sizes, and ablations validate the benefit of content-dependent depth-wise selection. We further integrate AttnRes into the Kimi Linear architecture (48B total / 3B activated parameters) and pre-train on 1.4T tokens, where AttnRes mitigates PreNorm dilution, yielding more uniform output magnitudes and gradient distribution across depth, and improves downstream performance across all evaluated tasks.

cs.CL↗

arXiv2026

Evidential Domain Adaptation for Remaining Useful Life Prediction with Incomplete Degradation

Accurate Remaining Useful Life (RUL) prediction without labeled target domain data is a critical challenge, and domain adaptation (DA) has been widely adopted to address it by transferring knowledge from a labeled source domain to an unlabeled target domain. Despite its success, existing DA methods struggle significantly when faced with incomplete degradation trajectories in the target domain, particularly due to the absence of late degradation stages. This missing data introduces a key extrapolation challenge. When applied to such incomplete RUL prediction tasks, current DA methods encounter two primary limitations. First, most DA approaches primarily focus on global alignment, which can misaligns late degradation stage in the source domain with early degradation stage in the target domain. Second, due to varying operating conditions in RUL prediction, degradation patterns may differ even within the same degradation stage, resulting in different learned features. As a result, even if degradation stages are partially aligned, simple feature matching cannot fully align two domains. To overcome these limitations, we propose a novel evidential adaptation approach called EviAdapt, which leverages evidential learning to enhance domain adaptation. The method first segments the source and target domain data into distinct degradation stages based on degradation rate, enabling stage-wise alignment that ensures samples from corresponding stages are accurately matched. To address the second limitation, we introduce an evidential uncertainty alignment technique that estimates uncertainty using evidential learning and aligns the uncertainty across matched stages.

cs.LG↗

arXiv2026

An Extended Study of Gear-Ratio-Aware Standard Cell Layout Generation for DTCO Exploration

Advanced nodes decouple contacted poly pitch (CPP) and lower-metal pitch to improve routability. We present CPCell, an efficient standard-cell layout generation framework, to support arbitrary gear ratio (GR) and offset parameters through a fine-grained layered grid graph and constraint-programming-based placement-routing co-optimization. Layout quality is improved via Middle-of-Line routing, M0 pin enablement, pin accessibility constraints and a weighted multi-objective formulation that jointly optimizes cell layouts. To scale to netlists with up to 48 transistors, we incorporate acceleration techniques including transistor clustering, identical transistor partitioning, routing lower bound tightening and early termination strategies. Comprehensive cell-level and block-level studies are conducted to evaluate GR and offset choices, quantify the benefits of the proposed objectives and assess their impact on power, performance, area and IR-drop outcomes.

cs.AR↗

arXiv2026

Linking Perception, Confidence and Accuracy in MLLMs

Recent advances in Multi-modal Large Language Models (MLLMs) have predominantly focused on enhancing visual perception to improve accuracy. However, a critical question remains unexplored: Do models know when they do not know? Through a probing experiment, we reveal a severe confidence miscalibration problem in MLLMs. To address this, we propose Confidence-Driven Reinforcement Learning (CDRL), which uses original-noise image pairs and a novel confidence-based reward to enhance perceptual sensitivity and robustly calibrate the model's confidence. Beyond training benefits, calibrated confidence enables more effective test-time scaling as a free lunch. We further propose Confidence-Aware Test-Time Scaling (CA-TTS), which dynamically coordinates Self-Consistency, Self-Reflection, and Visual Self-Check modules guided by confidence signals. An Expert Model acts in multiple roles (e.g., Planner, Critic, Voter) to schedule these modules and provide external verification. Our integrated framework establishes new state-of-the-art results with consistent 8.8% gains across four benchmarks. More ablation studies demonstrate the effectiveness of each module and scaling superiority.

