The Mixture-of-Experts (MoE) technique has proven to be a promising solution to efficiently scale the model size, which has been widely applied in recent LLM advancements. However, the substantial memory overhead of MoE models has made their compression an important research direction. In this work, we provide a theoretical analysis of expert merging, a recently proposed technique for compressing MoE models. Rather than interpreting expert merging from the conventional perspective of parameter aggregation, we approach it from the perspective of merging experts' outputs. Our key insight is that the merging process can be interpreted as inserting additional matrices into the forward computation, which naturally leads to an optimization formulation. Building on this analysis, we introduce MergeMoE, a method that leverages mathematical optimization to construct the compression matrices. We evaluate MergeMoE on multiple MoE models and show that our algorithm consistently outperforms the baselines with the same compression ratios.

Although large language models (LLMs) have made significant strides across various tasks, they still face significant challenges in complex reasoning and planning. For example, even with carefully designed prompts and prior information explicitly provided, GPT-4o achieves only a 7% Final Pass Rate on the TravelPlanner dataset in the sole-planning mode. Similarly, even in the thinking mode, Qwen3-8B-Instruct and DeepSeek-R1-671B, only achieve Final Pass Rates of 5.9% and 40%, respectively. Although well-organized Multi-Agent Systems (MAS) can offer improved collective reasoning, they often suffer from high reasoning costs due to multi-round internal interactions, long per-response latency, and difficulties in end-to-end training. To address these challenges, we propose a general and scalable framework called IMAGINE, short for Integrating Multi-Agent System into One Model. This framework not only integrates the reasoning and planning capabilities of MAS into a single, compact model, but also significantly surpass the capabilities of the MAS through a simple end-to-end training. Through this pipeline, a single small-scale model is not only able to acquire the structured reasoning and planning capabilities of a well-organized MAS but can also significantly outperform it. Experimental results demonstrate that, when using Qwen3-8B-Instruct as the base model and training it with our method, the model achieves an 82.7% Final Pass Rate on the TravelPlanner benchmark, far exceeding the 40% of DeepSeek-R1-671B, while maintaining a much smaller model size.

Large language model (LLM) inference systems face a fundamental tension between minimizing Time-to-First-Token (TTFT) latency for new requests and maintaining a high, steady token generation rate (low Time-Per-Output-Token, or TPOT) for ongoing requests. Existing stall-free batching schedulers proposed by Sarathi, while effective at preventing decode stalls, introduce significant computational unfairness. They prioritize decode tasks excessively, simultaneously leading to underutilized decode slack and unnecessary prefill queuing delays, which collectively degrade the system's overall quality of service (QoS). This work identifies the root cause of this unfairness: the non-monotonic nature of Time-Between-Tokens (TBT) as a scheduling metric and the rigid decode-prioritizing policy that fails to adapt to dynamic workload bursts. We therefore propose FairBatching, a novel LLM inference scheduler that enforces fair resource allocation between prefill and decode tasks. It features an adaptive batch capacity determination mechanism, which dynamically adjusts the computational budget to improve the GPU utilization without triggering SLO violations. Its fair and dynamic batch formation algorithm breaks away from the decode-prioritizing paradigm, allowing computation resources to be reclaimed from bursting decode tasks to serve prefill surges, achieving global fairness. Furthermore, FairBatching provides a novel load estimation method, enabling more effective coordination with upper-level schedulers. Implemented and evaluated on realistic traces, FairBatching significantly reduces TTFT tail latency by up to 2.29x while robustly maintaining TPOT SLOs, achieving overall 20.0% improvement in single-node capacity and 54.3% improvement in cluster-level capacity.

