Searcharxiv⌕ Search

arXiv subjects

Yi Yuan

Publications and source records attributed to Yi Yuan.

At least 19 recordsLinked to original sources

Future Muon Physics Experiments: Muon Beams and Experimental Apparatus

Muons are powerful probes of fundamental physics at the precision and intensity frontiers. We review the status and planned upgrades of selected muon-beam facilities worldwide, together with the experimental apparatus associated with representative experiments across three categories. Charged-lepton flavor violation (CLFV) searches test charged-lepton flavor conservation through rare muon decays, muon-to-electron conversion in nuclei, and muonium--antimuonium conversion. Precision measurements of the muon magnetic and electric dipole moments, muonium spectroscopy, and muonic-atom spectroscopy test the Standard Model and bound-state QED and determine fundamental constants and nuclear charge and magnetization distributions. Muon-scattering programs span low-momentum elastic lepton--proton scattering with PSI MUSE to study the proton radius, lepton universality, and two-photon exchange; high-energy MUonE measurements of hadronic vacuum polarization; NA64$μ$ missing-momentum searches for invisible dark sectors; and the phased PKMu program, which investigates muonphilic dark matter, light mediators, LFV muon--electron processes, and atomic effects in resonant annihilation. We emphasize the beam intensity, timing, purity, and phase-space control, as well as beamline and detector design, background suppression, and experimental techniques needed to achieve next-generation sensitivity and precision.

hep-ex↗

Single-shot state preparation thresholds in quantum LDPC codes: soundness, confinement, and statistical mechanics

Single-shot error correction does not by itself establish equally direct and robust logical-state initialization. We study preparation of CSS quantum low-density parity-check code states using one noisy round of the code's checks. We prove that linear confinement over a weight range growing at least logarithmically, together with at least logarithmic distance, suffices for a nonzero threshold under independent local-stochastic readout and data noise in the relevant CSS sector. The threshold is certified by ideal recovery of a decoder-defined logical channel. The key is a correction rule that preserves the logical class under local syndrome changes, allowing the proof to proceed without a local-stochastic bound on the preparation residual. This gives a single-block guarantee for quantum expander codes without dimension jumps or batching. Under the stronger linear-soundness hypothesis, we establish local-stochastic residual noise, enabling composition with compatible fault-tolerant gadgets. For a retained-record formulation, we also derive an exact statistical-mechanical representation of optimal final recovery at fixed preparation decoder. A decoder-dependent logical-sector order parameter characterizes asymptotically successful block decoding. Monte Carlo results for the three-dimensional toric code provide finite-size evidence for a threshold beyond both linear sufficient conditions. These results distinguish logical robustness from local-stochastic residual-noise control and identify decoder consistency as a route to single-shot initialization.

quant-ph↗

SplatFlux: Feed-Forward 4D Gaussian Reconstruction and Generation from Video Latents

This work presents SplatFlux, a feed-forward latent-to-4DGS framework for both video-based 4D reconstruction and conditional 4D generation. Our key observation is that observed videos and conditionally generated content can be represented in the same pretrained video latent space, allowing both tasks to be formulated as latent-to-4DGS decoding. To construct compact and temporally coherent 4D scenes, SplatFlux introduces a static-dynamic decoupled 4D Gaussian representation that explicitly predicts dynamic probabilities for Gaussian primitives. This representation separates static scene structures from moving regions, enabling complete 4D scene construction in a feed-forward manner through static aggregation and dynamic motion interpolation. To learn static-dynamic decomposition and dynamic Gaussian motion without costly annotations, we further develop an auxiliary distillation strategy that provides pseudo dynamic masks and optical-flow cues. Extensive experiments across 3D and 4D reconstruction and generation demonstrate that SplatFlux achieves state-of-the-art performance.

cs.CV↗

Anchor3DGS: Feed-Forward 3D Gaussian Splatting with Compact Anchor-Based Representation