cs.CV↗

arXiv2026

An Updated Assessment of Reinforcement Learning for Macro Placement

We provide an improved assessment of Google Brain's deep reinforcement learning approach to macro placement and its updated Circuit Training (CT) implementation in GitHub. A stronger simulated annealing (SA) baseline leverages the "go-with-the-winners" metaheuristic and a multi-threading implementation. We develop and release new public benchmarks in sub-10nm technology: LEF/DEF for Google's 7nm TSMC Ariane protobuf and scaled variants, as well as testcases implemented in the open-source ASAP7 7nm research enablement. We evaluate from-scratch training and fine-tuning results for the latest "AlphaChip" release of Circuit Training, alongside multiple alternative macro placers. We also study the recently-published pre-training guidance in. A commercial place-and-route tool is used to provide "true reward" post-route power, performance and area metrics. All data, evaluation flows and related scripts are publicly available in the MacroPlacement GitHub repository. Our study affords insights into reproducibility and reporting in the research literature, and points out still-missing confirmations (e.g., of CT's scalability and pre-training methodology) that remain open questions for the research community.

cs.LG↗

arXiv2026

CARE-Edit: Condition-Aware Routing of Experts for Contextual Image Editing

Unified diffusion editors often rely on a fixed, shared backbone for diverse tasks, suffering from task interference and poor adaptation to heterogeneous demands (e.g., local vs global, semantic vs photometric). In particular, prevalent ControlNet and OmniControl variants combine multiple conditioning signals (e.g., text, mask, reference) via static concatenation or additive adapters which cannot dynamically prioritize or suppress conflicting modalities, thus resulting in artifacts like color bleeding across mask boundaries, identity or style drift, and unpredictable behavior under multi-condition inputs. To address this, we propose Condition-Aware Routing of Experts (CARE-Edit) that aligns model computation with specific editing competencies. At its core, a lightweight latent-attention router assigns encoded diffusion tokens to four specialized experts--Text, Mask, Reference, and Base--based on multi-modal conditions and diffusion timesteps: (i) a Mask Repaint module first refines coarse user-defined masks for precise spatial guidance; (ii) the router applies sparse top-K selection to dynamically allocate computation to the most relevant experts; (iii) a Latent Mixture module subsequently fuses expert outputs, coherently integrating semantic, spatial, and stylistic information to the base images. Experiments validate CARE-Edit's strong performance on contextual editing tasks, including erasure, replacement, text-driven edits, and style transfer. Empirical analysis further reveals task-specific behavior of specialized experts, showcasing the importance of dynamic, condition-aware processing to mitigate multi-condition conflicts.

cs.CV↗

arXiv2026

Unmasking Reasoning Processes: A Process-aware Benchmark for Evaluating Structural Mathematical Reasoning in LLMs

Recent large language models (LLMs) achieve near-saturation accuracy on many established mathematical reasoning benchmarks, raising concerns about their ability to diagnose genuine reasoning competence. This saturation largely stems from the dominance of template-based computation and shallow arithmetic decomposition in existing datasets, which underrepresent reasoning skills such as multi-constraint coordination, constructive logical synthesis, and spatial inference. To address this gap, we introduce ReasoningMath-Plus, a benchmark of 150 carefully curated problems explicitly designed to evaluate structural reasoning. Each problem emphasizes reasoning under interacting constraints, constructive solution formation, or non-trivial structural insight, and is annotated with a minimal reasoning skeleton to support fine-grained process-level evaluation. Alongside the dataset, we introduce HCRS (Hazard-aware Chain-based Rule Score), a deterministic step-level scoring function, and train a Process Reward Model (PRM) on the annotated reasoning traces. Empirically, while leading models attain relatively high final-answer accuracy (up to 5.8/10), HCRS-based holistic evaluation yields substantially lower scores (average 4.36/10, best 5.14/10), showing that answer-only metrics can overestimate reasoning robustness.

cs.AI↗

arXiv2026

The fundamental localization phases in quasiperiodic systems: A unified framework and exact results

The disordered quantum systems host three classes of quantum states, the extended, localized, and critical, which bring up seven distinct fundamental phases in nature: three pure phases and four coexisting ones with mobility edges, yet a unified theory built on universal mechanism and full realization of all these phases has not been developed. Here we propose a unified framework based on a spinful quasiperiodic (QP) system which realizes all the fundamental localization phases, with the exact and universal results being obtained for their characterization. First, we show that the pure phases are obtained when the chiral(-like) symmetry preserves in the proposed spinful QP model, giving a criterion for emergence of the pure phases and otherwise the coexisting ones. Further, we uncover a novel mechanism for the critical states that their emergence is protected by the generalized incommensurate matrix element zeros in the spinful QP model, which considerably broadens rigorous realizations of the exotic critical states. We then show criteria of exact solvability for the present spinful QP system, with which we construct various exactly solvable models for all distinct localization phases. In particular, we propose two novel models, dubbed spin-selective QP lattice model and QP optical Raman lattice model, to achieve all basic types of mobility edges and all the seven fundamental phases of Anderson localization physics, respectively. The experimental scheme is proposed and studied in detail to realize these models with high feasibility. This study establishes a complete and profound theoretical framework which enables an in-depth exploration of the broad classes of all fundamental localization phenomena in QP systems, and offers key insights for constructing their exactly solvable models with experimental feasibility.

