arXiv subjects

Bo Wang

Publications and source records attributed to Bo Wang.

At least 253 records · Page 14Linked to original sources

arXiv2025

Benchmarking and Evaluation of AI Models in Biology: Outcomes and Recommendations from the CZI Virtual Cells Workshop

Artificial intelligence holds immense promise for transforming biology, yet a lack of standardized, cross domain, benchmarks undermines our ability to build robust, trustworthy models. Here, we present insights from a recent workshop that convened machine learning and computational biology experts across imaging, transcriptomics, proteomics, and genomics to tackle this gap. We identify major technical and systemic bottlenecks such as data heterogeneity and noise, reproducibility challenges, biases, and the fragmented ecosystem of publicly available resources and propose a set of recommendations for building benchmarking frameworks that can efficiently compare ML models of biological systems across tasks and data modalities. By promoting high quality data curation, standardized tooling, comprehensive evaluation metrics, and open, collaborative platforms, we aim to accelerate the development of robust benchmarks for AI driven Virtual Cells. These benchmarks are crucial for ensuring rigor, reproducibility, and biological relevance, and will ultimately advance the field toward integrated models that drive new discoveries, therapeutic insights, and a deeper understanding of cellular systems.

cs.LG↗

arXiv2025

Towards Applying Large Language Models to Complement Single-Cell Foundation Models

Single-cell foundation models such as scGPT represent a significant advancement in single-cell omics, with an ability to achieve state-of-the-art performance on various downstream biological tasks. However, these models are inherently limited in that a vast amount of information in biology exists as text, which they are unable to leverage. There have therefore been several recent works that propose the use of LLMs as an alternative to single-cell foundation models, achieving competitive results. However, there is little understanding of what factors drive this performance, along with a strong focus on using LLMs as an alternative, rather than complementary approach to single-cell foundation models. In this study, we therefore investigate what biological insights contribute toward the performance of LLMs when applied to single-cell data, and introduce scMPT; a model which leverages synergies between scGPT, and single-cell representations from LLMs that capture these insights. scMPT demonstrates stronger, more consistent performance than either of its component models, which frequently have large performance gaps between each other across datasets. We also experiment with alternate fusion methods, demonstrating the potential of combining specialized reasoning models with scGPT to improve performance. This study ultimately showcases the potential for LLMs to complement single-cell foundation models and drive improvements in single-cell analysis.

cs.LG↗

arXiv2025

Rattan: An Extensible and Scalable Modular Internet Path Emulator

The rapid growth of Internet paths in heterogeneity, scale, and dynamics has made existing emulators increasingly insufficient in flexibility, scalability, and usability. To address these limitations, we present Rattan, an extensible and scalable software network path emulator for modern Internet conditions. Rattan's core innovation lies in its cell-based architecture: by splitting emulation functions into modular "cells" with well-documented asynchronous interfaces, users are allowed to easily compose different cells by hierarchically linking them and easily construct new cells by using standard cell interfaces. This design enables: (1) scalability, supporting hundreds of concurrent gigabit-level paths on a single machine and cluster-level experiments composed of multiple machines; (2) extensibility, simulating new network conditions by constructing new cells. Rattan empowers developers and researchers to efficiently and confidently evaluate, validate, and diagnose diverse network transport innovations for online services.

cs.NI↗

arXiv2025

Late Chunking: Contextual Chunk Embeddings Using Long-Context Embedding Models

Many use cases require retrieving smaller portions of text, and dense vector-based retrieval systems often perform better with shorter text segments, as the semantics are less likely to be over-compressed in the embeddings. Consequently, practitioners often split text documents into smaller chunks and encode them separately. However, chunk embeddings created in this way can lose contextual information from surrounding chunks, resulting in sub-optimal representations. In this paper, we introduce a novel method called late chunking, which leverages long context embedding models to first embed all tokens of the long text, with chunking applied after the transformer model and just before mean pooling - hence the term late in its naming. The resulting chunk embeddings capture the full contextual information, leading to superior results across various retrieval tasks. The method is generic enough to be applied to a wide range of long-context embedding models and works without additional training. To further increase the effectiveness of late chunking, we propose a dedicated fine-tuning approach for embedding models.