Motherboard defect detection is critical for ensuring reliability in high-volume electronics manufacturing. While prior research in PCB inspection has largely targeted bare-board or trace-level defects, assembly-level inspection of full motherboards inspection remains underexplored. In this work, we present BoardVision, a reproducible framework for detecting assembly-level defects such as missing screws, loose fan wiring, and surface scratches. We benchmark two representative detectors - YOLOv7 and Faster R-CNN, under controlled conditions on the MiracleFactory motherboard dataset, providing the first systematic comparison in this domain. To mitigate the limitations of single models, where YOLO excels in precision but underperforms in recall and Faster R-CNN shows the reverse, we propose a lightweight ensemble, Confidence-Temporal Voting (CTV Voter), that balances precision and recall through interpretable rules. We further evaluate robustness under realistic perturbations including sharpness, brightness, and orientation changes, highlighting stability challenges often overlooked in motherboard defect detection. Finally, we release a deployable GUI-driven inspection tool that bridges research evaluation with operator usability. Together, these contributions demonstrate how computer vision techniques can transition from benchmark results to practical quality assurance for assembly-level motherboard manufacturing.

Building agents that autonomously operate mobile devices has attracted increasing attention. While Vision-Language Models (VLMs) show promise, most existing approaches rely on direct state-to-action mappings, which lack structured reasoning and planning, and thus generalize poorly to novel tasks or unseen UI layouts. We introduce Hi-Agent, a trainable hierarchical vision-language agent for mobile control, featuring a high-level reasoning model and a low-level action model that are jointly optimized. For efficient training, we reformulate multi-step decision-making as a sequence of single-step subgoals and propose a foresight advantage function, which leverages execution feedback from the low-level model to guide high-level optimization. This design alleviates the path explosion issue encountered by Group Relative Policy Optimization (GRPO) in long-horizon tasks and enables stable, critic-free joint training. Hi-Agent achieves a new State-Of-The-Art (SOTA) 87.9% task success rate on the Android-in-the-Wild (AitW) benchmark, significantly outperforming prior methods across three paradigms: prompt-based (AppAgent: 17.7%), supervised (Filtered BC: 54.5%), and reinforcement learning-based (DigiRL: 71.9%). It also demonstrates competitive zero-shot generalization on the ScreenSpot-v2 benchmark. On the more challenging AndroidWorld benchmark, Hi-Agent also scales effectively with larger backbones, showing strong adaptability in high-complexity mobile control scenarios.

Math reasoning has been one crucial ability of large language models (LLMs), where significant advancements have been achieved in recent years. However, most efforts focus on LLMs by curating high-quality annotation data and intricate training (or inference) paradigms, while the math reasoning performance of multi-modal LLMs (MLLMs) remains lagging behind. Since the MLLM typically consists of an LLM and a vision block, we wonder: Can MLLMs directly absorb math reasoning abilities from off-the-shelf math LLMs without tuning? Recent model-merging approaches may offer insights into this question. However, they overlook the alignment between the MLLM and LLM, where we find that there is a large gap between their parameter spaces, resulting in lower performance. Our empirical evidence reveals two key factors behind this issue: the identification of crucial reasoning-associated layers in the model and the mitigation of the gaps in parameter space. Based on the empirical insights, we propose IP-Merging that first identifies the reasoning-associated parameters in both MLLM and Math LLM, then projects them into the subspace of MLLM, aiming to maintain the alignment, and finally merges parameters in this subspace. IP-Merging is a tuning-free approach since parameters are directly adjusted. Extensive experiments demonstrate that our IP-Merging method can enhance the math reasoning ability of MLLMs directly from Math LLMs without compromising their other capabilities.

Recent advances in large language models (LLMs) and retrieval-augmented generation (RAG) have enabled progress on question answering (QA) when relevant evidence is in one (single-hop) or multiple (multi-hop) passages. Yet many realistic questions about recurring report data - medical records, compliance filings, maintenance logs - require aggregation across all documents, with no clear stopping point for retrieval and high sensitivity to even one missed passage. We term these pluri-hop questions and formalize them by three criteria: recall sensitivity, exhaustiveness, and exactness. To study this setting, we introduce PluriHopWIND, a diagnostic multilingual dataset of 48 pluri-hop questions built from 191 real-world wind industry reports in German and English. We show that PluriHopWIND is 8-40% more repetitive than other common datasets and thus has higher density of distractor documents, better reflecting practical challenges of recurring report corpora. We test a traditional RAG pipeline as well as graph-based and multimodal variants, and find that none of the tested approaches exceed 40% in statement-wise F1 score. Motivated by this, we propose PluriHopRAG, a RAG architecture that follows a "check all documents individually, filter cheaply" approach: it (i) decomposes queries into document-level subquestions and (ii) uses a cross-encoder filter to discard irrelevant documents before costly LLM reasoning. We find that PluriHopRAG achieves relative F1 score improvements of 18-52% depending on base LLM. Despite its modest size, PluriHopWIND exposes the limitations of current QA systems on repetitive, distractor-rich corpora. PluriHopRAG's performance highlights the value of exhaustive retrieval and early filtering as a powerful alternative to top-k methods.

