Locate, Steer, and Improve: A Practical Survey of Actionable Mechanistic Interpretability in Large Language Models

Mechanistic Interpretability (MI) has emerged as a vital approach to demystify the opaque decision-making of Large Language Models (LLMs). However, existing reviews primarily treat MI as an observational science, summarizing analytical insights while lacking a systematic framework for actionable intervention. To bridge this gap, we present a practical survey structured around the pipeline: "Locate, Steer, and Improve." We formally categorize Localizing (diagnosis) and Steering (intervention) methods based on specific Interpretable Objects to establish a rigorous intervention protocol. Furthermore, we demonstrate how this framework enables tangible improvements in Alignment, Capability, and Efficiency, effectively operationalizing MI as an actionable methodology for model optimization. The curated paper list of this work is available at https://github.com/rattlesnakey/Awesome-Actionable-MI-Survey.

A General One-Shot Multimodal Active Perception Framework for Robotic Manipulation: Learning to Predict Optimal Viewpoint

Active perception in vision-based robotic manipulation aims to move the camera toward more informative observation viewpoints, thereby providing high-quality perceptual inputs for downstream tasks. Most existing active perception methods rely on iterative optimization, leading to high time and motion costs, and are tightly coupled with task-specific objectives, which limits their transferability. In this paper, we propose a general one-shot multimodal active perception framework for robotic manipulation. The framework enables direct inference of optimal viewpoints and comprises a data collection pipeline and an optimal viewpoint prediction network. Specifically, the framework decouples viewpoint quality evaluation from the overall architecture, supporting heterogeneous task requirements. Optimal viewpoints are defined through systematic sampling and evaluation of candidate viewpoints, after which large-scale training datasets are constructed via domain randomization. Moreover, a multimodal optimal viewpoint prediction network is developed, leveraging cross-attention to align and fuse multimodal features and directly predict camera pose adjustments. The proposed framework is instantiated in robotic grasping under viewpoint-constrained environments. Experimental results demonstrate that active perception guided by the framework significantly improves grasp success rates. Notably, real-world evaluations achieve nearly double the grasp success rate and enable seamless sim-to-real transfer without additional fine-tuning, demonstrating the effectiveness of the proposed framework.

Domain Adaptation in Structural Health Monitoring of Civil Infrastructure: A Systematic Review

This study provides a comprehensive review of domain adaptation (DA) techniques in vibration-based structural health monitoring (SHM). As data-driven models increasingly support the assessment of civil structures, the persistent challenge of transferring knowledge across varying geometries, materials, and environmental conditions remains a major obstacle. DA offers a systematic approach to mitigate these discrepancies by aligning feature distributions between simulated, laboratory, and field domains while preserving the sensitivity of damage-related information. Drawing on more than sixty representative studies, this paper analyzes the evolution of DA methods for SHM, including statistical alignment, adversarial and subdomain learning, physics-informed adaptation, and generative modeling for simulation-to-real transfer. The review summarizes their contributions and limitations across bridge and building applications, revealing that while DA has improved generalization significantly, key challenges persist: managing domain discrepancy, addressing data scarcity, enhancing model interpretability, and enabling adaptability to multiple sources and time-varying conditions. Future research directions emphasize integrating physical constraints into learning objectives, developing physics-consistent generative frameworks to enhance data realism, establishing interpretable and certifiable DA systems for engineering practice, and advancing multi-source and lifelong adaptation for scalable monitoring. Overall, this review consolidates the methodological foundation of DA for SHM, identifies existing barriers to generalization and trust, and outlines the technological trajectory toward transparent, physics-aware, and adaptive monitoring systems that support the long-term resilience of civil infrastructure.

Anatomical Region-Guided Contrastive Decoding: A Plug-and-Play Strategy for Mitigating Hallucinations in Medical VLMs

Medical Vision-Language Models (MedVLMs) show immense promise in clinical applicability. However, their reliability is hindered by hallucinations, where models often fail to derive answers from visual evidence, instead relying on learned textual priors. Existing mitigation strategies for MedVLMs have distinct limitations: training-based methods rely on costly expert annotations, limiting scalability, while training-free interventions like contrastive decoding, though data-efficient, apply a global, untargeted correction whose effects in complex real-world clinical settings can be unreliable. To address these challenges, we introduce Anatomical Region-Guided Contrastive Decoding (ARCD), a plug-and-play strategy that mitigates hallucinations by providing targeted, region-specific guidance. Our module leverages an anatomical mask to direct a three-tiered contrastive decoding process. By dynamically re-weighting at the token, attention, and logits levels, it verifiably steers the model's focus onto specified regions, reinforcing anatomical understanding and suppressing factually incorrect outputs. Extensive experiments across diverse datasets, including chest X-ray, CT, brain MRI, and ocular ultrasound, demonstrate our method's effectiveness in improving regional understanding, reducing hallucinations, and enhancing overall diagnostic accuracy.