cs.CL↗

arXiv2025

jina-embeddings-v4: Universal Embeddings for Multimodal Multilingual Retrieval

We introduce jina-embeddings-v4, a 3.8 billion parameter multimodal embedding model that unifies text and image representations through a novel architecture supporting both single-vector and multi-vector embeddings in the late interaction style. The model incorporates task-specific Low-Rank Adaptation (LoRA) adapters to optimize performance across diverse retrieval scenarios, including query-document retrieval, semantic text similarity, and code search. Comprehensive evaluations demonstrate that jina-embeddings-v4 achieves state-of-the-art performance on both single-modal and cross-modal retrieval tasks, with particular strength in processing visually rich content such as tables, charts, diagrams, and mixed-media formats. To facilitate evaluation of this capability, we also introduce Jina-VDR, a novel benchmark specifically designed for visually rich image retrieval.

cs.AI↗

arXiv2025

Implicit Reward as the Bridge: A Unified View of SFT and DPO Connections

Post-training processes are essential phases in grounding pre-trained language models to real-world tasks, with learning from demonstrations or preference signals playing a crucial role in this adaptation. We present a unified theoretical framework bridging Supervised Fine-Tuning (SFT) and preference learning in Large Language Model (LLM) post-training. Through rigorous mathematical derivation, we demonstrate that both SFT and preference learning methods like Direct Preference Optimization (DPO) operate within the same optimal policy-reward subspace, with SFT representing a special case of implicit reward learning. Our analysis reveals a critical limitation in conventional SFT: the KL divergence term in distribution matching becomes constant with respect to the policy during optimization, failing to constrain model updates. To address this, we propose a simple yet effective learning rate reduction approach that yields significant performance improvements (up to \textbf{25\%} relative gain and \textbf{6\%} absolute win rate increase in instruction following tasks. Additionally, we derive alternative SFT objectives from various f-divergence functions that preserve the KL term during optimization, further enhancing post-DPO model performance. Finally, we extend the theoretical relationship between LLM logits and Q-functions from preference learning to the SFT context, providing mathematical derivations and experimental validation.

cs.LG↗

arXiv2025

UnitModule: A Lightweight Joint Image Enhancement Module for Underwater Object Detection

Underwater object detection faces the problem of underwater image degradation, which affects the performance of the detector. Underwater object detection methods based on noise reduction and image enhancement usually do not provide images preferred by the detector or require additional datasets. In this paper, we propose a plug-and-play \textbf{U}nderwater joi\textbf{n}t \textbf{i}mage enhancemen\textbf{t} \textbf{Module} (UnitModule) that provides the input image preferred by the detector. We design an unsupervised learning loss for the joint training of UnitModule with the detector without additional datasets to improve the interaction between UnitModule and the detector. Furthermore, a color cast predictor with the assisting color cast loss and a data augmentation called Underwater Color Random Transfer (UCRT) are designed to improve the performance of UnitModule on underwater images with different color casts. Extensive experiments are conducted on DUO for different object detection models, where UnitModule achieves the highest performance improvement of 2.6 AP for YOLOv5-S and gains the improvement of 3.3 AP on the brand-new test set ($\text{URPC}_{test}$). And UnitModule significantly improves the performance of all object detection models we test, especially for models with a small number of parameters. In addition, UnitModule with a small number of parameters of 31K has little effect on the inference speed of the original object detection model. Our quantitative and visual analysis also demonstrates the effectiveness of UnitModule in enhancing the input image and improving the perception ability of the detector for object features. The code is available at https://github.com/LEFTeyex/UnitModule.

cs.CV↗

arXiv2025

ShareCMP: Polarization-Aware RGB-P Semantic Segmentation

Multimodal semantic segmentation is developing rapidly, but the modality of RGB-\textbf{P}olarization remains underexplored. To delve into this problem, we construct a UPLight RGB-P segmentation benchmark with 12 typical underwater semantic classes. In this work, we design the ShareCMP, an RGB-P semantic segmentation framework with a shared dual-branch architecture (ShareCMP Encoder), which reduces the parameters and memory space by about 33.8\% compared to previous dual-branch models. It encompasses a Polarization Generate Attention (PGA) module designed to generate polarization modal images with richer polarization properties for the encoder. In addition, we introduce the Class Polarization-Aware Loss (CPALoss) with Class Polarization-Aware Auxiliary Head (CPAAHead) to improve the learning and understanding of the encoder for polarization modal information and to optimize the PGA module. With extensive experiments on a total of three RGB-P benchmarks, our ShareCMP achieves the best performance in mIoU with fewer parameters on the UPLight (92.45{\small (+0.32)}\%), ZJU (92.7{\small (+0.1)}\%), and MCubeS (50.99{\small (+1.51)}\%) datasets. And our ShareCMP (w/o PGA) achieves competitive or even higher performance on other RGB-X datasets compared to the corresponding state-of-the-art RGB-X methods. The code and datasets are available at https://github.com/LEFTeyex/ShareCMP.

