arXiv subjects

Fangyuan Li

Publications and source records attributed to Fangyuan Li.

16 recordsLinked to original sources

arXiv2026

Cycle Based Computational Pipeline for Extracting Instantaneous Whisking Frequency

Whisking is a rhythmic and adaptive behavior that rodents use to probe and interact with their environment, and the frequency of movement reflects both sensorimotor processing and internal brain states. A robust and traditional method of whisker frequency estimation uses power spectral analysis of whisker position spanning several cycles. To improve the temporal resolution of whisker movement, we here estimate the period for each cycle, hence indirectly extracting an instantaneous frequency. We do this using markerless estimation of whisker position and identifying the peak and trough for each cycle. The cycle period is extracted, and artifacts are rejected with a ripple exclusion validator based on peak prominence and sequential amplitude filtering. The method is compared with power spectral estimation, using the Fourier transform of a temporal window of 0.5 seconds. We find that frequency estimation using a fixed window does not capture transient variability, while the cycle by cycle method recovers higher, time-resolved frequencies. The cycle by cycle approach also reveals the expected cycle-level variability. Artifact rejection through subsequence filtering removed spurious frequencies above 30 Hz, aligning refined frequencies with established physiological bounds (4 to 28 Hz). This pipeline provides an alternative solution for real time compatible frequency estimation, which better captures temporal variation at the expense of precision in frequency estimation.

q-bio.QM↗

arXiv2026

MARS$^2$: Scaling Multi-Agent Tree Search via Reinforcement Learning for Code Generation

Reinforcement learning (RL) paradigms have demonstrated strong performance on reasoning-intensive tasks such as code generation. However, limited trajectory diversity often leads to diminishing returns, which constrains the achievable performance ceiling. Search-enhanced RL alleviates this issue by introducing structured exploration, which remains constrained by the single-agent policy priors. Meanwhile, leveraging multiple interacting policies can acquire more diverse exploratory signals, but existing approaches are typically decoupled from structured search. We propose \textbf{MARS$^2$} (Multi-Agent Reinforced Tree-Search Scaling), a unified RL framework in which multiple independently-optimized agents collaborate within a shared tree-structured search environment. MARS$^2$ models the search tree as a learnable multi-agent interaction environment, enabling heterogeneous agents to collaboratively generate and refine candidate solutions within a shared search topology. To support effective learning, we introduce a path-level group advantage formulation based on tree-consistent reward shaping, which facilitates effective credit assignment across complex search trajectories. Experiments on code generation benchmarks show that MARS$^2$ consistently improves performance across diverse model combinations and training settings, demonstrating the effectiveness of coupling multi-agent collaboration with tree search for enhancing reinforcement learning. Our code is publicly available at https://github.com/TsinghuaC3I/MARTI.

cs.AI↗

arXiv2026

WIST: Web-Grounded Iterative Self-Play Tree for Domain-Targeted Reasoning Improvement

Recent progress in reinforcement learning with verifiable rewards (RLVR) offers a practical path to self-improvement of language models, but existing methods face a key trade-off: endogenous self-play can drift over iterations, while corpus-grounded approaches rely on curated data environments. We present \textbf{WIST}, a \textbf{W}eb-grounded \textbf{I}terative \textbf{S}elf-play \textbf{T}ree framework for domain-targeted reasoning improvement that learns directly from the open web without requiring any pre-arranged domain corpus. WIST incrementally expands a domain tree for exploration, and retrieves and cleans path-consistent web corpus to construct a controllable training environment. It then performs Challenger--Solver self-play with verifiable rewards, and feeds learnability signals back to update node posteriors and guide subsequent exploration through an adaptive curriculum. Across four backbones, WIST consistently improves over the base models and typically outperforms both purely endogenous self-evolution and corpus-grounded self-play baselines, with the Overall gains reaching \textbf{+9.8} (\textit{Qwen3-4B-Base}) and \textbf{+9.7} (\textit{OctoThinker-8B}). WIST is also domain-steerable, improving \textit{Qwen3-8B-Base} by \textbf{+14.79} in medicine and \textit{Qwen3-4B-Base} by \textbf{+5.28} on PhyBench. Ablations further confirm the importance of WIST's key components for stable open-web learning. Our Code is available at https://github.com/lfy-123/WIST.