CheXPO-v2: Preference Optimization for Chest X-ray VLMs with Knowledge Graph Consistency

Medical Vision-Language Models (VLMs) are prone to hallucinations, compromising clinical reliability. While reinforcement learning methods like Group Relative Policy Optimization (GRPO) offer a low-cost alignment solution, their reliance on sparse, outcome-based rewards inadvertently encourages models to "overthink" -- generating verbose, convoluted, and unverifiable Chain-of-Thought reasoning to justify answers. This focus on outcomes obscures factual errors and poses significant safety risks. To address this, we propose CheXPO-v2, a novel alignment framework that shifts from outcome to process supervision. Our core innovation is a Knowledge Graph Consistency Reward mechanism driven by Entity-Relation Matching. By explicitly parsing reasoning steps into structured "Disease, Relation, Anatomy" triplets, we provide fine-grained supervision that penalizes incoherent logic and hallucinations at the atomic level. Integrating this with a hard-example mining strategy, our approach significantly outperforms GRPO and state-of-the-art models on benchmarks like MIMIC-CXR-VQA. Crucially, CheXPO-v2 achieves new state-of-the-art accuracy using only 5k samples, demonstrating exceptional data efficiency while producing clinically sound and verifiable reasoning. The project source code is publicly available at: https://github.com/ecoxial2007/CheX-Phi4MM.

PrediFlow: A Flow-Based Prediction-Refinement Framework for Real-Time Human Motion Prediction in Human-Robot Collaboration

Stochastic human motion prediction is critical for safe and effective human-robot collaboration (HRC) in industrial remanufacturing, as it captures human motion uncertainties and multi-modal behaviors that deterministic methods cannot handle. While earlier works emphasize highly diverse predictions, they often generate unrealistic human motions. More recent methods focus on accuracy and real-time performance, yet there remains potential to improve prediction quality further without exceeding time budgets. Additionally, current research on stochastic human motion prediction in HRC typically considers human motion in isolation, neglecting the influence of robot motion on human behavior. To address these research gaps and enable real-time, realistic, and interaction-aware human motion prediction, we propose a novel prediction-refinement framework that integrates both human and robot observed motion to refine the initial predictions produced by a pretrained state-of-the-art predictor. The refinement module employs a Flow Matching structure to account for uncertainty. Experimental studies on the HRC desktop disassembly dataset demonstrate that our method significantly improves prediction accuracy while preserving the uncertainties and multi-modalities of human motion. Moreover, the total inference time of the proposed framework remains within the time budget, highlighting the effectiveness and practicality of our approach.

SAIL-Embedding Technical Report: Omni-modal Embedding Foundation Model

Multimodal embedding models aim to yield informative unified representations that empower diverse cross-modal tasks. Despite promising developments in the evolution from CLIP-based dual-tower architectures to large vision-language models, prior works still face unavoidable challenges in real-world applications and business scenarios, such as the limited modality support, unstable training mechanisms, and industrial domain gaps. In this work, we introduce SAIL-Embedding, an omni-modal embedding foundation model that addresses these issues through tailored training strategies and architectural design. In the optimization procedure, we propose a multi-stage training scheme to boost the multifaceted effectiveness of representation learning. Specifically, the content-aware progressive training aims to enhance the model's adaptability to diverse downstream tasks and master enriched cross-modal proficiency. The collaboration-aware recommendation enhancement training further adapts multimodal representations for recommendation scenarios by distilling knowledge from sequence-to-item and ID-to-item embeddings while mining user historical interests. Concurrently, we develop the stochastic specialization and dataset-driven pattern matching to strengthen model training flexibility and generalizability. Experimental results show that SAIL-Embedding achieves SOTA performance compared to other methods in different retrieval tasks. In online experiments across various real-world scenarios integrated with our model, we observe a significant increase in Lifetime (LT), which is a crucial indicator for the recommendation experience. For instance, the model delivers the 7-day LT gain of +0.158% and the 14-day LT gain of +0.144% in the Douyin-Selected scenario. For the Douyin feed rank model, the match features produced by SAIL-Embedding yield a +0.08% AUC gain.