cs.CV↗

arXiv2025

Commander-GPT: Fully Unleashing the Sarcasm Detection Capability of Multi-Modal Large Language Models

Sarcasm detection, as a crucial research direction in the field of Natural Language Processing (NLP), has attracted widespread attention. Traditional sarcasm detection tasks have typically focused on single-modal approaches (e.g., text), but due to the implicit and subtle nature of sarcasm, such methods often fail to yield satisfactory results. In recent years, researchers have shifted the focus of sarcasm detection to multi-modal approaches. However, effectively leveraging multi-modal information to accurately identify sarcastic content remains a challenge that warrants further exploration. Leveraging the powerful integrated processing capabilities of Multi-Modal Large Language Models (MLLMs) for various information sources, we propose an innovative multi-modal Commander-GPT framework. Inspired by military strategy, we first decompose the sarcasm detection task into six distinct sub-tasks. A central commander (decision-maker) then assigns the best-suited large language model to address each specific sub-task. Ultimately, the detection results from each model are aggregated to identify sarcasm. We conducted extensive experiments on MMSD and MMSD 2.0, utilizing four multi-modal large language models and six prompting strategies. Our experiments demonstrate that our approach achieves state-of-the-art performance, with a 19.3% improvement in F1 score, without necessitating fine-tuning or ground-truth rationales.

cs.CL↗

arXiv2025

PI-WAN: A Physics-Informed Wind-Adaptive Network for Quadrotor Dynamics Prediction in Unknown Environments

Accurate dynamics modeling is essential for quadrotors to achieve precise trajectory tracking in various applications. Traditional physical knowledge-driven modeling methods face substantial limitations in unknown environments characterized by variable payloads, wind disturbances, and external perturbations. On the other hand, data-driven modeling methods suffer from poor generalization when handling out-of-distribution (OoD) data, restricting their effectiveness in unknown scenarios. To address these challenges, we introduce the Physics-Informed Wind-Adaptive Network (PI-WAN), which combines knowledge-driven and data-driven modeling methods by embedding physical constraints directly into the training process for robust quadrotor dynamics learning. Specifically, PI-WAN employs a Temporal Convolutional Network (TCN) architecture that efficiently captures temporal dependencies from historical flight data, while a physics-informed loss function applies physical principles to improve model generalization and robustness across previously unseen conditions. By incorporating real-time prediction results into a model predictive control (MPC) framework, we achieve improvements in closed-loop tracking performance. Comprehensive simulations and real-world flight experiments demonstrate that our approach outperforms baseline methods in terms of prediction accuracy, tracking precision, and robustness to unknown environments.

cs.RO↗

arXiv2025

LightBSR: Towards Lightweight Blind Super-Resolution via Discriminative Implicit Degradation Representation Learning

Implicit degradation estimation-based blind super-resolution (IDE-BSR) hinges on extracting the implicit degradation representation (IDR) of the LR image and adapting it to LR image features to guide HR detail restoration. Although IDE-BSR has shown potential in dealing with noise interference and complex degradations, existing methods ignore the importance of IDR discriminability for BSR and instead over-complicate the adaptation process to improve effect, resulting in a significant increase in the model's parameters and computations. In this paper, we focus on the discriminability optimization of IDR and propose a new powerful and lightweight BSR model termed LightBSR. Specifically, we employ a knowledge distillation-based learning framework. We first introduce a well-designed degradation-prior-constrained contrastive learning technique during teacher stage to make the model more focused on distinguishing different degradation types. Then we utilize a feature alignment technique to transfer the degradation-related knowledge acquired by the teacher to the student for practical inferencing. Extensive experiments demonstrate the effectiveness of IDR discriminability-driven BSR model design. The proposed LightBSR can achieve outstanding performance with minimal complexity across a range of blind SR tasks. Our code is accessible at: https://github.com/MJ-NCEPU/LightBSR.