High-performing medical Large Language Models (LLMs) typically require extensive fine-tuning with substantial computational resources, limiting accessibility for resource-constrained healthcare institutions. This study introduces a confidence-driven multi-model framework that leverages model diversity to enhance medical question answering without fine-tuning. Our framework employs a two-stage architecture: a confidence detection module assesses the primary model's certainty, and an adaptive routing mechanism directs low-confidence queries to Helper models with complementary knowledge for collaborative reasoning. We evaluate our approach using Qwen3-30B-A3B-Instruct, Phi-4 14B, and Gemma 2 12B across three medical benchmarks; MedQA, MedMCQA, and PubMedQA. Result demonstrate that our framework achieves competitive performance, with particularly strong results in PubMedQA (95.0\%) and MedMCQA (78.0\%). Ablation studies confirm that confidence-aware routing combined with multi-model collaboration substantially outperforms single-model approaches and uniform reasoning strategies. This work establishes that strategic model collaboration offers a practical, computationally efficient pathway to improve medical AI systems, with significant implications for democratizing access to advanced medical AI in resource-limited settings.

Mobile app markets host millions of apps, yet undesired behaviors (e.g., disruptive ads, illegal redirection, payment deception) remain hard to catch because they often do not rely on permission-protected APIs and can be easily camouflaged via UI or metadata edits. We present BINCTX, a learning approach that builds multi-modal representations of an app from (i) a global bytecode-as-image view that captures code-level semantics and family-style patterns, (ii) a contextual view (manifested actions, components, declared permissions, URL/IP constants) indicating how behaviors are triggered, and (iii) a third-party-library usage view summarizing invocation frequencies along inter-component call paths. The three views are embedded and fused to train a contextual-aware classifier. On real-world malware and benign apps, BINCTX attains a macro F1 of 94.73%, outperforming strong baselines by at least 14.92%. It remains robust under commercial obfuscation (F1 84% post-obfuscation) and is more resistant to adversarial samples than state-of-the-art bytecode-only systems.

Mammography, the current standard for breast cancer screening, has reduced sensitivity in women with dense breast tissue, contributing to missed or delayed diagnoses. Thermalytix, an AI-based thermal imaging modality, captures functional vascular and metabolic cues that may complement mammographic structural data. This study investigates whether a breast density-informed multi-modal AI framework can improve cancer detection by dynamically selecting the appropriate imaging modality based on breast tissue composition. A total of 324 women underwent both mammography and thermal imaging. Mammography images were analyzed using a multi-view deep learning model, while Thermalytix assessed thermal images through vascular and thermal radiomics. The proposed framework utilized Mammography AI for fatty breasts and Thermalytix AI for dense breasts, optimizing predictions based on tissue type. This multi-modal AI framework achieved a sensitivity of 94.55% (95% CI: 88.54-100) and specificity of 79.93% (95% CI: 75.14-84.71), outperforming standalone mammography AI (sensitivity 81.82%, specificity 86.25%) and Thermalytix AI (sensitivity 92.73%, specificity 75.46%). Importantly, the sensitivity of Mammography dropped significantly in dense breasts (67.86%) versus fatty breasts (96.30%), whereas Thermalytix AI maintained high and consistent sensitivity in both (92.59% and 92.86%, respectively). This demonstrates that a density-informed multi-modal AI framework can overcome key limitations of unimodal screening and deliver high performance across diverse breast compositions. The proposed framework is interpretable, low-cost, and easily deployable, offering a practical path to improving breast cancer screening outcomes in both high-resource and resource-limited settings.