Dynamic Generation of Multi-LLM Agents Communication Topologies with Graph Diffusion Models

The efficiency of multi-agent systems driven by large language models (LLMs) largely hinges on their communication topology. However, designing an optimal topology is a non-trivial challenge, as it requires balancing competing objectives such as task performance, communication cost, and robustness. Existing frameworks often rely on static or hand-crafted topologies, which inherently fail to adapt to diverse task requirements, leading to either excessive token consumption for simple problems or performance bottlenecks for complex ones. To address this challenge, we introduce a novel generative framework called \textit{Guided Topology Diffusion (GTD)}. Inspired by conditional discrete graph diffusion models, GTD formulates topology synthesis as an iterative construction process. At each step, the generation is steered by a lightweight proxy model that predicts multi-objective rewards (e.g., accuracy, utility, cost), enabling real-time, gradient-free optimization towards task-adaptive topologies. This iterative, guided synthesis process distinguishes GTD from single-step generative frameworks, enabling it to better navigate complex design trade-offs. We validated GTD across multiple benchmarks, and experiments show that this framework can generate highly task-adaptive, sparse, and efficient communication topologies, significantly outperforming existing methods in LLM agent collaboration.

SAIL-VL2 Technical Report

We introduce SAIL-VL2, an open-suite vision-language foundation model (LVM) for comprehensive multimodal understanding and reasoning. As the successor to SAIL-VL, SAIL-VL2 achieves state-of-the-art performance at the 2B and 8B parameter scales across diverse image and video benchmarks, demonstrating strong capabilities from fine-grained perception to complex reasoning. Its effectiveness is driven by three core innovations. First, a large-scale data curation pipeline with scoring and filtering strategies enhances both quality and distribution across captioning, OCR, QA, and video data, improving training efficiency. Second, a progressive training framework begins with a powerful pre-trained vision encoder (SAIL-ViT), advances through multimodal pre-training, and culminates in a thinking-fusion SFT-RL hybrid paradigm that systematically strengthens model capabilities. Third, architectural advances extend beyond dense LLMs to efficient sparse Mixture-of-Experts (MoE) designs. With these contributions, SAIL-VL2 demonstrates competitive performance across 106 datasets and achieves state-of-the-art results on challenging reasoning benchmarks such as MMMU and MathVista. Furthermore, on the OpenCompass leaderboard, SAIL-VL2-2B ranks first among officially released open-source models under the 4B parameter scale, while serving as an efficient and extensible foundation for the open-source multimodal community.

UniSearch: Rethinking Search System with a Unified Generative Architecture

Modern search systems play a crucial role in facilitating information acquisition. Traditional search engines typically rely on a cascaded architecture, where results are retrieved through recall, pre-ranking, and ranking stages. The complexity of designing and maintaining multiple modules makes it difficult to achieve holistic performance gains. Recent advances in generative recommendation have motivated the exploration of unified generative search as an alternative. However, existing approaches are not genuinely end-to-end: they typically train an item encoder to tokenize candidates first and then optimize a generator separately, leading to objective inconsistency and limited generalization. To address these limitations, we propose UniSearch, a unified generative search framework for Kuaishou Search. UniSearch replaces the cascaded pipeline with an end-to-end architecture that integrates a Search Generator and a Video Encoder. The Generator produces semantic identifiers of relevant items given a user query, while the Video Encoder learns latent item embeddings and provides their tokenized representations. A unified training framework jointly optimizes both components, enabling mutual enhancement and improving representation quality and generation accuracy. Furthermore, we introduce Search Preference Optimization (SPO), which leverages a reward model and real user feedback to better align generation with user preferences. Extensive experiments on industrial-scale datasets, together with online A/B testing in both short-video and live search scenarios, demonstrate the strong effectiveness and deployment potential of UniSearch. Notably, its deployment in live search yields the largest single-experiment improvement in recent years of our product's history, highlighting its practical value for real-world applications.