Recent feed-forward 3D Gaussian Splatting (3DGS) methods have enabled effective multi-view 3D reconstruction by predicting Gaussian primitives in a single forward pass. However, existing approaches predominantly adopt a pixel-aligned paradigm that predicts one Gaussian per input pixel, coupling the Gaussian count to the input resolution and number of views. This also leads to redundant primitives in textureless regions and insufficient coverage in geometrically complex areas. We present Anchor3DGS, a pose-free feed-forward 3DGS framework that replaces pixel-aligned prediction with a compact anchor-based representation. Guided by information entropy, our method places a set of 3D anchors in the scene, concentrating them in regions of higher visual complexity while maintaining broad spatial coverage. Each anchor performs occlusion-aware aggregation of multi-view image features, exchanges information with other anchors, and is decoded into a set of local Gaussians. A lightweight feature enhancement module further refines the coarse RGB output from Gaussian splatting and improves rendering fidelity. Extensive experiments on diverse benchmark datasets demonstrate that our Anchor3DGS achieves state-of-the-art novel view synthesis quality while only using an order of magnitude fewer Gaussian primitives.

cs.CV↗

More Methods, Better Agents? A Reality Check on Self-Distillation in Agentic Reinforcement Learning

Self-distillation in agentic reinforcement learning has produced a growing number of methods. However, existing evaluations often reuse published baseline scores under differing experimental conditions, provide few direct comparisons among contemporaneous methods, and give limited attention to training stability and computational cost. These limitations obscure the source of reported gains and the practical strengths and weaknesses of competing methods. Costly agent training further motivates a common basis for comparison. To address these gaps, we compare ten representative methods against GRPO on ALFWorld, ScienceWorld, Search-QA, and WebShop across three model sizes. Our study comprises 132 training runs totaling 12,124.4 recorded H800 GPU-hours, with repeated evaluations over eight rollout seeds for the pretrained backbones and all eleven trained methods. Beyond final task performance, we examine training progress, stability, and measured computational cost. Our evaluation yields three key findings. First, gains over GRPO remain limited in scope, with no method consistently outperforming it across settings. Second, strong final performance can mask unstable training. Third, additional training cost does not consistently translate into better performance. We establish this benchmark to support fair, comprehensive comparison and guide future research toward reliable and cost-effective improvements in agent training. Project page: https://aim-uofa.github.io/Self-Distillation-AgentRL/

cs.CL↗

Zeno-Enhanced Probabilistic Error Cancellation with Quantum Error Detection Codes

Probabilistic error cancellation (PEC) is unbiased but suffers exponential sampling overhead set by noise-weighted circuit volume, whereas quantum error-detecting codes (QEDCs) remove many physical faults by stabilizer post-selection but leave an undetectable logical residue. We exploit this complementarity by using post-selection to map physical noise to a weaker accepted logical channel, and then applying PEC only to the residual channel. The resulting feedback-free QED+PEC scheme interleaves Clifford logical blocks, stabilizer measurements, post-selection, and probabilistic cancellation on accepted trajectories, without real-time decoding or active recovery. A key complication is that post-selection correlates accepted fault branches through stabilizer-commutation constraints, so the sparse Pauli-Lindblad factorization underlying bare PEC no longer applies directly. We therefore construct the inverse channel perturbatively: for fixed order $K$, only accepted fault branches up to order $K$ are retained, reducing preprocessing from $2^m$ branches to $O(m^K)$ per block. The order-$K$ protocol cancels the normalized post-selected channel through degree $K$, leaving a per-block error $O(W^{K+1})$ that accumulates at most linearly. For logical GHZ-state preparation with the $[[n,n-2,2]]$ Iceberg code under circuit-level depolarizing noise and ideal stabilizer measurements, first-order QED+PEC reaches $n=200$ physical qubits and lowers sampling overhead by three to four orders of magnitude relative to standard PEC while maintaining $F\simeq0.956$. Syndrome-noise tests show that readout-only flips mainly increase post-selection cost, whereas noisy GHZ-assisted global stabilizer extraction can remove the advantage. This identifies a discrete-Zeno trade-off: cheap detection reshapes the effective channel PEC must invert, rather than simply adding overhead.

