This paper presents a feasibility study on the deployment of a European Deep Inference Fabric (eDIF), an NDIF-compatible infrastructure designed to support mechanistic interpretability research on large language models. The need for widespread accessibility of LLM interpretability infrastructure in Europe drives this initiative to democratize advanced model analysis capabilities for the research community. The project introduces a GPU-based cluster hosted at Ansbach University of Applied Sciences and interconnected with partner institutions, enabling remote model inspection via the NNsight API. A structured pilot study involving 16 researchers from across Europe evaluated the platform's technical performance, usability, and scientific utility. Users conducted interventions such as activation patching, causal tracing, and representation analysis on models including GPT-2 and DeepSeek-R1-70B. The study revealed a gradual increase in user engagement, stable platform performance throughout, and a positive reception of the remote experimentation capabilities. It also marked the starting point for building a user community around the platform. Identified limitations such as prolonged download durations for activation data as well as intermittent execution interruptions are addressed in the roadmap for future development. This initiative marks a significant step towards widespread accessibility of LLM interpretability infrastructure in Europe and lays the groundwork for broader deployment, expanded tooling, and sustained community collaboration in mechanistic interpretability research.

Multimodal Large Language Models (MLLMs) have demonstrated remarkable capabilities across a diverse range of mul-timodal tasks. However, these models suffer from a core problem known as text dominance: they depend heavily on text for their inference, while underutilizing other modalities. While prior work has acknowledged this phenomenon in vision-language tasks, often attributing it to data biases or model architectures. In this paper, we conduct the first systematic investigation of text dominance across diverse data modalities, including images, videos, audio, time-series, and graphs. To measure this imbalance, we propose two evaluation metrics: the Modality Dominance Index (MDI) and the Attention Efficiency Index (AEI). Our comprehensive analysis reveals that text dominance is both significant and pervasive across all tested modalities. Our in-depth analysis identifies three underlying causes: attention dilution from severe token redundancy in non-textual modalities, the influence of fusion architecture design, and task formulations that implicitly favor textual inputs. Furthermore, we propose a simple token compression method that effectively rebalances model attention. Applying this method to LLaV A-7B, for instance, drastically reduces its MDI from 10.23 to a well-balanced value of 0.86.

Reward sparsity in long-horizon reinforcement learning (RL) tasks remains a significant challenge, while existing outcome-based reward shaping struggles to define meaningful immediate rewards without introducing bias or requiring explicit task decomposition. Alternatively, verification-based reward shaping uses stepwise critics, but misalignment between immediate rewards and long-term objectives can lead to reward hacking and suboptimal policies. In this work, we address this problem in the context of software engineering (SWE) tasks, where multi-turn reasoning and rule-based verification are critical. We introduce the SWE-oriented RL Framework, a unified system supporting multi-turn interaction, docker-based execution, and customizable reward functions. Additionally, we propose Gated Reward Accumulation (G-RA), a novel method that accumulates immediate rewards only when high-level (long-term) rewards meet a predefined threshold, ensuring stable RL optimization. Experiments on SWE-bench V eri-fied and kBench demonstrate that G-RA leads to an increase in completion rates (47.6% 93.8% and 22.0% 86.0%) and modification rates (19.6% 23.8% and 12.0% 42.0%), while avoiding policy degradation caused by reward misalignment. Our findings highlight the importance of balanced reward accumulation in long-horizon RL and provide a practical solution.