cs.LG↗

arXiv2026

MARTI-MARS$^2$: Scaling Multi-Agent Self-Search via Reinforcement Learning for Code Generation

While the complex reasoning capability of Large Language Models (LLMs) has attracted significant attention, single-agent systems often encounter inherent performance ceilings in complex tasks such as code generation. Multi-agent collaboration offers a promising avenue to transcend these boundaries. However, existing frameworks typically rely on prompt-based test-time interactions or multi-role configurations trained with homogeneous parameters, limiting error correction capabilities and strategic diversity. In this paper, we propose a Multi-Agent Reinforced Training and Inference Framework with Self-Search Scaling (MARTI-MARS2), which integrates policy learning with multi-agent tree search by formulating the multi-agent collaborative exploration process as a dynamic and learnable environment. By allowing agents to iteratively explore and refine within the environment, the framework facilitates evolution from parameter-sharing homogeneous multi-role training to heterogeneous multi-agent training, breaking through single-agent capability limits. We also introduce an efficient inference strategy MARTI-MARS2-T+ to fully exploit the scaling potential of multi-agent collaboration at test time. We conduct extensive experiments across varied model scales (8B, 14B, and 32B) on challenging code generation benchmarks. Utilizing two collaborating 32B models, MARTI-MARS2 achieves 77.7%, outperforming strong baselines like GPT-5.1. Furthermore, MARTI-MARS2 reveals a novel scaling law: shifting from single-agent to homogeneous multi-role and ultimately to heterogeneous multi-agent paradigms progressively yields higher RL performance ceilings, robust TTS capabilities, and greater policy diversity, suggesting that policy diversity is critical for scaling intelligence via multi-agent reinforcement learning.

cs.LG↗

arXiv2026

TF-UNet: Resolving Complex Speckles for Single-Shot Reconstruction of 512^2-Matrix Images Using a Micron-Sized Optical Fiber

Tapered optical fibers (TFs), with diameters gradually reduced from hundreds of microns to the micron scale, offer key advantages over conventional flat optical fibers (FFs), including uniform illumination, efficient long-range signal collection, and minimal invasiveness for applications in high-sensitivity biosensing, optogenetics, and photodynamic therapy. However, high-fidelity, single-shot imaging through a single TF remains underexplored due to intermodal coupling from the tapering geometry, which distorts output speckle patterns and poses challenges for image reconstruction using existing deep learning methods. Here, we propose a physics-inspired TF-UNet architecture that augments skip connections with hierarchical grouped-MLP fusion to effectively capture non-local, cross-scale dependencies caused by intermodal coupling in TFs. We experimentally validate our method on both FFs and TFs, demonstrating that TF-UNet outperforms standard U-Net variants in structural and perceptual fidelity while maintaining competitive PSNR at quadratic complexity. Our study offers a promising approach for deep learning-based imaging through micron-sized, ultrafine optical fibers, enabling scanning-free single-shot reconstruction on a 512x512 reconstruction matrix, and further validating the framework on biologically meaningful neuronal and vascular datasets for physically interpretable characterization.

physics.optics↗

arXiv2025

Integrated Ring-based Quantum Key Distribution with Weak Measurement Enhanced Fiber-Optic Sensing Disturbance Magnitude and Location

The deep integration of quantum communication and fiber-optic sensing is pivotal for the development of next-generation multifunctional and highly reliable secure information infrastructure. Here, we present a Sagnac-loop integrated system (SLIS) that, for the first time, combines ring-based quantum key distribution (QKD) with fiber-based weak measurement (WM) enhanced sensing and disturbance localization capabilities. In the event of communication interruption due to external disturbances, the SLIS seamlessly switches to perception system, employing interference measurement and WM techniques to monitor channel disturbances. By integrating null-frequencies localization (NFL) mode, the system precisely determines the disturbance location, enabling rapid identification of security vulnerabilities along the link. Experimental results demonstrate that, over a 30 km Sagnac loop channel, the SLIS achieves a raw key generation rate of 22.4 kbps with stable operation and clear scalability toward network expansion. In terms of perception performance, the SLIS exhibits strong capability for both dynamic and quasi-static disturbances. For dynamic perturbations, the system detects transient impacts and PZT-driven frequency variations down to 100 Hz, and enables long-distance localization via NFL alignment, with improved localization performance as the disturbance position moves farther away along the loop. For quasi-static disturbances, gravitational changes as small as 100 g are resolved, corresponding to a time-delay variation of 9.81 as. This work provides a novel technical pathway toward self-diagnosing, robust quantum networks through integrated communication and sensing functionalities.

quant-ph↗

arXiv2025

SmartBench: Is Your LLM Truly a Good Chinese Smartphone Assistant?