cond-mat.dis-nn↗

arXiv2026

Physics-Informed Glass-Structure Descriptors for Assessing the Intrinsic Reactivity of Mixed Amorphous-Crystalline Precursors in Alkali-Activated Materials

Rapid and reliable assessment of the intrinsic reactivity of amorphous aluminosilicates is critical for their application in alkali-activated materials (AAMs) and blended cements. Although physics-informed glass-structure descriptors have demonstrated strong structure-reactivity relationships for predominantly amorphous systems, their extension to heterogeneous precursors with mixed crystalline-amorphous phases has been limited. Here, quantitative X-ray diffraction combined with bulk compositional analysis was used to reconstruct the effective amorphous compositions of five fly ashes (FAs) and three ground granulated blast-furnace slags (GGBSs). These compositions served as inputs for molecular dynamics simulations employing a melt-and-quench approach to generate atomic-scale structural models of the glassy phases. Based on these structures, the previously introduced descriptors, i.e., average metal oxygen dissociation energy and average metal oxygen bond strength, were refined to cover a broader compositional space spanning SiO2-Al2O3-TiO2-Fe2O3-CaO-MgO-MnO-Na2O-K2O. The refined descriptors exhibit strong inverse correlations with multiple independent reactivity indicators, including cumulative heat release from isothermal calorimetry, bound water content from thermogravimetric analysis, and compressive strength, for both single precursors and binary FA-GGBS blends activated with NaOH. These results demonstrate that physics-informed glass-structure descriptors can be extended from ideal amorphous systems to heterogeneous mixed-phase precursors and capture relative intrinsic reactivity trends in alkaline solutions. The proposed framework provides a transferable, structure-informed basis for comparative assessment of precursor reactivity that complements experimental testing and may inform precursor screening and mix designs for AAM and blended cement systems.

cond-mat.mtrl-sci↗

arXiv2026

Extended Landauer-Büttiker Formula for Current through Open Quantum Systems with Gain or Loss

The Landauer-Büttiker formula, which characterizes the current flowing through a finite region connected to leads, has significantly advanced our understanding of transport. We extend this formula to describe particle and energy currents with gain or loss in the intermediate region by using the Lindblad-Keldysh formalism. Based on the derived formula, several novel effects induced by gain or loss in the current are discussed: the breaking of inversion symmetry in the gain and loss terms or in the system can lead to current generation; the anomalous phenomenon that disorder can induce current generation; the presence of gain and loss makes the thermal and electrical conductances continuous and ensures they follow the Wiedemann-Franz law even outside the energy band; the effect of bond loss-induced skin effect on current. This work deepens and extends our understanding of transport phenomena in open systems.

cond-mat.mes-hall↗

arXiv2026

Controllable Video Generation: A Survey

With the rapid development of AI-generated content (AIGC), video generation has emerged as one of its most dynamic and impactful subfields. In particular, the advancement of video generation foundation models has led to growing demand for controllable video generation methods that can more accurately reflect user intent. Most existing foundation models are designed for text-to-video generation, where text prompts alone are often insufficient to express complex, multi-modal, and fine-grained user requirements. This limitation makes it challenging for users to generate videos with precise control using current models. To address this issue, recent research has explored the integration of additional non-textual conditions, such as camera motion, depth maps, and human pose, to extend pretrained video generation models and enable more controllable video synthesis. These approaches aim to enhance the flexibility and practical applicability of AIGC-driven video generation systems. In this survey, we provide a systematic review of controllable video generation, covering both theoretical foundations and recent advances in the field. We begin by introducing the key concepts and commonly used open-source video generation models. We then focus on control mechanisms in video diffusion models, analyzing how different types of conditions can be incorporated into the denoising process to guide generation. Finally, we categorize existing methods based on the types of control signals they leverage, including single-condition generation, multi-condition generation, and universal controllable generation. For a complete list of the literature on controllable video generation reviewed, please visit our curated repository at https://github.com/mayuelala/Awesome-Controllable-Video-Generation.

