Foundation model (FM) powered agent services are regarded as a promising solution to develop intelligent and personalized applications for advancing toward Artificial General Intelligence (AGI). To achieve high reliability and scalability in deploying these agent services, it is essential to collaboratively optimize computational and communication resources, thereby ensuring effective resource allocation and seamless service delivery. In pursuit of this vision, this paper proposes a unified framework aimed at providing a comprehensive survey on deploying FM-based agent services across heterogeneous devices, with the emphasis on the integration of model and resource optimization to establish a robust infrastructure for these services. Particularly, this paper begins with exploring various low-level optimization strategies during inference and studies approaches that enhance system scalability, such as parallelism techniques and resource scaling methods. The paper then discusses several prominent FMs and investigates research efforts focused on inference acceleration, including techniques such as model compression and token reduction. Moreover, the paper also investigates critical components for constructing agent services and highlights notable intelligent applications. Finally, the paper presents potential research directions for developing real-time agent services with high Quality of Service (QoS).

Experience Goal Visual Rearrangement task stands as a However, these methods have disadvantages: 2D and 3D foundational challenge within Embodied AI, requiring an semantic maps store object pose and semantic information agent to construct a robust world model that accurately in a grid; this approach provides limited resolution, does captures the goal state. The agent uses this world model to not inherently capture interactions between objects and is restore a shuffled scene to its original configuration, making prone to sensitivity issues and quantization errors. Although an accurate representation of the world essential for pointcloud based representation can provide more robustness successfully completing the task. In this work, we present to sensitivity, it lacks structural semantics: identifying a novel framework that leverages on 3D Gaussian Splatting objects and their interactions with the world in a noisy as a 3D scene representation for experience goal visual pointcloud is challenging. Scene graph based methods often rearrangement task. Recent advances in volumetric assume a clear and well defined relationship between scene representation like 3D Gaussian Splatting, offer fast objects, which often limits the granularity of scene understanding, rendering of high quality and photo-realistic novel views.

Multi-hop question answering (QA) often requires sequential retrieval (multi-hop retrieval), where each hop retrieves missing knowledge based on information from previous hops. To facilitate more effective retrieval, we aim to distill knowledge from a posterior retrieval, which has access to posterior information like an answer, into a prior retrieval used during inference when such information is unavailable. Unfortunately, current methods for knowledge distillation in one-time retrieval are ineffective for multi-hop QA due to two issues: 1) Posterior information is often defined as the response (i.e. the answer), which may not clearly connect to the query without intermediate retrieval; and 2) The large knowledge gap between prior and posterior retrievals makes existing distillation methods unstable, even resulting in performance loss. As such, we propose MoPo (Momentum Posterior Regularization) with two key innovations: 1) Posterior information of one hop is defined as a query-focus summary from the golden knowledge of the previous and current hops; 2) We develop an effective training strategy where the posterior retrieval is updated along with the prior retrieval via momentum moving average method, allowing smoother and effective distillation. Experiments on HotpotQA and StrategyQA demonstrate that MoPo outperforms existing baselines in both retrieval and downstream QA tasks.

Scientific question answering (SQA) is an important task aimed at answering questions based on papers. However, current SQA datasets have limited reasoning types and neglect the relevance between tables and text, creating a significant gap with real scenarios. To address these challenges, we propose a QA benchmark for scientific tables and text with diverse reasoning types (SciTaT). To cover more reasoning types, we summarize various reasoning types from real-world questions. To involve both tables and text, we require the questions to incorporate tables and text as much as possible. Based on SciTaT, we propose a strong baseline (CaR), which combines various reasoning methods to address different reasoning types and process tables and text at the same time. CaR brings average improvements of 12.9% over other baselines on SciTaT, validating its effectiveness. Error analysis reveals the challenges of SciTaT, such as complex numerical calculations and domain knowledge.

A crucial step in cohort studies is to extract the required cohort from one or more study datasets. This step is time-consuming, especially when a researcher is presented with a dataset that they have not previously worked with. When the cohort has to be extracted from multiple datasets, cohort extraction can be extremely laborious. In this study, we present an approach for partially automating cohort extraction from multiple electronic health record (EHR) databases. We formulate the guided multi-dataset cohort extraction problem in which selection criteria are first converted into queries, translating them from natural language text to language that maps to database entities. Then, using FLMs, columns of interest identified from the queries are automatically matched between the study databases. Finally, the generated queries are run across all databases to extract the study cohort. We propose and evaluate an algorithm for automating column matching on two large, popular and publicly-accessible EHR databases -- MIMIC-III and eICU. Our approach achieves a high top-three accuracy of $92\%$, correctly matching $12$ out of the $13$ columns of interest, when using a small, pre-trained general purpose language model. Furthermore, this accuracy is maintained even as the search space (i.e., size of the database) increases.

