Faithful Bi-Directional Model Steering via Distribution Matching and Distributed Interchange Interventions

Intervention-based model steering offers a lightweight and interpretable alternative to prompting and fine-tuning. However, by adapting strong optimization objectives from fine-tuning, current methods are susceptible to overfitting and often underperform, sometimes generating unnatural outputs. We hypothesize that this is because effective steering requires the faithful identification of internal model mechanisms, not the enforcement of external preferences. To this end, we build on the principles of distributed alignment search (DAS), the standard for causal variable localization, to propose a new steering method: Concept DAS (CDAS). While we adopt the core mechanism of DAS, distributed interchange intervention (DII), we introduce a novel distribution matching objective tailored for the steering task by aligning intervened output distributions with counterfactual distributions. CDAS differs from prior work in two main ways: first, it learns interventions via weak-supervised distribution matching rather than probability maximization; second, it uses DIIs that naturally enable bi-directional steering and allow steering factors to be derived from data, reducing the effort required for hyperparameter tuning and resulting in more faithful and stable control. On AxBench, a large-scale model steering benchmark, we show that CDAS does not always outperform preference-optimization methods but may benefit more from increased model scale. In two safety-related case studies, overriding refusal behaviors of safety-aligned models and neutralizing a chain-of-thought backdoor, CDAS achieves systematic steering while maintaining general model utility. These results indicate that CDAS is complementary to preference-optimization approaches and conditionally constitutes a robust approach to intervention-based model steering. Our code is available at https://github.com/colored-dye/concept_das.

Ground What You See: Hallucination-Resistant MLLMs via Caption Feedback, Diversity-Aware Sampling, and Conflict Regularization

While Multimodal Large Language Models (MLLMs) have achieved remarkable success across diverse tasks, their practical deployment is severely hindered by hallucination issues, which become particularly acute during Reinforcement Learning (RL) optimization. This paper systematically analyzes the root causes of hallucinations in MLLMs under RL training, identifying three critical factors: (1) an over-reliance on chained visual reasoning, where inaccurate initial descriptions or redundant information anchor subsequent inferences to incorrect premises; (2) insufficient exploration diversity during policy optimization, leading the model to generate overly confident but erroneous outputs; and (3) destructive conflicts between training samples, where Neural Tangent Kernel (NTK) similarity causes false associations and unstable parameter updates. To address these challenges, we propose a comprehensive framework comprising three core modules. First, we enhance visual localization by introducing dedicated planning and captioning stages before the reasoning phase, employing a quality-based caption reward to ensure accurate initial anchoring. Second, to improve exploration, we categorize samples based on the mean and variance of their reward distributions, prioritizing samples with high variance to focus the model on diverse and informative data. Finally, to mitigate sample interference, we regulate NTK similarity by grouping sample pairs and applying an InfoNCE loss to push overly similar pairs apart and pull dissimilar ones closer, thereby guiding gradient interactions toward a balanced range. Experimental results demonstrate that our proposed method significantly reduces hallucination rates and effectively enhances the inference accuracy of MLLMs.

ToolGate: Contract-Grounded and Verified Tool Execution for LLMs

Large Language Models (LLMs) augmented with external tools have demonstrated remarkable capabilities in complex reasoning tasks. However, existing frameworks rely heavily on natural language reasoning to determine when tools can be invoked and whether their results should be committed, lacking formal guarantees for logical safety and verifiability. We present \textbf{ToolGate}, a forward execution framework that provides logical safety guarantees and verifiable state evolution for LLM tool calling. ToolGate maintains an explicit symbolic state space as a typed key-value mapping representing trusted world information throughout the reasoning process. Each tool is formalized as a Hoare-style contract consisting of a precondition and a postcondition, where the precondition gates tool invocation by checking whether the current state satisfies the required conditions, and the postcondition determines whether the tool's result can be committed to update the state through runtime verification. Our approach guarantees that the symbolic state evolves only through verified tool executions, preventing invalid or hallucinated results from corrupting the world representation. Experimental validation demonstrates that ToolGate significantly improves the reliability and verifiability of tool-augmented LLM systems while maintaining competitive performance on complex multi-step reasoning tasks. This work establishes a foundation for building more trustworthy and debuggable AI systems that integrate language models with external tools.

