Large Language Models offer impressive language capabilities but suffer from well-known limitations, including hallucinations, biases, privacy concerns, and high computational costs. These issues are largely driven by the combination of linguistic competence and factual memorization within a single monolithic model. This paper introduces and empirically supports the Fundamental Language Model (FLM) paradigm, which advocates for smaller, linguistically competent models that offload factual retrieval to external tools. We evaluate models ranging from 135M to 32B parameters across three dimensions: linguistic competence, external factual knowledge, and internal factual knowledge. Our findings reveal that while both linguistic competence and factual knowledge improve with scale, internal factual knowledge grows significantly faster, suggesting that model size is more closely tied to memorization than to core language ability. These results support a modular approach to language modeling, where compact, linguistically proficient models serve as the foundation for tool-augmented systems. The FLM paradigm offers a path toward more efficient, interpretable, and sustainable NLP solutions.

In multivariate time series (MTS) forecasting, many deep learning based methods have been proposed for modeling dependencies at multiple spatial (inter-variate) or temporal (intra-variate) scales. However, existing methods may fail to model dependencies across multiple spatial-temporal scales (ST-scales, i.e., scales that jointly consider spatial and temporal scopes). In this work, we propose ST-Hyper to model the high-order dependencies across multiple ST-scales through adaptive hypergraph modeling. Specifically, we introduce a Spatial-Temporal Pyramid Modeling (STPM) module to extract features at multiple ST-scales. Furthermore, we introduce an Adaptive Hypergraph Modeling (AHM) module that learns a sparse hypergraph to capture robust high-order dependencies among features. In addition, we interact with these features through tri-phase hypergraph propagation, which can comprehensively capture multi-scale spatial-temporal dynamics. Experimental results on six real-world MTS datasets demonstrate that ST-Hyper achieves the state-of-the-art performance, outperforming the best baselines with an average MAE reduction of 3.8\% and 6.8\% for long-term and short-term forecasting, respectively.

As large language models (LLMs) advance in conversational and reasoning capabilities, their practical application in healthcare has become a critical research focus. However, there is a notable gap between the performance of medical LLMs on static benchmarks such as USMLE and their utility in real-world clinical decision-making. This discrepancy arises because traditional exams fail to capture the dynamic, interactive nature of medical consultations. To address this challenge, we introduce a novel dynamic verification framework that moves beyond static answer verifier, establishing a large-scale, high-fidelity interactive reinforcement learning system. Our framework comprises two key components: a Patient Simulator that creates realistic clinical environments using de-identified medical records, and a Clinical Rubrics Generator that dynamically produces multi-dimensional evaluation metrics. Building on this foundation, we develop Baichuan-M2, a 32B-parameter medical augmented reasoning model trained through a multi-stage reinforcement learning strategy with an improved Group Relative Policy Optimization (GRPO) algorithm. Evaluated on HealthBench, Baichuan-M2 outperforms all other open-source models and most advanced closed-source counterparts, achieving a score above 32 on the challenging HealthBench Hard benchmark-previously exceeded only by GPT-5. Our work demonstrates that robust dynamic verifier system is essential for aligning LLM capabilities with practical clinical applications, establishing a new Pareto front in the performance-parameter trade-off for medical AI deployment.

Large Language Models tend to struggle when dealing with specialized domains. While all aspects of evaluation hold importance, factuality is the most critical one. Similarly, reliable fact-checking tools and data sources are essential for hallucination mitigation. We address these issues by providing a comprehensive Fact-checking Benchmark FActBench covering four generation tasks and six state-of-the-art Large Language Models (LLMs) for the Medical domain. We use two state-of-the-art Fact-checking techniques: Chain-of-Thought (CoT) Prompting and Natural Language Inference (NLI). Our experiments show that the fact-checking scores acquired through the Unanimous Voting of both techniques correlate best with Domain Expert Evaluation.