cs.CV↗

arXiv2025

Diversity by Design: Addressing Mode Collapse Improves scRNA-seq Perturbation Modeling on Well-Calibrated Metrics

Recent benchmarks reveal that models for single-cell perturbation response are often outperformed by simply predicting the dataset mean. We trace this anomaly to a metric artifact: control-referenced deltas and unweighted error metrics reward mode collapse whenever the control is biased or the biological signal is sparse. Large-scale \textit{in silico} simulations and analysis of two real-world perturbation datasets confirm that shared reference shifts, not genuine biological change, drives high performance in these evaluations. We introduce differentially expressed gene (DEG)-aware metrics, weighted mean-squared error (WMSE) and weighted delta $R^{2}$ ($R^{2}_{w}(Δ)$) with respect to all perturbations, that measure error in niche signals with high sensitivity. We further introduce negative and positive performance baselines to calibrate these metrics. With these improvements, the mean baseline sinks to null performance while genuine predictors are correctly rewarded. Finally, we show that using WMSE as a loss function reduces mode collapse and improves model performance.

q-bio.GN↗

arXiv2025

Transformer-based RGB-T Tracking with Channel and Spatial Feature Fusion

The main problem in RGB-T tracking is the correct and optimal merging of the cross-modal features of visible and thermal images. Some previous methods either do not fully exploit the potential of RGB and TIR information for channel and spatial feature fusion or lack a direct interaction between the template and the search area, which limits the model's ability to fully utilize the original semantic information of both modalities. To address these limitations, we investigate how to achieve a direct fusion of cross-modal channels and spatial features in RGB-T tracking and propose CSTNet. It uses the Vision Transformer (ViT) as the backbone and adds a Joint Spatial and Channel Fusion Module (JSCFM) and Spatial Fusion Module (SFM) integrated between the transformer blocks to facilitate cross-modal feature interaction. The JSCFM module achieves joint modeling of channel and multi-level spatial features. The SFM module includes a cross-attention-like architecture for cross modeling and joint learning of RGB and TIR features. Comprehensive experiments show that CSTNet achieves state-of-the-art performance. To enhance practicality, we retrain the model without JSCFM and SFM modules and use CSNet as the pretraining weight, and propose CSTNet-small, which achieves 50% speedup with an average decrease of 1-2% in SR and PR performance. CSTNet and CSTNet-small achieve real-time speeds of 21 fps and 33 fps on the Nvidia Jetson Xavier, meeting actual deployment requirements. Code is available at https://github.com/LiYunfengLYF/CSTNet.

cs.CV↗

arXiv2025

NTIRE 2025 Image Shadow Removal Challenge Report

This work examines the findings of the NTIRE 2025 Shadow Removal Challenge. A total of 306 participants have registered, with 17 teams successfully submitting their solutions during the final evaluation phase. Following the last two editions, this challenge had two evaluation tracks: one focusing on reconstruction fidelity and the other on visual perception through a user study. Both tracks were evaluated with images from the WSRD+ dataset, simulating interactions between self- and cast-shadows with a large number of diverse objects, textures, and materials.

cs.CV↗

arXiv2025

Knowledge Bridger: Towards Training-free Missing Modality Completion

Previous successful approaches to missing modality completion rely on carefully designed fusion techniques and extensive pre-training on complete data, which can limit their generalizability in out-of-domain (OOD) scenarios. In this study, we pose a new challenge: can we develop a missing modality completion model that is both resource-efficient and robust to OOD generalization? To address this, we present a training-free framework for missing modality completion that leverages large multimodal models (LMMs). Our approach, termed the "Knowledge Bridger", is modality-agnostic and integrates generation and ranking of missing modalities. By defining domain-specific priors, our method automatically extracts structured information from available modalities to construct knowledge graphs. These extracted graphs connect the missing modality generation and ranking modules through the LMM, resulting in high-quality imputations of missing modalities. Experimental results across both general and medical domains show that our approach consistently outperforms competing methods, including in OOD generalization. Additionally, our knowledge-driven generation and ranking techniques demonstrate superiority over variants that directly employ LMMs for generation and ranking, offering insights that may be valuable for applications in other domains.

cs.LG↗

arXiv2025

Decoding Knowledge Attribution in Mixture-of-Experts: A Framework of Basic-Refinement Collaboration and Efficiency Analysis