Taming Preference Mode Collapse via Directional Decoupling Alignment in Diffusion Reinforcement Learning

Recent studies have demonstrated significant progress in aligning text-to-image diffusion models with human preference via Reinforcement Learning from Human Feedback. However, while existing methods achieve high scores on automated reward metrics, they often lead to Preference Mode Collapse (PMC)-a specific form of reward hacking where models converge on narrow, high-scoring outputs (e.g., images with monolithic styles or pervasive overexposure), severely degrading generative diversity. In this work, we introduce and quantify this phenomenon, proposing DivGenBench, a novel benchmark designed to measure the extent of PMC. We posit that this collapse is driven by over-optimization along the reward model's inherent biases. Building on this analysis, we propose Directional Decoupling Alignment (D$^2$-Align), a novel framework that mitigates PMC by directionally correcting the reward signal. Specifically, our method first learns a directional correction within the reward model's embedding space while keeping the model frozen. This correction is then applied to the reward signal during the optimization process, preventing the model from collapsing into specific modes and thereby maintaining diversity. Our comprehensive evaluation, combining qualitative analysis with quantitative metrics for both quality and diversity, reveals that D$^2$-Align achieves superior alignment with human preference.

AstraNav-Memory: Contexts Compression for Long Memory

Lifelong embodied navigation requires agents to accumulate, retain, and exploit spatial-semantic experience across tasks, enabling efficient exploration in novel environments and rapid goal reaching in familiar ones. While object-centric memory is interpretable, it depends on detection and reconstruction pipelines that limit robustness and scalability. We propose an image-centric memory framework that achieves long-term implicit memory via an efficient visual context compression module end-to-end coupled with a Qwen2.5-VL-based navigation policy. Built atop a ViT backbone with frozen DINOv3 features and lightweight PixelUnshuffle+Conv blocks, our visual tokenizer supports configurable compression rates; for example, under a representative 16$\times$ compression setting, each image is encoded with about 30 tokens, expanding the effective context capacity from tens to hundreds of images. Experimental results on GOAT-Bench and HM3D-OVON show that our method achieves state-of-the-art navigation performance, improving exploration in unfamiliar environments and shortening paths in familiar ones. Ablation studies further reveal that moderate compression provides the best balance between efficiency and accuracy. These findings position compressed image-centric memory as a practical and scalable interface for lifelong embodied agents, enabling them to reason over long visual histories and navigate with human-like efficiency.

AstraNav-World: World Model for Foresight Control and Consistency

Embodied navigation in open, dynamic environments demands accurate foresight of how the world will evolve and how actions will unfold over time. We propose AstraNav-World, an end-to-end world model that jointly reasons about future visual states and action sequences within a unified probabilistic framework. Our framework integrates a diffusion-based video generator with a vision-language policy, enabling synchronized rollouts where predicted scenes and planned actions are updated simultaneously. Training optimizes two complementary objectives: generating action-conditioned multi-step visual predictions and deriving trajectories conditioned on those predicted visuals. This bidirectional constraint makes visual predictions executable and keeps decisions grounded in physically consistent, task-relevant futures, mitigating cumulative errors common in decoupled "envision-then-plan" pipelines. Experiments across diverse embodied navigation benchmarks show improved trajectory accuracy and higher success rates. Ablations confirm the necessity of tight vision-action coupling and unified training, with either branch removal degrading both prediction quality and policy reliability. In real-world testing, AstraNav-World demonstrated exceptional zero-shot capabilities, adapting to previously unseen scenarios without any real-world fine-tuning. These results suggest that AstraNav-World captures transferable spatial understanding and planning-relevant navigation dynamics, rather than merely overfitting to simulation-specific data distribution. Overall, by unifying foresight vision and control within a single generative model, we move closer to reliable, interpretable, and general-purpose embodied agents that operate robustly in open-ended real-world settings.

Do Not Merge My Model! Safeguarding Open-Source LLMs Against Unauthorized Model Merging

Model merging has emerged as an efficient technique for expanding large language models (LLMs) by integrating specialized expert models. However, it also introduces a new threat: model merging stealing, where free-riders exploit models through unauthorized model merging. Unfortunately, existing defense mechanisms fail to provide effective protection. Specifically, we identify three critical protection properties that existing methods fail to simultaneously satisfy: (1) proactively preventing unauthorized merging; (2) ensuring compatibility with general open-source settings; (3) achieving high security with negligible performance loss. To address the above issues, we propose MergeBarrier, a plug-and-play defense that proactively prevents unauthorized merging. The core design of MergeBarrier is to disrupt the Linear Mode Connectivity (LMC) between the protected model and its homologous counterparts, thereby eliminating the low-loss path required for effective model merging. Extensive experiments show that MergeBarrier effectively prevents model merging stealing with negligible accuracy loss.