Recent advancements in 3D editing have highlighted the potential of text-driven methods in real-time, user-friendly AR/VR applications. However, current methods rely on 2D diffusion models without adequately considering multi-view information, resulting in multi-view inconsistency. While 3D Gaussian Splatting (3DGS) significantly improves rendering quality and speed, its 3D editing process encounters difficulties with inefficient optimization, as pre-trained Gaussians retain excessive source information, hindering optimization. To address these limitations, we propose \textbf{EditSplat}, a novel 3D editing framework that integrates Multi-view Fusion Guidance (MFG) and Attention-Guided Trimming (AGT). Our MFG ensures multi-view consistency by incorporating essential multi-view information into the diffusion process, leveraging classifier-free guidance from the text-to-image diffusion model and the geometric properties of 3DGS. Additionally, our AGT leverages the explicit representation of 3DGS to selectively prune and optimize 3D Gaussians, enhancing optimization efficiency and enabling precise, semantically rich local edits. Through extensive qualitative and quantitative evaluations, EditSplat achieves superior multi-view consistency and editing quality over existing methods, significantly enhancing overall efficiency.

Constructing precise global maps is a key task in robotics and is required for localization, surveying, monitoring, or constructing digital twins. To build accurate maps, data from mobile 3D LiDAR sensors is often used. Mapping requires correctly aligning the individual point clouds to each other to obtain a globally consistent map. In this paper, we investigate the problem of multi-scan alignment to obtain globally consistent point cloud maps. We propose a 3D LiDAR bundle adjustment approach to solve the global alignment problem and jointly optimize the available data. Utilizing a continuous-time trajectory allows us to consider the ego-motion of the LiDAR scanner while recording a single scan directly in the least squares adjustment. Furthermore, pruning the search space of correspondences and utilizing out-of-core circular buffer enables our approach to align thousands of point clouds efficiently. We successfully align point clouds recorded with a handheld LiDAR, as well as ones mounted on a vehicle, and are able to perform multi-session alignment.

Vision-language models (VLMs) have shown remarkable advancements in multimodal reasoning tasks. However, they still often generate inaccurate or irrelevant responses due to issues like hallucinated image understandings or unrefined reasoning paths. To address these challenges, we introduce Critic-V, a novel framework inspired by the Actor-Critic paradigm to boost the reasoning capability of VLMs. This framework decouples the reasoning process and critic process by integrating two independent components: the Reasoner, which generates reasoning paths based on visual and textual inputs, and the Critic, which provides constructive critique to refine these paths. In this approach, the Reasoner generates reasoning responses according to text prompts, which can evolve iteratively as a policy based on feedback from the Critic. This interaction process was theoretically driven by a reinforcement learning framework where the Critic offers natural language critiques instead of scalar rewards, enabling more nuanced feedback to boost the Reasoner's capability on complex reasoning tasks. The Critic model is trained using Direct Preference Optimization (DPO), leveraging a preference dataset of critiques ranked by Rule-based Reward~(RBR) to enhance its critic capabilities. Evaluation results show that the Critic-V framework significantly outperforms existing methods, including GPT-4V, on 5 out of 8 benchmarks, especially regarding reasoning accuracy and efficiency. Combining a dynamic text-based policy for the Reasoner and constructive feedback from the preference-optimized Critic enables a more reliable and context-sensitive multimodal reasoning process. Our approach provides a promising solution to enhance the reliability of VLMs, improving their performance in real-world reasoning-heavy multimodal applications such as autonomous driving and embodied intelligence.

The advanced processing and reasoning capabilities of multimodal large language models (MLLMs) have driven substantial progress in vision-language (VL) understanding tasks. However, while effective for tasks governed by straightforward logic, MLLMs often encounter challenges when reasoning over complex, interdependent logic structures. To address this limitation, we introduce \textit{AgentPS}, a novel framework that integrates Agentic Process Supervision into MLLMs via multi-round question answering during fine-tuning. \textit{AgentPS} demonstrates significant performance improvements over baseline MLLMs on proprietary TikTok datasets, due to its integration of process supervision and structured sequential reasoning. Furthermore, we show that replacing human-annotated labels with LLM-generated labels retains much of the performance gain, highlighting the framework's practical scalability in industrial applications. These results position \textit{AgentPS} as a highly effective and efficient architecture for multimodal classification tasks. Its adaptability and scalability, especially when enhanced by automated annotation generation, make it a powerful tool for handling large-scale, real-world challenges.

Chain-of-thought (CoT) reasoning has emerged as an effective approach for activating latent capabilities in large language models (LLMs). We observe that CoT shares significant similarities with self-training in terms of their learning processes. Motivated by these parallels, this paper explores the underlying relationship between CoT and self-training, demonstrating how insights from self-training can enhance CoT performance. Specifically, our study first reveals that CoT, like self-training, follows the principle of semantic entropy minimization. Leveraging this insight, we propose a novel CoT framework that incorporates two key components: (i) a task-specific prompt module designed to guide LLMs in generating high-quality initial reasoning processes, and (ii) an adaptive reasoning iteration module for progressively refining the reasoning process.