Real-world multimodal applications often require any-to-any capabilities, enabling both understanding and generation across modalities including text, image, audio, and video. However, integrating the strengths of autoregressive language models (LLMs) for reasoning and diffusion models for high-fidelity generation remains challenging. Existing approaches rely on rigid pipelines or tightly coupled architectures, limiting flexibility and scalability. We propose MAGUS (Multi-Agent Guided Unified Multimodal System), a modular framework that unifies multimodal understanding and generation via two decoupled phases: Cognition and Deliberation. MAGUS enables symbolic multi-agent collaboration within a shared textual workspace. In the Cognition phase, three role-conditioned multimodal LLM agents-- Perceiver, Planner, and Reflector --engage in collaborative dialogue to perform structured understanding and planning. The Deliberation phase incorporates a Growth-A ware Search mechanism that orchestrates LLM-based reasoning and diffusion-based generation in a mutually reinforcing manner. MAGUS supports plug-and-play extensibility, scalable any-to-any modality conversion, and semantic alignment--all without the need for joint training. Experiments across multiple benchmarks, including image, video, and audio generation, as well as cross-modal instruction following, demonstrate that MAGUS outperforms strong baselines and state-of-the-art systems.

Full-process clinical diagnosis in the real world encompasses the entire diagnostic workflow that begins with only an ambiguous chief complaint. While artificial intelligence (AI), particularly large language models (LLMs), is transforming clinical diagnosis, its role remains largely as an assistant to physicians. This AI-assisted working pattern makes AI can only answer specific medical questions at certain parts within the diagnostic process, but lack the ability to drive the entire diagnostic process starting from an ambiguous complaint, which still relies heavily on human physicians. This gap limits AI's ability to fully reduce physicians' workload and enhance diagnostic efficiency. To address this, we propose a paradigm shift that reverses the relationship between physicians and AI: repositioning AI as the primary director, with physicians serving as its assistants. So we present DxDirector-7B, an LLM endowed with advanced deep thinking capabilities, enabling it to drive the full-process diagnosis with minimal physician involvement. Furthermore, DxDirector-7B establishes a robust accountability framework for misdiagnoses, delineating responsibility between AI and human physicians. In evaluations across rare, complex, and real-world cases under full-process diagnosis setting, DxDirector-7B not only achieves significant superior diagnostic accuracy but also substantially reduces physician workload than state-of-the-art medical LLMs as well as general-purpose LLMs. Fine-grained analyses across multiple clinical departments and tasks validate its efficacy, with expert evaluations indicating its potential to serve as a viable substitute for medical specialists. These findings mark a new era where AI, traditionally a physicians' assistant, now drives the entire diagnostic process to drastically reduce physicians' workload, indicating an efficient and accurate diagnostic solution.

Contemporary spatial services such as online maps predominantly rely on user queries for location searches. However, the user experience is limited when performing complex tasks, such as searching for a group of locations simultaneously. In this study, we examine the extended scenario known as Spatial Exemplar Query (SEQ), where multiple relevant locations are jointly searched based on user-specified examples. We introduce SEQ-GPT, a spatial query system powered by Large Language Models (LLMs) towards more versatile SEQ search using natural language. The language capabilities of LLMs enable unique interactive operations in the SEQ process, including asking users to clarify query details and dynamically adjusting the search based on user feedback. We also propose a tailored LLM adaptation pipeline that aligns natural language with structured spatial data and queries through dialogue synthesis and multi-model cooperation. SEQ-GPT offers an end-to-end demonstration for broadening spatial search with realistic data and application scenarios.

Question answering over Scholarly Knowledge Graphs (SKGs) remains a challenging task due to the complexity of scholarly content and the intricate structure of these graphs. Large Language Model (LLM) approaches could be used to translate natural language questions (NLQs) into SPARQL queries; however, these LLM-based approaches struggle with SPARQL query generation due to limited exposure to SKG-specific content and the underlying schema. We identified two main types of errors in the LLM-generated SPARQL queries: (i) structural inconsistencies, such as missing or redundant triples in the queries, and (ii) semantic inaccuracies, where incorrect entities or properties are shown in the queries despite a correct query structure. To address these issues, we propose FIRESPARQL, a modular framework that supports fine-tuned LLMs as a core component, with optional context provided via retrieval-augmented generation (RAG) and a SPARQL query correction layer. We evaluate the framework on the SciQA Benchmark using various configurations (zero-shot, zero-shot with RAG, one-shot, fine-tuning, and fine-tuning with RAG) and compare the performance with baseline and state-of-the-art approaches. We measure query accuracy using BLEU and ROUGE metrics, and query result accuracy using relaxed exact match(RelaxedEM), with respect to the gold standards containing the NLQs, SPARQL queries, and the results of the queries. Experimental results demonstrate that fine-tuning achieves the highest overall performance, reaching 0.90 ROUGE-L for query accuracy and 0.85 RelaxedEM for result accuracy on the test set.