quant-ph↗

PMOPD: Task Ordering, Cycling, and Parameter-Update Subspace Protection in Multi-Teacher On-Policy Distillation

Multi-teacher on-policy distillation (MOPD) has emerged as a popular post-training paradigm for integrating specialized capabilities in frontier language models. Existing OPD research has primarily focused on optimizing single-task distillation through objective design, distillation scope, and teacher signal construction, whereas MOPD must aggregate multiple capabilities in shared parameters and address the resulting capability seesaw, in which improving one domain suppresses capabilities acquired from another. Inspired by the distinctive update geometry of OPD, we find that parameter updates from different tasks rapidly concentrate in their respective low-dimensional subspaces during MOPD, providing a direct geometric basis for identifying and controlling cross-task interference. We therefore propose PMOPD (Projection-based Multi-Teacher On-Policy Distillation), which constructs subspace memories from the cumulative parameter displacements of different tasks and projects both gradients and optimizer updates to remove components that interfere with protected task directions. We further develop a lightweight conflict probe to characterize task interactions and guide task ordering, together with a cycling strategy that balances subspace estimation and timely task revisitation. Experiments on representative Code, Reason, and Math tasks show that PMOPD improves every evaluated capability over MOPD, raising the average score across the three tasks by 2.54 points on Qwen2.5-7B and 2.09 points on Llama-3.1-8B. These consistent gains establish geometry-aware optimization as an effective and transferable approach to balanced multi-teacher distillation.

cs.LG↗

Beyond Retrieval: Progressive Latent Memory Evolution for Streaming Video Understanding

Streaming video understanding requires multimodal large language models (MLLMs) to process continuous visual inputs and respond to user queries under strict causality and bounded memory. Existing approaches typically compress historical observations into an external memory bank and retrieve query-relevant evidence as additional visual context. Though effective, this store-and-retrieve paradigm keeps historical evidence as external visual context, preventing it from being internalized into a compact, evolving latent memory that can continuously guide streaming reasoning. To bridge this gap, we introduce LatentStream, a progressive latent working memory framework that shifts streaming memory from store-and-retrieve to retrieve-and-internalize. Specifically, LatentStream comprises three coordinated components. First, Query-agnostic Hierarchical Streaming Memory organizes visual history into short-, mid-, and long-term levels under a fixed memory budget through Jenks-guided adaptive consolidation. Once a query arrives, Hierarchical Latent Memory Evolution equips groups of latent memory tokens with progressively expanding memory receptive fields, enabling them to iteratively retrieve historical evidence from their corresponding scopes and internalize it into a compact, fixed-length latent memory. Finally, Progressive Confidence-guided Latent Memory Optimization constructs a hierarchical progression reward from group-wise predictive entropy and jointly refines the latent memory tokens and retrieved evidence, encouraging increasingly confident streaming reasoning. Extensive experiments demonstrate that LatentStream achieves new state-of-the-art results on existing online and offline video benchmarks.

cs.CV↗

FlowSep 2: Self-Supervised Flow Matching for Language-Queried Audio Source Separation

Language-queried audio source separation (LASS) aims to extract target sources from audio mixtures according to natural language descriptions, offering a flexible and scalable interface for audio source separation. However, most existing LASS methods rely on discriminative, mask-based models, which estimate masks from the input mixture. These methods often over-suppress target sounds or fail to fully separate them, especially when multiple sound events strongly overlap in complex acoustic scenes. In this work, we propose FlowSep2, a text-conditioned flow-matching generative model for LASS. Instead of directly predicting a separation mask, FlowSep2 learns to generate the target source representation from Gaussian noise in a latent space, conditioned on both the mixture representation and the text query. Specifically, we employ rectified flow matching with a Diffusion Transformer backbone. We further incorporate Self-Flow, a self-supervised flow-matching paradigm, into our LASS framework. By encouraging semantically structured latent representations under the generative objective, Self-Flow improves the model's ability to separate target sources according to text queries. Experiments on multiple LASS benchmarks show that FlowSep2 achieves state-of-the-art performance and demonstrates enhanced sound separation results in challenging scenarios with overlapping sound events.