Omni-Effects: Unified and Spatially-Controllable Visual Effects Generation

Visual effects (VFX) are essential visual enhancements fundamental to modern cinematic production. Although video generation models offer cost-efficient solutions for VFX production, current methods are constrained by per-effect LoRA training, which limits generation to single effects. This fundamental limitation impedes applications that require spatially controllable composite effects, i.e., the concurrent generation of multiple effects at designated locations. However, integrating diverse effects into a unified framework faces major challenges: interference from effect variations and spatial uncontrollability during multi-VFX joint training. To tackle these challenges, we propose Omni-Effects, a first unified framework capable of generating prompt-guided effects and spatially controllable composite effects. The core of our framework comprises two key innovations: (1) LoRA-based Mixture of Experts (LoRA-MoE), which employs a group of expert LoRAs, integrating diverse effects within a unified model while effectively mitigating cross-task interference. (2) Spatial-Aware Prompt (SAP) incorporates spatial mask information into the text token, enabling precise spatial control. Furthermore, we introduce an Independent-Information Flow (IIF) module integrated within the SAP, isolating the control signals corresponding to individual effects to prevent any unwanted blending. To facilitate this research, we construct a comprehensive VFX dataset Omni-VFX via a novel data collection pipeline combining image editing and First-Last Frame-to-Video (FLF2V) synthesis, and introduce a dedicated VFX evaluation framework for validating model performance. Extensive experiments demonstrate that Omni-Effects achieves precise spatial control and diverse effect generation, enabling users to specify both the category and location of desired effects.

TK-Mamba: Marrying KAN with Mamba for Text-Driven 3D Medical Image Segmentation

3D medical image segmentation is vital for clinical diagnosis and treatment but is challenged by high-dimensional data and complex spatial dependencies. Traditional single-modality networks, such as CNNs and Transformers, are often limited by computational inefficiency and constrained contextual modeling in 3D settings. We introduce a novel multimodal framework that leverages Mamba and Kolmogorov-Arnold Networks (KAN) as an efficient backbone for long-sequence modeling. Our approach features three key innovations: First, an EGSC (Enhanced Gated Spatial Convolution) module captures spatial information when unfolding 3D images into 1D sequences. Second, we extend Group-Rational KAN (GR-KAN), a Kolmogorov-Arnold Networks variant with rational basis functions, into 3D-Group-Rational KAN (3D-GR-KAN) for 3D medical imaging - its first application in this domain - enabling superior feature representation tailored to volumetric data. Third, a dual-branch text-driven strategy leverages CLIP's text embeddings: one branch swaps one-hot labels for semantic vectors to preserve inter-organ semantic relationships, while the other aligns images with detailed organ descriptions to enhance semantic alignment. Experiments on the Medical Segmentation Decathlon (MSD) and KiTS23 datasets show our method achieving state-of-the-art performance, surpassing existing approaches in accuracy and efficiency. This work highlights the power of combining advanced sequence modeling, extended network architectures, and vision-language synergy to push forward 3D medical image segmentation, delivering a scalable solution for clinical use. The source code is openly available at https://github.com/yhy-whu/TK-Mamba.

A General Infrastructure and Workflow for Quadrotor Deep Reinforcement Learning and Reality Deployment

Deploying robot learning methods to a quadrotor in unstructured outdoor environments is an exciting task. Quadrotors operating in real-world environments by learning-based methods encounter several challenges: a large amount of simulator generated data required for training, strict demands for real-time processing onboard, and the sim-to-real gap caused by dynamic and noisy conditions. Current works have made a great breakthrough in applying learning-based methods to end-to-end control of quadrotors, but rarely mention the infrastructure system training from scratch and deploying to reality, which makes it difficult to reproduce methods and applications. To bridge this gap, we propose a platform that enables the seamless transfer of end-to-end deep reinforcement learning (DRL) policies. We integrate the training environment, flight dynamics control, DRL algorithms, the MAVROS middleware stack, and hardware into a comprehensive workflow and architecture that enables quadrotors' policies to be trained from scratch to real-world deployment in several minutes. Our platform provides rich types of environments including hovering, dynamic obstacle avoidance, trajectory tracking, balloon hitting, and planning in unknown environments, as a physical experiment benchmark. Through extensive empirical validation, we demonstrate the efficiency of proposed sim-to-real platform, and robust outdoor flight performance under real-world perturbations. Details can be found from our website https://emnavi.tech/AirGym/.