Graph neural networks (GNNs) have achieved state-of-the-art results in computer vision and medical image classification tasks by capturing structural dependencies across data instances. However, their decision-making remains largely opaque limiting their trustworthiness in high-stakes clinical applications, where interpretability is essential. Existing explainability techniques for GNNs are typically post-hoc and global, offering limited insight into individual node decisions or local reasoning. We introduce X-Node, a self-explaining GNN framework in which each node generates its own explanation as part of the prediction process. For every node, we construct a structured context vector encoding interpretable cues, such as degree, centrality, clustering, feature saliency, and label agreement within its local topology. A lightweight Reasoner module maps this context into a compact explanation vector, which serves three purposes: (1) reconstructing the node's latent embedding via a Decoder to enforce faithfulness, (2) generating a natural language explanation using a pre-trained LLM (e.g., Grok or Gemini), and (3) guiding the GNN itself via a "text-injection" mechanism that feeds explanations back into the message-passing pipeline. We evaluate X-Node on two graph datasets derived from MedMNIST and MorphoMNIST, integrating it with GCN, GAT, and GIN backbones. Our results show that X-Node maintains competitive classification accuracy while producing faithful, per-node explanations.

Counterfactual explanations provide human-understandable reasoning for AI-made decisions by describing minimal changes to input features that would alter a model's prediction. To be truly useful in practice, such explanations must be realistic and feasible -- they should respect both the underlying data distribution and user-defined feasibility constraints. Existing approaches often enforce inter-feature dependencies through rigid, hand-crafted constraints or domain-specific knowledge, which limits their generalizability and ability to capture complex, nonlinear relations inherent in data. Moreover, they rarely accommodate user-specified preferences and suggest explanations that are causally implausible or infeasible to act upon. We introduce RealAC, a domain-agnostic framework for generating realistic and actionable counterfactuals. RealAC automatically preserves complex inter-feature dependencies without relying on explicit domain knowledge -- by aligning the joint distributions of feature pairs between factual and counterfactual instances. The framework also allows end-users to ``freeze'' attributes they cannot or do not wish to change by suppressing change in frozen features during optimization. Evaluations on three synthetic and two real datasets demonstrate that RealAC balances realism with actionability. Our method outperforms state-of-the-art baselines and Large Language Model-based counterfactual generation techniques in causal edge score, dependency preservation score, and IM1 realism metric and offers a solution for causality-aware and user-centric counterfactual generation.

Generating textual rationales from large vision-language models (LVLMs) to support trainable multimodal misinformation detectors has emerged as a promising paradigm. However, its effectiveness is fundamentally limited by three core challenges: (i) insufficient diversity in generated rationales, (ii) factual inaccuracies due to hallucinations, and (iii) irrelevant or conflicting content that introduces noise. We introduce DiFaR, a detector-agnostic framework that produces diverse, factual, and relevant rationales to enhance misinformation detection. DiFaR employs five chain-of-thought prompts to elicit varied reasoning traces from LVLMs and incorporates a lightweight post-hoc filtering module to select rationale sentences based on sentence-level factuality and relevance scores. Extensive experiments on four popular benchmarks demonstrate that DiFaR outperforms four baseline categories by up to 5.9% and boosts existing detectors by as much as 8.7%. Both automatic metrics and human evaluations confirm that DiFaR significantly improves rationale quality across all three dimensions.