cs.GR↗

arXiv2026

Rewarding the Rare: Uniqueness-Aware RL for Creative Problem Solving in LLMs

Reinforcement learning (RL) has become a central paradigm for post-training large language models (LLMs), particularly for complex reasoning tasks, yet it often suffers from exploration collapse: policies prematurely concentrate on a small set of dominant reasoning patterns, improving pass@1 while limiting rollout-level diversity and gains in pass@k. We argue that this failure stems from regularizing local token behavior rather than diversity over sets of solutions. To address this, we propose Uniqueness-Aware Reinforcement Learning, a rollout-level objective that explicitly rewards correct solutions that exhibit rare high-level strategies. Our method uses an LLM-based judge to cluster rollouts for the same problem according to their high-level solution strategies, ignoring superficial variations, and reweights policy advantages inversely with cluster size. As a result, correct but novel strategies receive higher rewards than redundant ones. Across mathematics, physics, and medical reasoning benchmarks, our approach consistently improves pass@$k$ across large sampling budgets and increases the area under the pass@$k$ curve (AUC@$K$) without sacrificing pass@1, while sustaining exploration and uncovering more diverse solution strategies at scale.

cs.LG↗

arXiv2026

Collaborative Multi-Agent Test-Time Reinforcement Learning for Reasoning

Multi-agent systems have evolved into practical LLM-driven collaborators for many applications, gaining robustness from diversity and cross-checking. However, multi-agent RL (MARL) training is resource-intensive and unstable: co-adapting teammates induce non-stationarity, and rewards are often sparse and high-variance. Therefore, we introduce \textbf{Multi-Agent Test-Time Reinforcement Learning (MATTRL)}, a framework that injects structured textual experience into multi-agent deliberation at inference time. MATTRL forms a multi-expert team of specialists for multi-turn discussions, retrieves and integrates test-time experiences, and reaches consensus for final decision-making. We also study credit assignment for constructing a turn-level experience pool, then reinjecting it into the dialogue. Across challenging benchmarks in medicine, math, and education, MATTRL improves accuracy by an average of 3.67\% over a multi-agent baseline, and by 8.67\% over comparable single-agent baselines. Ablation studies examine different credit-assignment schemes and provide a detailed comparison of how they affect training outcomes. MATTRL offers a stable, effective and efficient path to distribution-shift-robust multi-agent reasoning without tuning.

cs.AI↗

arXiv2026

A Unified Shape-Aware Foundation Model for Time Series Classification

Foundation models pre-trained on large-scale source datasets are reshaping the traditional training paradigm for time series classification. However, existing time series foundation models primarily focus on forecasting tasks and often overlook classification-specific challenges, such as modeling interpretable shapelets that capture class-discriminative temporal features. To bridge this gap, we propose UniShape, a unified shape-aware foundation model designed for time series classification. UniShape incorporates a shape-aware adapter that adaptively aggregates multiscale discriminative subsequences (shapes) into class tokens, effectively selecting the most relevant subsequence scales to enhance model interpretability. Meanwhile, a prototype-based pretraining module is introduced to jointly learn instance- and shape-level representations, enabling the capture of transferable shape patterns. Pre-trained on a large-scale multi-domain time series dataset comprising 1.89 million samples, UniShape exhibits superior generalization across diverse target domains. Experiments on 128 UCR datasets and 30 additional time series datasets demonstrate that UniShape achieves state-of-the-art classification performance, with interpretability and ablation analyses further validating its effectiveness.

cs.LG↗

arXiv2025

Deriving the Eigenstate Thermalization Hypothesis from Eigenstate Typicality and Kinematic Principles

