Kimi K2.5: Visual Agentic Intelligence

We introduce Kimi K2.5, an open-source multimodal agentic model designed to advance general agentic intelligence. K2.5 emphasizes the joint optimization of text and vision so that two modalities enhance each other. This includes a series of techniques such as joint text-vision pre-training, zero-vision SFT, and joint text-vision reinforcement learning. Building on this multimodal foundation, K2.5 introduces Agent Swarm, a self-directed parallel agent orchestration framework that dynamically decomposes complex tasks into heterogeneous sub-problems and executes them concurrently. Extensive evaluations show that Kimi K2.5 achieves state-of-the-art results across various domains including coding, vision, reasoning, and agentic tasks. Agent Swarm also reduces latency by up to $4.5\times$ over single-agent baselines. We release the post-trained Kimi K2.5 model checkpoint to facilitate future research and real-world applications of agentic intelligence.

Self-Improving Pretraining: using post-trained models to pretrain better models

Ensuring safety, factuality and overall quality in the generations of large language models is a critical challenge, especially as these models are increasingly deployed in real-world applications. The prevailing approach to addressing these issues involves collecting expensive, carefully curated datasets and applying multiple stages of fine-tuning and alignment. However, even this complex pipeline cannot guarantee the correction of patterns learned during pretraining. Therefore, addressing these issues during pretraining is crucial, as it shapes a model's core behaviors and prevents unsafe or hallucinated outputs from becoming deeply embedded. To tackle this issue, we introduce a new pretraining method that streams documents and uses reinforcement learning (RL) to improve the next K generated tokens at each step. A strong, post-trained model judges candidate generations -- including model rollouts, the original suffix, and a rewritten suffix -- for quality, safety, and factuality. Early in training, the process relies on the original and rewritten suffixes; as the model improves, RL rewards high-quality rollouts. This approach builds higher quality, safer, and more factual models from the ground up. In experiments, our method gives 36.2% and 18.5% relative improvements over standard pretraining in terms of factuality and safety, and up to 86.3% win rate improvements in overall generation quality.

MiLorE-SSL: Scaling Multilingual Capabilities in Self-Supervised Models without Forgetting

Self-supervised learning (SSL) has greatly advanced speech representation learning, but multilingual SSL models remain constrained to languages encountered during pretraining. Retraining from scratch to incorporate new languages is computationally expensive, while sequential training without migitation strategies often leads to catastrophic forgetting. To address this, we propose MiLorE-SSL, a lightweight framework that combines LoRA modules with a soft mixture-of-experts (MoE) mechanism for efficient continual multilingual training. LoRA provides efficient low-rank adaptation, while soft MoE promotes flexible expert sharing across languages, reducing cross-lingual interference. To further mitigate forgetting, we introduce limited replay data from existing languages, avoiding reliance on large historical corpora. Experiments on ML-SUPERB demonstrate that MiLorE-SSL achieves strong performance in new languages and improves the ability in existing ones with only 2.14% trainable parameters.

PROST-LLM: Progressively Enhancing the Speech-to-Speech Translation Capability in LLMs

Although Large Language Models (LLMs) excel in many tasks, their application to Speech-to-Speech Translation (S2ST) is underexplored and hindered by data scarcity. To bridge this gap, we propose PROST-LLM (PROgressive Speech-to-speech Translation) to enhance the S2ST capabilities in LLMs progressively. First, we fine-tune the LLMs with the CVSS corpus, employing designed tri-task learning and chain of modality methods to boost the initial performance. Then, leveraging the fine-tuned model, we generate preference pairs through self-sampling and back-translation without human evaluation. Finally, these preference pairs are used for preference optimization to enhance the model's S2ST capability further. Extensive experiments confirm the effectiveness of our proposed PROST-LLM in improving the S2ST capability of LLMs.

A Quantum-Driven Evolutionary Framework for Solving High-Dimensional Sharpe Ratio Portfolio Optimization

High-dimensional portfolio optimization faces significant computational challenges under complex constraints, with traditional optimization methods struggling to balance convergence speed and global exploration capability. To address this, firstly, we introduce an enhanced Sharpe ratio-based model that incorporates all constraints into the objective function using adaptive penalty terms, transforming the original constrained problem into an unconstrained single-objective formulation. This approach preserves financial interpretability while simplifying algorithmic implementation. To efficiently solve the resulting high-dimensional optimization problem, we propose a Quantum Hybrid Differential Evolution (QHDE) algorithm, which integrates Quantum-inspired probabilistic behavior into the standard DE framework. QHDE employs a Schrodinger-inspired probabilistic mechanism for population evolution, enabling more flexible and diversified solution updates. To further enhance performance, a good point set-chaos reverse learning strategy is adopted to generate a well-dispersed initial population, and a dynamic elite pool combined with Cauchy-Gaussian hybrid perturbations strengthens global exploration and mitigates premature convergence. Experimental validation on CEC benchmarks and real-world portfolios involving 20 to 80 assets demonstrates that QHDE's performance improves by up to 73.4%. It attains faster convergence, higher solution precision, and greater robustness than seven state-of-the-art counterparts, thereby confirming its suitability for complex, high-dimensional portfolio optimization and advancing quantum-inspired evolutionary research in computational finance.