The interpretability of Mixture-of-Experts (MoE) models, especially those with heterogeneous designs, remains underexplored. Existing attribution methods for dense models fail to capture dynamic routing-expert interactions in sparse MoE architectures. To address this issue, we propose a cross-level attribution algorithm to analyze sparse MoE architectures (Qwen 1.5-MoE, OLMoE, Mixtral-8x7B) against dense models (Qwen 1.5-7B, Llama-7B, Mistral-7B). Results show MoE models achieve 37% higher per-layer efficiency via a "mid-activation, late-amplification" pattern: early layers screen experts, while late layers refine knowledge collaboratively. Ablation studies reveal a "basic-refinement" framework--shared experts handle general tasks (entity recognition), while routed experts specialize in domain-specific processing (geographic attributes). Semantic-driven routing is evidenced by strong correlations between attention heads and experts (r=0.68), enabling task-aware coordination. Notably, architectural depth dictates robustness: deep Qwen 1.5-MoE mitigates expert failures (e.g., 43% MRR drop in geographic tasks when blocking top-10 experts) through shared expert redundancy, whereas shallow OLMoE suffers severe degradation (76% drop). Task sensitivity further guides design: core-sensitive tasks (geography) require concentrated expertise, while distributed-tolerant tasks (object attributes) leverage broader participation. These insights advance MoE interpretability, offering principles to balance efficiency, specialization, and robustness.

cs.CL↗

arXiv2025

GenBreak: Red Teaming Text-to-Image Generators Using Large Language Models

Text-to-image (T2I) models such as Stable Diffusion have advanced rapidly and are now widely used in content creation. However, these models can be misused to generate harmful content, including nudity or violence, posing significant safety risks. While most platforms employ content moderation systems, underlying vulnerabilities can still be exploited by determined adversaries. Recent research on red-teaming and adversarial attacks against T2I models has notable limitations: some studies successfully generate highly toxic images but use adversarial prompts that are easily detected and blocked by safety filters, while others focus on bypassing safety mechanisms but fail to produce genuinely harmful outputs, neglecting the discovery of truly high-risk prompts. Consequently, there remains a lack of reliable tools for evaluating the safety of defended T2I models. To address this gap, we propose GenBreak, a framework that fine-tunes a red-team large language model (LLM) to systematically explore underlying vulnerabilities in T2I generators. Our approach combines supervised fine-tuning on curated datasets with reinforcement learning via interaction with a surrogate T2I model. By integrating multiple reward signals, we guide the LLM to craft adversarial prompts that enhance both evasion capability and image toxicity, while maintaining semantic coherence and diversity. These prompts demonstrate strong effectiveness in black-box attacks against commercial T2I generators, revealing practical and concerning safety weaknesses.

cs.CR↗

arXiv2025

STI-SNN: A 0.14 GOPS/W/PE Single-Timestep Inference FPGA-based SNN Accelerator with Algorithm and Hardware Co-Design

Brain-inspired Spiking Neural Networks (SNNs) have attracted attention for their event-driven characteristics and high energy efficiency. However, the temporal dependency and irregularity of spikes present significant challenges for hardware parallel processing and data reuse, leading to some existing accelerators falling short in processing latency and energy efficiency. To overcome these challenges, we introduce the STI-SNN accelerator, designed for resource-constrained applications with high energy efficiency, flexibility, and low latency. The accelerator is designed through algorithm and hardware co-design. Firstly, STI-SNN can perform inference in a single timestep. At the algorithm level, we introduce a temporal pruning approach based on the temporal efficient training (TET) loss function. This approach alleviates spike disappearance during timestep reduction, maintains inference accuracy, and expands TET's application. In hardware design, we analyze data access patterns and adopt the output stationary (OS) dataflow, eliminating the need to store membrane potentials and access memory operations. Furthermore, based on the OS dataflow, we propose a compressed and sorted representation of spikes, then cached in the line buffer to reduce the memory access cost and improve reuse efficiency. Secondly, STI-SNN supports different convolution methods. By adjusting the computation mode of processing elements (PEs) and parameterizing the computation array, STI-SNN can accommodate lightweight models based on depthwise separable convolutions (DSCs), further enhancing hardware flexibility. Lastly, STI-SNN also supports both inter-layer and intra-layer parallel processing. For inter-layer parallelism, we ...

cs.AR↗