--Automatic differentiation (AD) is a set of techniques that systematically applies the chain rule to compute the gradients of functions without requiring human intervention. Although the fundamentals of this technology were established decades ago, it is experiencing a renaissance as it plays a key role in efficiently computing gradients for backpropagation in machine learning algorithms. AD is also crucial for many applications in scientific computing domains, particularly emerging techniques that integrate machine learning models within scientific simulations and schemes. Existing AD frameworks have four main limitations: limited support of programming languages, requiring code modifications for AD compatibility, limited performance on scientific computing codes, and a naive store-all solution for forward-pass data required for gradient calculations. These limitations force domain scientists to manually compute the gradients for large problems. This work presents DaCe AD, a general, efficient automatic differentiation engine that requires no code modifications. DaCe AD uses a novel ILP-based algorithm to optimize the trade-off between storing and recomputing to achieve maximum performance within a given memory constraint. We showcase the generality of our method by applying it to NPBench, a suite of HPC benchmarks with diverse scientific computing patterns, where we outperform JAX, a Python framework with state-of-the-art general AD capabilities, by more than 92 times on average without requiring any code changes. Automatic differentiation (AD) is a technique for calculating the derivatives of programs [1] that outperforms traditional methods like symbolic and numerical differentiation, particularly for complex algorithms and mathematical functions [2]. AD is critical for training neural networks, as it computes gradients required for backpropagation [3], [4]. Advancements in deep learning were significantly facilitated by AD's efficiency in computing gradients for complex loss functions [5], enabling the training of large-scale models such as modern Large Language Models [6]. Beyond the confines of machine learning in fields such as atmospheric sciences and oceanography, AD is crucial for sensitivity analysis [7], parameter estimation [8], and data assimilation [9]. Furthermore, many recent scientific algorithms have integrated machine learning models within their systems [10], making AD more relevant Figure 1: DaCe AD Contribution Overview.

In this paper, we present a data-driven Forward Selection with Neighborhood Refinement (FSNR) algorithm to determine the number and placement of Phasor Measurement Units (PMUs) for maximizing deep-learning-based fault diagnosis performance. Candidate PMU locations are ranked via a cross-validated Support Vector Machine (SVM) classifier, and each selection is refined through local neighborhood exploration to produce a near-optimal sensor set. The resulting PMU subset is then supplied to a 1D Convolutional Neural Network (CNN) for faulted-line localization and fault-type classification from time-series measurements. Evaluation on modified IEEE 34- and IEEE 123-bus systems demonstrates that the proposed FSNR-SVM method identifies a minimal PMU configuration that achieves the best overall CNN performance, attaining over 96 percent accuracy in fault location and over 99 percent accuracy in fault-type classification on the IEEE 34 system, and approximately 94 percent accuracy in fault location and around 99.8 percent accuracy in fault-type classification on the IEEE 123 system.

Integrating Large Language Models (LLMs) into robotic systems has revolutionised embodied artificial intelligence, enabling advanced decision-making and adaptability. However, ensuring reliability--encompassing both security against adversarial attacks and safety in complex environments--remains a critical challenge. To address this, we propose a unified framework that mitigates prompt injection attacks while enforcing operational safety through robust validation mechanisms. Our approach combines prompt assembling, state management, and safety validation, evaluated using both performance and security metrics. Experiments show a 30.8% improvement under injection attacks and up to a 325% improvement in complex environment settings under adversarial conditions compared to baseline scenarios. The framework is open-sourced with simulation and physical deployment demos at https://llmeyesim.vercel.app/. Introduction The integration of Large Language Models (LLMs) into embodied robotic systems represents a significant leap in robotic autonomy and adaptability [11]. Recent advances enable robots to interpret natural language instructions, fuse multimodal sensor data, and make planning decisions using the general-purpose reasoning capabilities of models like GPT -4o [22]. These capabilities promise generalist agents that can execute complex, interactive tasks without task-specific training [14]. By drawing on vast internet-scale training corpora, LLMs can produce structured action plans from ambiguous user goals, acting as high-level controllers in dynamic and unpredictable environments [12]. However, these benefits come with risks. Unlike traditional robotic architectures that rely on modular safety subsystems, such as collision avoidance, mission timeouts, and hardware constraints, LLM-based controllers can bypass these safeguards via incorrect inference or adversarial inputs. The semantic sensitivity of LLMs to phrasing, ambiguity, or hallucinated knowledge introduces vulnerabilities not addressed by existing robotics safety protocols [6]. Moreover, integrating multimodal perception (e.g., camera, LiDAR) expands the input space but also introduces new failure modes, where partial, spoofed, or contextually misleading inputs can lead to unsafe behaviours [31]. The current literature lacks a unified methodology to secure and validate the behaviour of LLM-driven robots. Most prior work evaluates vision-language reasoning or robotic planning in isolation and does not consider how prompt injection attacks or input spoofing a ffect downstream physical actions. Similarly, existing LLM safety work focuses on digital assistants or text-only settings, leaving a critical gap in embodied use cases such as autonomous navigation and exploration [37, 20]. As robots begin to operate in open-world human environments, the absence of integrated security and safety layers poses real risks to both mission success and human-robot interaction.