cs.SD↗

VideoGAIA: A Benchmark for General AI Assistants on Agentic Video Understanding

Video understanding is a fundamental task for evaluating the capabilities of multimodal large language models (MLLMs). However, existing leading models have already achieved approximately 90% accuracy on the Video-MME leaderboard, suggesting that conventional single-turn video understanding tasks are becoming increasingly saturated and insufficient for assessing the intelligence of advanced MLLMs. Towards this end, we introduce VideoGAIA, an agentic video understanding benchmark for general artificial intelligence (AI) assistants. Moving beyond one-shot video question answering, VideoGAIA formulates video understanding as a multi-turn, tool-augmented interaction process, where models must iteratively perceive videos, invoke external tools, gather complementary information, and integrate multimodal evidence across turns. VideoGAIA contains 271 model-human co-designed tasks covering diverse and complex real-world scenarios. Each video-question-answer instance is independently verified by three human experts to ensure both correctness and appropriate difficulty. All evaluated MLLMs, including frontier models such as GPT-5.5 and Kimi-K3, achieve less than 60% accuracy on VideoGAIA, highlighting its value as a high-quality and timely benchmark for evaluating next-generation MLLMs. We hope that VideoGAIA will facilitate the transition from conventional video understanding toward agentic video understanding.

cs.CV↗

Conditional entropy realization and approximation by uniquely ergodic measures

This paper studies conditional entropy realization and weak* approximation by uniquely ergodic measures with compact support. We prove that, after fixing an admissible potential average and an entropy strictly below the entropy supremum over the corresponding average fiber, every invariant measure satisfying these two exact constraints can be approximated weakly* by uniquely ergodic measures with compact support satisfying the same constraints. Each approximating measure is the unique invariant measure on its minimal support, whose topological entropy equals the prescribed metric entropy. This result holds for three broad classes of systems: topologically expanding maps (including topologically Anosov systems), transitive countable Markov shifts, and symbolic systems with non-uniform structure. The proof uses a nested multi-horseshoe construction, with separate arguments addressing non-invertibility, non-compactness and non-uniformity.

math.DS↗

Efficient Text-to-Audio Generation via Pruning

Diffusion-based text-to-audio generative models such as AudioLDM achieve high perceptual quality and strong semantic consistency; however, their practical deployment is hindered by the substantial computational cost of the U-Net denoising backbone. In this work, we apply model pruning to improve the computational efficiency of AudioLDM, a U-Net-based text-conditioned audio latent diffusion model. We analyse parameter redundancy across U-Net convolutional blocks and evaluate a filter-pruning strategy. Pruning is guided by norm-based criteria and followed by lightweight finetuning to recover performance losses. Experimental results demonstrate that up to 83% of the parameters and 39% of the multiply-accumulate operations of U-Net have been reduced while maintaining, and in some cases improving, generation quality compared to the baseline unpruned network. We find that pruning affects AudioLDM's ability to generate certain sound events including safety-critical sounds such as gunshots, sirens, and explosions, as well as mechanical sounds such as drills and sewing machines, and other sounds such as sprays and tick-tocks, which are mostly recovered by lightweight finetuning of the pruned model.

eess.AS↗

Scaling with Confidence: Calibrating Confidence of LLMs for Adaptive Test Time Scaling