Large Language Models (LLMs) have become integral to daily life, especially advancing as intelligent assistants through on-device deployment on smartphones. However, existing LLM evaluation benchmarks predominantly focus on objective tasks like mathematics and coding in English, which do not necessarily reflect the practical use cases of on-device LLMs in real-world mobile scenarios, especially for Chinese users. To address these gaps, we introduce SmartBench, the first benchmark designed to evaluate the capabilities of on-device LLMs in Chinese mobile contexts. We analyze functionalities provided by representative smartphone manufacturers and divide them into five categories: text summarization, text Q&A, information extraction, content creation, and notification management, further detailed into 20 specific tasks. For each task, we construct high-quality datasets comprising 50 to 200 question-answer pairs that reflect everyday mobile interactions, and we develop automated evaluation criteria tailored for these tasks. We conduct comprehensive evaluations of on-device LLMs and MLLMs using SmartBench and also assess their performance after quantized deployment on real smartphone NPUs. Our contributions provide a standardized framework for evaluating on-device LLMs in Chinese, promoting further development and optimization in this critical area. Code and data will be available at https://github.com/vivo-ai-lab/SmartBench.

cs.CL↗

arXiv2025

EdgeInfinite-Instruct: Bridging SFT-Based Optimization and NPU-Level Efficiency for Edge Devices

Deploying Transformer-based large language models (LLMs) on resource-constrained edge devices for long-sequence tasks remains challenging due to the quadratic time complexity of self-attention and growing Key-Value (KV) cache demands. While existing KV cache optimizations improve memory efficiency, they often fail to reduce time to first token (TTFT) and may degrade performance through token pruning. Alternative sequence modeling architectures address some of these limitations, but typically require full retraining and lack infrastructure support. EdgeInfinite offers an efficient solution by fine-tuning only a small subset of parameters, maintaining quality while reducing both computational and memory costs, including improved TTFT. However, its instruction-following ability is limited, and it lacks mobile-specific optimizations. To address these issues, we propose EdgeInfinite-Instruct, which introduces a Segmented Supervised Fine-Tuning (S-SFT) strategy tailored to long-sequence tasks such as summarization and question answering. We further optimized EdgeInfinite-Instruct for efficient deployment on edge NPUs by employing fine-grained post-training quantization (PTQ) to reduce computational demands while maintaining accuracy, and by implementing a fixed-shape computation graph that balances memory usage and on-device efficiency through scenario-specific customization of input token and cache sizes. Experiments on long-context benchmarks and real-world mobile tasks show that our approach improves domain-specific performance while maintaining efficiency on NPU-accelerated edge devices.

cs.CL↗

arXiv2025

BlueLM-2.5-3B Technical Report

We present BlueLM-2.5-3B, a compact and unified dense Multimodal Large Language Model (MLLM) designed for efficient edge-device deployment, offering strong general-purpose and reasoning capabilities. To the best of our knowledge, this is the first 3B-scale MLLM to support both thinking and non-thinking modes, while also enabling explicit control over thinking token budget. BlueLM-2.5-3B is developed through diversified data curation, key data resampling, hybrid heterogeneous reinforcement learning, and a high-performance training infrastructure. Our model achieves superior multimodal capacity while preserving competitive pure-text performance with only 2.9 billion parameters. We conduct comprehensive evaluations across a broad range of multimodal and text-only benchmarks. In thinking mode, BlueLM-2.5-3B achieves comparable performance to Qwen3-4B on text-only benchmarks, and trails the larger Kimi-VL-A3B-16B by only about 5% on average across multimodal evaluations. In non-thinking mode, it outperforms Qwen2.5-VL-3B on the majority of multimodal benchmarks. Additionally, BlueLM-2.5-3B exhibits exceptional data efficiency. All of the aforementioned performance is achieved with substantially less total training data than Qwen2.5-VL-3B and Qwen3-4B. We hope our work contributes to the advancement of high-performance, on-device MLLMs and provides meaningful insights to the research community.