The eigenstate thermalization hypothesis (ETH) provides a powerful framework for understanding thermalization in isolated quantum many-body systems, yet a complete and conceptually transparent derivation has remained elusive. In this work, we derive the structure of ETH from a minimal dynamical principle, which we term the eigenstate typicality principle (ETP), together with general kinematic ingredients arising from entropy maximization, Hilbert-space geometry, and locality. ETP asserts that in quantum-chaotic systems, energy eigenstates are statistically indistinguishable, with respect to local measurements, from states drawn from the Haar measure on a narrow microcanonical shell. Within this framework, diagonal ETH arises from concentration of measure, provided that eigenstate typicality holds. The structure of off-diagonal matrix elements is then fixed by entropic scaling and the finite-time dynamical correlations of local observables, with ETP serving as the dynamical bridge to energy eigenstates, without invoking random-matrix assumptions. Our results establish ETH as a consequence of entropy, Hilbert-space geometry, and chaos-induced eigenstate typicality, and clarify its regime of validity across generic quantum-chaotic many-body systems, thereby deepening our understanding of quantum thermalization and the emergence of statistical mechanics from unitary many-body dynamics.

cond-mat.stat-mech↗

arXiv2025

Retrieving Filter Spectra in CNN for Explainable Sleep Stage Classification

Despite significant advances in deep learning-based sleep stage classification, the clinical adoption of automatic classification models remains slow. One key challenge is the lack of explainability, as many models function as black boxes with millions of parameters. In response, recent work has increasingly focussed on enhancing model explainability. This study contributes to these efforts by introducing an explainability tool for spectral processing of individual EEG channels. Specifically, this tools retrieves the filter spectrum of low-level convolutional feature extraction and compares it with the classification-relevant spectral information in the data. We apply our tool on the EEGNet and MSA-CNN models using the ISRUC-S3 and Sleep-EDF-20 datasets. The tool reveals that spectral processing plays a significant role in the lower frequency bands. In addition, comparing the correlation between filter spectrum and data-derived spectral information with univariate performance indicates that the model naturally prioritises the most informative channels in a multimodal setting. We specify how these insights can be leveraged to enhance model performance. The code for the filter spectrum retrieval and its analysis is available at https://github.com/sgoerttler/MSA-CNN.

eess.SP↗

arXiv2025

MonoDream: Monocular Vision-Language Navigation with Panoramic Dreaming

Vision-Language Navigation (VLN) tasks often leverage panoramic RGB and depth inputs to provide rich spatial cues for action planning, but these sensors can be costly or less accessible in real-world deployments. Recent approaches based on Vision-Language Action (VLA) models achieve strong results with monocular input, yet they still lag behind methods using panoramic RGB-D information. We present MonoDream, a lightweight VLA framework that enables monocular agents to learn a Unified Navigation Representation (UNR). This shared feature representation jointly aligns navigation-relevant visual semantics (e.g., global layout, depth, and future cues) and language-grounded action intent, enabling more reliable action prediction. MonoDream further introduces Latent Panoramic Dreaming (LPD) tasks to supervise the UNR, which train the model to predict latent features of panoramic RGB and depth observations at both current and future steps based on only monocular input. Experiments on multiple VLN benchmarks show that MonoDream consistently improves monocular navigation performance and significantly narrows the gap with panoramic-based agents.

cs.CV↗

arXiv2025

A Survey on Speech Large Language Models for Understanding

Speech understanding is essential for interpreting the diverse forms of information embedded in spoken language, including linguistic, paralinguistic, and non-linguistic cues that are vital for effective human-computer interaction. The rapid advancement of large language models (LLMs) has catalyzed the emergence of Speech Large Language Models (Speech LLMs), which marks a transformative shift toward general-purpose speech understanding systems. To further clarify and systematically delineate task objectives, in this paper, we formally define the concept of speech understanding and introduce a structured taxonomy encompassing its informational, functional, and format dimensions. Within this scope of definition, we present a comprehensive review of current Speech LLMs, analyzing their architectures through a three-stage abstraction: Modality Feature Extraction, Modality Information Fusion, and LLM Inference. In addition, we examine training strategies, discuss representative datasets, and review evaluation methodologies adopted in the field. Based on empirical analyses and experimental evidence, we identify two key challenges currently facing Speech LLMs: instruction sensitivity and degradation in semantic reasoning and propose concrete directions for addressing these issues. Through this systematic and detailed survey, we aim to offer a foundational reference for researchers and practitioners working toward more robust, generalizable, and human-aligned Speech LLMs.

eess.AS↗