Training large language models (LLMs) with reinforcement learning (RL) has significantly advanced their performance on reasoning and question-answering tasks. However, prevailing RL reward designs typically prioritize response correctness, neglecting to incentivize models to express their confidence accurately. This leads to a critical problem: performance gains are often accompanied by poor calibration between confidence and accuracy, misleading models to overconfidently hallucinate when uncertain. To address this limitation, we propose $\textbf{C}$orrectness and $\textbf{C}$onfidence $\textbf{C}$alibration $\textbf{R}$einforcement $\textbf{L}$earning ($\textbf{C3RL}$), a novel RL algorithm integrating correctness, calibration and dataset-informed reference accuracy rewards together. Comprehensive evaluation across 8 text and multimodal datasets demonstrates that C3RL enhances calibration without sacrificing accuracy, outperforming the current state-of-the-art method in both performance and calibration metrics. Utilizing the well-calibrated verbalized confidence from C3RL, we further introduce $\textbf{C}$onfidence-based $\textbf{A}$daptive Test Time $\textbf{S}$caling ($\textbf{CAS}$), an adjustable inference-time strategy that allocates computational resources based on response confidence. Experiments show that CAS surpasses majority voting on both in-domain and out-of-domain datasets while reducing the inference budget by up to 12.33 times. We believe the synergy of C3RL and CAS paves the way for deploying more reliable and resource-efficient LLMs. The code, data and models will be released.

cs.AI↗

MemDreamer: Decoupling Perception and Reasoning for Long Video Understanding via Hierarchical Graph Memory and Agentic Retrieval Mechanism

Current Vision-Language Models struggle with hours-long videos because processing full-length visual sequences induces prohibitive token explosion and attention dilution. To overcome this, we introduce MemDreamer to decouple perception and reasoning, shifting long-video understanding into an agentic exploration process. As a plug-and-play framework, it incrementally streams videos to construct a Hierarchical Graph Memory, a top-down three-tier architecture for semantic abstraction, anchored by a foundational graph capturing spatiotemporal and causal relations. During inference, the reasoning model employs agentic tool-augmented retrieval, navigating hierarchies, searching nodes, and traversing logical edges via an Observation-Reason-Action loop. Experiments show MemDreamer achieves SOTA results across four mainstream benchmarks, narrowing the gap with human experts to only 3.7 points. It constrains the reasoning context window to merely 2% of full-context ingestion while delivering a 12.5 point absolute accuracy gain. Furthermore, statistical analysis uncovers a strong positive linear correlation between an VLM's performance on logic reasoning and long-video understanding benchmarks, establishing agentic capability scaling as a new paradigm for multimodal comprehension.

cs.CV↗

SPRI: SVD-Partitioned Residual Initialization for Data-Constrained MoE Upcycling

Mixture-of-Experts (MoE) models enable efficient scaling, but training them from scratch remains prohibitively expensive. MoE upcycling mitigates this cost by converting pretrained dense models into sparse MoE models. However, existing upcycling methods typically rely on large-scale continued training and often perform poorly under data-constrained supervised adaptation, due to either homogeneous experts or overly disruptive perturbations to pretrained parameters. In this setting, effective upcycling must leverage pretrained weight structure while introducing sufficient diversity among routed experts. To this end, we propose SVD-Partitioned Residual Initialization (SPRI), which distributes SVD-partitioned residuals derived from pretrained feed-forward network (FFN) weights across routed experts, introducing controlled expert diversity grounded in pretrained spectral structure. We further introduce a two-stage training strategy to improve adaptation stability. We evaluate SPRI on multilingual speech-to-text translation, where limited supervised data challenges MoE upcycling and multiple target languages provide natural routing heterogeneity. On CoVoST2 across 15 En-to-XX directions, SPRI improves average BLEU and COMET over fully fine-tuned dense models by 2.58 and 3.32 points, respectively, and outperforms the prior best MoE upcycling baseline by 3.39 BLEU and 4.34 COMET points.

cs.LG↗

Towards Long-Horizon Vessel Trajectory and Destination Forecasting with Reasoning Large Language Models