cs.AI↗

arXiv2025

London Blue Light Collaboration Evaluation: A Comparative Analysis of Spatio temporal Patterns on Emergency Services by London Ambulance Service and London Fire Brigade

With rising demand for emergency services, the London Ambulance Service, LAS, and the London Fire Brigade, LFB, face growing challenges in resource coordination. This study investigates the temporal and spatial similarities in their service demands to assess potential for routine cross-agency collaboration. Time series analysis revealed aligned demand peaks in summer, on Fridays, during daytime hours, and were highly sensitive to high temperature weather conditions. Bivariate mapping and Moran I indicated significant spatial overlaps in central London and Hillingdon. Geographically Weighted Regression, GWR, examined the influence of socioeconomic factors, while Comap analysis uncovered spatiotemporal heterogeneity across fire service types. The findings highlight opportunities for targeted collaboration in high-overlap areas and peak periods, offering practical insights to enhance emergency service resilience and efficiency.

cs.CY↗

arXiv2025

EdgeInfinite: A Memory-Efficient Infinite-Context Transformer for Edge Devices

Transformer-based large language models (LLMs) encounter challenges in processing long sequences on edge devices due to the quadratic complexity of attention mechanisms and growing memory demands from Key-Value (KV) cache. Existing KV cache optimizations struggle with irreversible token eviction in long-output tasks, while alternative sequence modeling architectures prove costly to adopt within established Transformer infrastructure. We present EdgeInfinite, a memory-efficient solution for infinite contexts that integrates compressed memory into Transformer-based LLMs through a trainable memory-gating module. This approach maintains full compatibility with standard Transformer architectures, requiring fine-tuning only a small part of parameters, and enables selective activation of the memory-gating module for long and short context task routing. The experimental result shows that EdgeInfinite achieves comparable performance to baseline Transformer-based LLM on long context benchmarks while optimizing memory consumption and time to first token.

cs.CL↗

arXiv2025

GenieBlue: Integrating both Linguistic and Multimodal Capabilities for Large Language Models on Mobile Devices

Recent advancements in Multimodal Large Language Models (MLLMs) have enabled their deployment on mobile devices. However, challenges persist in maintaining strong language capabilities and ensuring hardware compatibility, both of which are crucial for user experience and practical deployment efficiency. In our deployment process, we observe that existing MLLMs often face performance degradation on pure language tasks, and the current NPU platforms on smartphones do not support the MoE architecture, which is commonly used to preserve pure language capabilities during multimodal training. To address these issues, we systematically analyze methods to maintain pure language capabilities during the training of MLLMs, focusing on both training data and model architecture aspects. Based on these analyses, we propose GenieBlue, an efficient MLLM structural design that integrates both linguistic and multimodal capabilities for LLMs on mobile devices. GenieBlue freezes the original LLM parameters during MLLM training to maintain pure language capabilities. It acquires multimodal capabilities by duplicating specific transformer blocks for full fine-tuning and integrating lightweight LoRA modules. This approach preserves language capabilities while achieving comparable multimodal performance through extensive training. Deployed on smartphone NPUs, GenieBlue demonstrates efficiency and practicality for applications on mobile devices.

cs.CL↗

arXiv2024

Fast and Slow Gradient Approximation for Binary Neural Network Optimization

Binary Neural Networks (BNNs) have garnered significant attention due to their immense potential for deployment on edge devices. However, the non-differentiability of the quantization function poses a challenge for the optimization of BNNs, as its derivative cannot be backpropagated. To address this issue, hypernetwork based methods, which utilize neural networks to learn the gradients of non-differentiable quantization functions, have emerged as a promising approach due to their adaptive learning capabilities to reduce estimation errors. However, existing hypernetwork based methods typically rely solely on current gradient information, neglecting the influence of historical gradients. This oversight can lead to accumulated gradient errors when calculating gradient momentum during optimization. To incorporate historical gradient information, we design a Historical Gradient Storage (HGS) module, which models the historical gradient sequence to generate the first-order momentum required for optimization. To further enhance gradient generation in hypernetworks, we propose a Fast and Slow Gradient Generation (FSG) method. Additionally, to produce more precise gradients, we introduce Layer Recognition Embeddings (LRE) into the hypernetwork, facilitating the generation of layer-specific fine gradients. Extensive comparative experiments on the CIFAR-10 and CIFAR-100 datasets demonstrate that our method achieves faster convergence and lower loss values, outperforming existing baselines.Code is available at http://github.com/two-tiger/FSG .

