Jiawei Zheng

Xin Zhang, Jianyang Xu, Hao Peng et al.

Knowledge distillation transfers knowledge from large teacher models to smaller students for efficient inference. While existing methods primarily focus on distillation strategies, they often overlook the importance of enhancing teacher knowledge quality. In this paper, we propose Text-guided Multi-view Knowledge Distillation (TMKD), which leverages dual-modality teachers, a visual teacher and a text teacher (CLIP), to provide richer supervisory signals. Specifically, we enhance the visual teacher with multi-view inputs incorporating visual priors (edge and high-frequency features), while the text teacher generates semantic weights through prior-aware prompts to guide adaptive feature fusion. Additionally, we introduce vision-language contrastive regularization to strengthen semantic knowledge in the student model. Extensive experiments on five benchmarks demonstrate that TMKD consistently improves knowledge distillation performance by up to 4.49\%, validating the effectiveness of our dual-teacher multi-view enhancement strategy. Code is available at https://anonymous.4open.science/r/TMKD-main-44D1.

Xudong Li, Mengdan Zhang, Peixian Chen et al.

Spatial intelligence remains a persistent challenge for Multimodal Large Language Models (MLLMs), as it requires coherent spatial scene representations beyond basic object recognition. Existing methods typically build such representations through textual reasoning or 3D reconstruction. However, they often falter during multi-step reasoning, particularly when required to dynamically re-anchor evidence to the specific camera-, object-, or direction-centric reference frames demanded by complex queries. To address this, we propose OmniView-Space, a framework designed to maintain spatial consistency through multimodal egocentric evidence. Our approach consists of three core components: (1) Multi-Perspective Spatial Mapping (MPSM), which re-anchors reconstructed geometry into a query-aligned visual cognitive map and a textual spatial graph; (2) Tool-Guided Egocentric Reasoning, an interleaved policy trained to actively select the ego anchor required by the query and request the corresponding MPSM evidence; and (3) Cognitive-Map Distillation, which uses MPSM-generated trajectories and ego-frame rewards to train the model to reason with self-generated cognitive maps. Experiments on single- and multi-image spatial reasoning benchmarks show that OmniView-Space achieves state-of-the-art performance. Furthermore, the distilled model maintains this performance while reducing reliance on external geometry pipelines.

Jingyuan Zheng, Dongjing Wang, Xin Zhang et al.

Reusable skill libraries are becoming important infrastructure for large language model (LLM) agents, yet existing selection methods often treat skills as retrievable documents and return fixed top-k lists. This paper presents SkillSelect-Serve, a budget-controllable and QoS-aware framework that formulates agent skill selection as Skill Service Recommendation and Composition. SkillSelect-Serve represents raw skills as structured Skill Services with functional descriptions, dependencies, context cost, risk, and QoS-related attributes. A local Micro-Agent Requirement Planner converts natural-language tasks into structured service requirements, while a shared discovery backbone retrieves candidate services from a large registry. The framework then performs dual-granularity utility modeling with skill-level marginal suitability estimation and bundle-level calibration for coverage, redundancy, cost, and risk trade-offs. Experiments on 35,353 skills and 586 task queries show that SkillSelect-Serve consistently improves same-budget bundle recall and mean utility over fixed top-k retrieval baselines.