Hair artifacts in dermoscopic images present significant challenges for accurate skin lesion analysis, potentially obscuring critical diagnostic features in dermatological assessments. This work introduces a fine-tuned SegFormer model augmented with dropout regularization to achieve precise hair mask segmentation. The proposed SegformerWithDropout architecture leverages the MiT-B2 encoder, pretrained on ImageNet, with an in-channel count of 3 and 2 output classes, incorporating a dropout probability of 0.3 in the segmentation head to prevent overfitting. Training is conducted on a specialized dataset of 500 dermoscopic skin lesion images with fine-grained hair mask annotations, employing 10-fold cross-validation, AdamW optimization with a learning rate of 0.001, and cross-entropy loss. Early stopping is applied based on validation loss, with a patience of 3 epochs and a maximum of 20 epochs per fold. Performance is evaluated using a comprehensive suite of metrics, including Intersection over Union (IoU), Dice coefficient, Peak Signal-to-Noise Ratio (PSNR), Structural Similarity Index (SSIM), and Learned Perceptual Image Patch Similarity (LPIPS). Experimental results from the cross-validation demonstrate robust performance, with average Dice coefficients reaching approximately 0.96 and IoU values of 0.93, alongside favorable PSNR (around 34 dB), SSIM (0.97), and low LPIPS (0.06), highlighting the model's effectiveness in accurate hair artifact segmentation and its potential to enhance preprocessing for downstream skin cancer detection tasks.

Large Language Models (LLMs) can inadvertently reflect societal biases present in their training data, leading to harmful or prejudiced outputs. In the Indian context, our empirical evaluations across a suite of models reveal that biases around caste and religion are particularly salient. Yet, most existing mitigation strategies are Western-centric and fail to address these local nuances. We propose AMBEDKAR, a framework inspired by the egalitarian vision of Dr B. R. Ambedkar, architect of the Indian Constitution, to guide LLM outputs toward fairness, neutrality, and inclusion in line with Articles 14 to 17. Our approach introduces a Constitution-Aware Decoding Layer, guided by the AI Constitution of India and applied only at inference time, without any parameter updates to the base model. We incorporate a speculative decoding algorithm that proactively reduces casteist and communal bias during generation. This mitigation layer operates directly within the decoding process, avoiding changes to model internals and lowering the computational and infrastructural costs associated with retraining. We reinterpret speculative decoding not merely as an efficiency tool but as a mechanism for fairness. In this framework, a Small Language Model (SLM) acts as a potentially biased generator, while a constitutionally guided Large Language Model (LLM) serves as the verifier. Rather than accelerating generation, the LLM enforces bias-robust trajectories in the SLM outputs. This inversion of roles gives rise to a fairness-by-speculation paradigm. Our approach yields an absolute reduction of bias up to 26.41 percent compared to baseline. Our source code, datasets, and results are available at https://anonymous.4open.science/r/AMBEDKAR-983B/

Visual Place Recognition (VPR) has evolved from handcrafted descriptors to deep learning approaches, yet significant challenges remain. Current approaches, including Vision Foundation Models (VFMs) and Multimodal Large Language Models (MLLMs), enhance semantic understanding but suffer from high computational overhead and limited cross-domain transferability when fine-tuned. To address these limitations, we propose a novel zero-shot framework employing Test-Time Scaling (TTS) that leverages MLLMs' vision-language alignment capabilities through Guidance-based methods for direct similarity scoring. Our approach eliminates two-stage processing by employing structured prompts that generate length-controllable JSON outputs. The TTS framework with Uncertainty-Aware Self-Consistency (UASC) enables real-time adaptation without additional training costs, achieving superior generalization across diverse environments. Experimental results demonstrate significant improvements in cross-domain VPR performance with up to 210 computational efficiency gains. Keywords: Multimodal Large language Model, Visual Place Recognition, Zero Shot Learning, Autonomous Systems. 1. INTRODUCTION Visual Place Recognition (VPR) represents a cornerstone capability in computer vision and robotics, enabling systems to determine whether a location has been previously visited based solely on visual input.