cs.LG↗

arXiv2024

Less is More: Efficient Model Merging with Binary Task Switch

As an effective approach to equip models with multi-task capabilities without additional training, model merging has garnered significant attention. However, existing methods face challenges of redundant parameter conflicts and the excessive storage burden of parameters. In this work, through controlled experiments, we reveal that for task vectors, only those parameters with magnitudes above a certain threshold contribute positively to the task, exhibiting a pulse-like characteristic. We then attempt leveraging this characteristic to binarize the task vectors and reduce storage overhead. Further controlled experiments show that the binarized task vectors incur almost no decrease in fine-tuning and merging performance, and even exhibit stronger performance improvements as the proportion of redundant parameters increases. Based on these insights, we propose Task Switch (T-Switch), which decomposes task vectors into three components: 1) an activation switch instantiated by a binarized mask vector, 2) a polarity switch instantiated by a binarized sign vector, and 3) a scaling knob instantiated by a scalar coefficient. By storing task vectors in a binarized form, T-Switch alleviates parameter conflicts while ensuring efficient task parameter storage. Furthermore, to enable automated switch combination in T-Switch, we further introduce Auto-Switch, which enables training-free switch combination via retrieval from a small query set. Experiments indicate that our methods achieve significant performance improvements over existing baselines, requiring only 1-3% of the storage space of full-precision parameters.

cs.LG↗

arXiv2024

Online DPO: Online Direct Preference Optimization with Fast-Slow Chasing

Direct Preference Optimization (DPO) improves the alignment of large language models (LLMs) with human values by training directly on human preference datasets, eliminating the need for reward models. However, due to the presence of cross-domain human preferences, direct continual training can lead to catastrophic forgetting, limiting DPO's performance and efficiency. Inspired by intraspecific competition driving species evolution, we propose a Online Fast-Slow chasing DPO (OFS-DPO) for preference alignment, simulating competition through fast and slow chasing among models to facilitate rapid adaptation. Specifically, we first derive the regret upper bound for online learning, validating our motivation with a min-max optimization pattern. Based on this, we introduce two identical modules using Low-rank Adaptive (LoRA) with different optimization speeds to simulate intraspecific competition, and propose a new regularization term to guide their learning. To further mitigate catastrophic forgetting in cross-domain scenarios, we extend the OFS-DPO with LoRA modules combination strategy, resulting in the Cross domain Online Fast-Slow chasing DPO (COFS-DPO). This method leverages linear combinations of fast modules parameters from different task domains, fully utilizing historical information to achive continual value alignment. Experimental results show that OFS-DPO outperforms DPO in in-domain alignment, while COFS-DPO excels in cross-domain continual learning scenarios.

cs.AI↗

arXiv2018

Closed-Loop Hierarchical Operation for Optimal Unit Commitment and Dispatch in Microgrids: A Hybrid System Approach

To contribute to a smart grid, the unit commitment and dispatch (UCD) is a very important problem to be revisited in the future microgrid environment. In this paper, the UCD problem is studied by taking account of both traditional thermal units and distributed generation units (DGs) while considering demand response (DR) and carbon emissions trading (CET). Then, the UCD problem is reformulated as the optimal switching problem of a hybrid system. The reformulation allows convenient handling of time-dependent start-up cost for dynamic programming (DP). Afterwards, a closed-loop hierarchical operation (CLHO) algorithm is developed which is capable of producing the optimal schedule in the closed-loop behavior. Theoretical analysis is also presented to show the equivalence of the reformulation and the effectiveness of the CLHO algorithm. Finally, the closed-loop behavior and the effectiveness of the algorithm are further verified by simulation studies. It is also shown that low emission quotas and high emission trading price help reduce the total carbon emissions.

math.OC↗