RealCamo: Boosting Real Camouflage Synthesis with Layout Controls and Textual-Visual Guidance

Camouflaged image generation (CIG) has recently emerged as an efficient alternative for acquiring high-quality training data for camouflaged object detection (COD). However, existing CIG methods still suffer from a substantial gap to real camouflaged imagery: generated images either lack sufficient camouflage due to weak visual similarity, or exhibit cluttered backgrounds that are semantically inconsistent with foreground targets. To address these limitations, we propose ReamCamo, a unified out-painting based framework for realistic camouflaged image generation. ReamCamo explicitly introduces additional layout controls to regulate global image structure, thereby improving semantic coherence between foreground objects and generated backgrounds. Moreover, we construct a multi-modal textual-visual condition by combining a unified fine-grained textual task description with texture-oriented background retrieval, which jointly guides the generation process to enhance visual fidelity and realism. To quantitatively assess camouflage quality, we further introduce a background-foreground distribution divergence metric that measures the effectiveness of camouflage in generated images. Extensive experiments and visualizations demonstrate the effectiveness of our proposed framework.

Embodied Tree of Thoughts: Deliberate Manipulation Planning with Embodied World Model

World models have emerged as a pivotal component in robot manipulation planning, enabling agents to predict future environmental states and reason about the consequences of actions before execution. While video-generation models are increasingly adopted, they often lack rigorous physical grounding, leading to hallucinations and a failure to maintain consistency in long-horizon physical constraints. To address these limitations, we propose Embodied Tree of Thoughts (EToT), a novel Real2Sim2Real planning framework that leverages a physics-based interactive digital twin as an embodied world model. EToT formulates manipulation planning as a tree search expanded through two synergistic mechanisms: (1) Priori Branching, which generates diverse candidate execution paths based on semantic and spatial analysis; and (2) Reflective Branching, which utilizes VLMs to diagnose execution failures within the simulator and iteratively refine the planning tree with corrective actions. By grounding high-level reasoning in a physics simulator, our framework ensures that generated plans adhere to rigid-body dynamics and collision constraints. We validate EToT on a suite of short- and long-horizon manipulation tasks, where it consistently outperforms baselines by effectively predicting physical dynamics and adapting to potential failures. Website at https://embodied-tree-of-thoughts.github.io .

A Quantum Tunneling and Bio-Phototactic Driven Enhanced Dwarf Mongoose Optimizer for UAV Trajectory Planning and Engineering Problem

With the widespread adoption of unmanned aerial vehicles (UAV), effective path planning has become increasingly important. Although traditional search methods have been extensively applied, metaheuristic algorithms have gained popularity due to their efficiency and problem-specific heuristics. However, challenges such as premature convergence and lack of solution diversity still hinder their performance in complex scenarios. To address these issues, this paper proposes an Enhanced Multi-Strategy Dwarf Mongoose Optimization (EDMO) algorithm, tailored for three-dimensional UAV trajectory planning in dynamic and obstacle-rich environments. EDMO integrates three novel strategies: (1) a Dynamic Quantum Tunneling Optimization Strategy (DQTOS) to enable particles to probabilistically escape local optima; (2) a Bio-phototactic Dynamic Focusing Search Strategy (BDFSS) inspired by microbial phototaxis for adaptive local refinement; and (3) an Orthogonal Lens Opposition-Based Learning (OLOBL) strategy to enhance global exploration through structured dimensional recombination. EDMO is benchmarked on 39 standard test functions from CEC2017 and CEC2020, outperforming 14 advanced algorithms in convergence speed, robustness, and optimization accuracy. Furthermore, real-world validations on UAV three-dimensional path planning and three engineering design tasks confirm its practical applicability and effectiveness in field robotics missions requiring intelligent, adaptive, and time-efficient planning.

ViTaMIn-B: A Reliable and Efficient Visuo-Tactile Bimanual Manipulation Interface

Handheld devices have opened up unprecedented opportunities to collect large-scale, high-quality demonstrations efficiently. However, existing systems often lack robust tactile sensing or reliable pose tracking to handle complex interaction scenarios, especially for bimanual and contact-rich tasks. In this work, we propose ViTaMIn-B, a more capable and efficient handheld data collection system for such tasks. We first design DuoTact, a novel compliant visuo-tactile sensor built with a flexible frame to withstand large contact forces during manipulation while capturing high-resolution contact geometry. To enhance the cross-sensor generalizability, we propose reconstructing the sensor's global deformation as a 3D point cloud and using it as the policy input. We further develop a robust, unified 6-DoF bimanual pose acquisition process using Meta Quest controllers, which eliminates the trajectory drift issue in common SLAM-based methods. Comprehensive user studies confirm the efficiency and high usability of ViTaMIn-B among novice and expert operators. Furthermore, experiments on four bimanual manipulation tasks demonstrate its superior task performance relative to existing systems.

Virtual-Real Collaborated Split Learning via Model Partitioning in IRS-Assisted IoT Networks

This paper investigates a novel computation and communication co-design framework for large-scale split learning in intelligent reflecting surface (IRS)-assisted internet of things (IoT) networks integrated with digital twin (DT) technique. The considered system consists of a multi-antenna access point (AP), multiple heterogeneous user devices (UDs), and an deployed IRS to enhance both uplink and downlink transmission. The training process of a deep neural network is partitioned between devices and the AP, where a DT replica is activated to replace UDs with insufficient local computation capabilities. We formulate a delay-optimal split learning problem, which optimizes five key variables: layer partitioning points, DT assignment decisions, IRS phase shift matrix, AP downlink power allocation, and DT frequency adjustment, aiming to minimize the overall end-to-end delay under communication and computation. The proposed optimization problem is a highly coupled non-convex mixed-integer problem. Therefore, we solve using an alternating optimization approach combining closed-form updates, semidefinite relaxation (SDR), and low-complexity heuristics. Extensive simulations demonstrate that the proposed scheme significantly reduces training delay compared to conventional baselines and achieves up to 35\% delay improvement, especially under high UD density and stringent power constraints.