Long-horizon maritime trajectory prediction is important for shipping management, logistics planning, and maritime risk analysis, yet month-level forecasting remains insufficiently studied. Existing deep learning methods mainly focus on short- and mid-term coordinate extrapolation and often struggle to preserve route feasibility and destination correctness over extended horizons. This paper investigates joint long-horizon vessel trajectory and destination forecasting with reasoning-capable large language models, and develops a Maritime LLM post-training framework based on Reinforcement Learning with Verifiable Reward (RLVR). An AIS-based benchmark is constructed with 60-day historical trajectories and 30-day forecasting horizons, where trajectories are converted into semantic textual representations for RL prompt construction. RLVR aligns LLMs with maritime forecasting objectives by enforcing physical validity, providing early-weighted trajectory supervision, and evaluating destination correctness through hierarchical matching and curriculum learning. Experimental results show that RLVR-trained LLMs substantially improve over zero-shot LLMs and representative deep learning baselines, especially on destination-related metrics. Among the evaluated RLVR-trained variants, 4B LLMs achieve the best overall performance, suggesting that reward-compatible optimization and task-specific capacity matching are more important than simply using larger 8B or 14B LLMs. The results also show that LSTM remains a strong deep learning baseline under limited fine-tuning data, while Transformer-style spatio-temporal models typically require larger datasets and richer structured inputs. Overall, this work advances semantic, verifier-aligned maritime forecasting for operational decision support.

cs.AI↗

Three-dimensional density and air-rock interface reconstruction with muography: Application to the TianQin tunnel

Muography is a non-invasive imaging technique that uses cosmic-ray muons, commonly divided into transmission (absorption) and scattering muography. For transmission muography, the inversion algorithm critically determines reconstruction quality. However, widely used schemes may produce smearing artifacts when measurement locations are limited and data are sparse. We develop an optimized Metropolis--Hastings (M--H) algorithm that mitigates smearing and retrieves sharper, more accurate density distributions without auxiliary data. Additionally, we implement an inverse distance weighting (IDW) approach to reconstruct the air--rock interface from muon measurements. The optimized M--H algorithm is applied in Monte Carlo simulations and applied to field data from the TianQin Tunnel experiment using the MuGrid-v2 detector. The IDW-reconstructed air--rock interface is validated against Light Detection and Ranging (LiDAR) measurements. In simulations, the optimized M--H algorithm improves high-density anomaly detection precision from $42\%$ to $100\%$ at threshold $5.1\,\mathrm{g/cm^3}$, with gains of $6\%$ to $42\%$ across other threshold and low-density scenarios, together with the TianQin Tunnel reconstructions, these results demonstrate the effectiveness of the proposed approach.

physics.ins-det↗

Dual Tuning for Reasoning Efficacy-Driven Data Curation in Multimodal LLM Training

Reasoning post-training improves Large Language Models (LLMs) on complex tasks such as mathematics and coding, but its benefits across diverse multimodal tasks remains uncertain. The trend of releasing parallel "Instruct" and "Thinking" models by leading teams is both resource-intensive and user-unfriendly. Prior work finds that the gains from reasoning training are influenced by multiple factors, such as base model capabilities, task characteristics, and Chain-of-Thought (CoT) data quality. However, principled criteria for determining when reasoning post-training is beneficial and which data should support it are still lacking. In this paper, we propose Dual Tuning, a reasoning efficacy-driven data curation framework for multimodal LLMs training. Given a target task and a base model, Dual Tuning jointly evaluates whether the training data is beneficial and whether reasoning training with current CoT content yields positive gains over non-reasoning alternatives. We apply Dual Tuning across spatial, mathematical, and multi-disciplinary tasks, and further analyze how reinforcement learning and thinking patterns affect reasoning efficacy. The Dual Tuning results guide data curation by identifying data that benefit reasoning training, data better suited to direct-answer training, and data that are detrimental under both training modes. Our work provides quantitative criteria for selecting appropriate training data and matching post-training strategies.

cs.CL↗