As Natural Language Generation (NLG) continues to be widely adopted, properly assessing it has become quite difficult. Lately, using large language models (LLMs) for evaluating these generations has gained traction, as they tend to align more closely with human preferences than conventional n-gram or embedding-based metrics. In our experiments, we show that LLM judges have low intra-rater reliability in their assigned scores across different runs. This variance makes their ratings inconsistent, almost arbitrary in the worst case, making it difficult to measure how good their judgments actually are. We quantify this inconsistency across different NLG tasks and benchmarks and see if judicious use of LLM judges can still be useful following proper guidelines.

Aggregating intracranial recordings across subjects is challenging since electrode count, placement, and covered regions vary widely. Spatial normalization methods like MNI coordinates offer a shared anatomical reference, but often fail to capture true functional similarity, particularly when localization is imprecise; even at matched anatomical coordinates, the targeted brain region and underlying neural dynamics can differ substantially between individuals. We propose a scalable representation-learning framework that (i) learns a subject-agnostic functional identity for each electrode from multi-region local field potentials using a Siamese encoder with contrastive objectives, inducing an embedding geometry that is locality-sensitive to region-specific neural signatures, and (ii) tokenizes these embeddings for a transformer that models inter-regional relationships with a variable number of channels. We evaluate this framework on a 20-subject dataset spanning basal ganglia-thalamic regions collected during flexible rest/movement recording sessions with heterogeneous electrode layouts. The learned functional space supports accurate within-subject discrimination and forms clear, region-consistent clusters; it transfers zero-shot to unseen channels. The transformer, operating on functional tokens without subject-specific heads or supervision, captures cross-region dependencies and enables reconstruction of masked channels, providing a subject-agnostic backbone for downstream decoding. Together, these results indicate a path toward large-scale, cross-subject aggregation and pretrain-ing for intracranial neural data where strict task structure and uniform sensor placement are unavailable. Building models that generalize across subjects in neuroscience requires representations that remain stable despite variability in data acquisition. Intracranial neural recordings lack this stability: electrode locations, counts, sampling, and coverage differ across individuals, reflecting clinical needs rather than standardized layouts. Without a shared representational system, cross-subject aggregation is unreliable, limiting scalable modeling and clinical translation. Such recordings are uniquely valuable for studying inter-regional communication, yet their heterogeneity makes them especially challenging to align. In practice, two obstacles dominate: Anatomical variability and inconsistent electrode coverage.

Safety guardrails in large language models(LLMs) are developed to prevent malicious users from generating toxic content at a large scale. However, these measures can inadvertently introduce or reflect new biases, as LLMs may refuse to generate harmful content targeting some demographic groups and not others. We explore this selective refusal bias in LLM guardrails through the lens of refusal rates of targeted individual and intersectional demographic groups, types of LLM responses, and length of generated refusals. Our results show evidence of selective refusal bias across gender, sexual orientation, nationality, and religion attributes. This leads us to investigate additional safety implications via an indirect attack, where we target previously refused groups. Our findings emphasize the need for more equitable and robust performance in safety guardrails across demographic groups.

Security applications are increasingly relying on large language models (LLMs) for cyber threat detection; however, their opaque reasoning often limits trust, particularly in decisions that require domain-specific cybersecurity knowledge. Because security threats evolve rapidly, LLMs must not only recall historical incidents but also adapt to emerging vulnerabilities and attack patterns. Retrieval-Augmented Generation (RAG) has demonstrated effectiveness in general LLM applications, but its potential for cybersecurity remains underexplored. In this work, we introduce a RAG-based framework designed to contextualize cybersecurity data and enhance LLM accuracy in knowledge retention and temporal reasoning. Using external datasets and the Llama-3-8B-Instruct model, we evaluate baseline RAG, an optimized hybrid retrieval approach, and conduct a comparative analysis across multiple performance metrics. Our findings highlight the promise of hybrid retrieval in strengthening the adaptability and reliability of LLMs for cybersecurity tasks.

This paper addresses the limitations of large-scale language models in safety alignment and robustness by proposing a fine-tuning method that combines contrastive distillation with noise-robust training. The method freezes the backbone model and transfers the knowledge boundaries of the teacher model to the student model through distillation, thereby improving semantic consistency and alignment accuracy. At the same time, noise perturbations and robust optimization constraints are introduced during training to ensure that the model maintains stable predictive outputs under noisy and uncertain inputs. The overall framework consists of distillation loss, robustness loss, and a regularization term, forming a unified optimization objective that balances alignment ability with resistance to interference. To systematically validate its effectiveness, the study designs experiments from multiple perspectives, including distillation weight sensitivity, stability analysis under computation budgets and mixed-precision environments, and the impact of data noise and distribution shifts on model performance. Results show that the method significantly outperforms existing baselines in knowledge transfer, robustness, and overall safety, achieving the best performance across several key metrics. This work not only enriches the theoretical system of parameter-efficient fine-tuning but also provides a new solution for building safer and more trustworthy alignment mechanisms.

We present MLPerf Automotive, the first standardized public benchmark for evaluating Machine Learning systems that are deployed for AI acceleration in automotive systems. Developed through a collaborative partnership between MLCommons and the Autonomous Vehicle Computing Consortium, this benchmark addresses the need for standardized performance evaluation methodologies in automotive machine learning systems. Existing benchmark suites cannot be utilized for these systems since automotive workloads have unique constraints including safety and real-time processing that distinguish them from the domains that previously introduced benchmarks target. Our benchmarking framework provides latency and accuracy metrics along with evaluation protocols that enable consistent and reproducible performance comparisons across different hardware platforms and software implementations. The first iteration of the benchmark consists of automotive perception tasks in 2D object detection, 2D semantic segmentation, and 3D object detection. We describe the methodology behind the benchmark design including the task selection, reference models, and submission rules. We also discuss the first round of benchmark submissions and the challenges involved in acquiring the datasets and the engineering efforts to develop the reference implementations. Our benchmark code is available at https://github.com/mlcommons/mlperf_automotive.

An LLM's factuality and refusal training can be compromised by simple changes to a prompt. Models often adopt user beliefs (sycophancy) or satisfy inappropriate requests which are wrapped within special text (jailbreaking). We explore \emph{consistency training}, a self-supervised paradigm that teaches a model to be invariant to certain irrelevant cues in the prompt. Instead of teaching the model what exact response to give on a particular prompt, we aim to teach the model to behave identically across prompt data augmentations (like adding leading questions or jailbreak text). We try enforcing this invariance in two ways: over the model's external outputs (\emph{Bias-augmented Consistency Training} (BCT) from Chua et al. [2025]) and over its internal activations (\emph{Activation Consistency Training} (ACT), a method we introduce). Both methods reduce Gemini 2.5 Flash's susceptibility to irrelevant cues. Because consistency training uses responses from the model itself as training data, it avoids issues that arise from stale training data, such as degrading model capabilities or enforcing outdated response guidelines. While BCT and ACT reduce sycophancy equally well, BCT does better at jailbreak reduction. We think that BCT can simplify training pipelines by removing reliance on static datasets. We argue that some alignment problems are better viewed not in terms of optimal responses, but rather as consistency issues.

We present Contamination Detection via Context (CoDeC), a practical and accurate method to detect and quantify training data contamination in large language models. CoDeC distinguishes between data memorized during training and data outside the training distribution by measuring how in-context learning affects model performance. We find that in-context examples typically boost confidence for unseen datasets but may reduce it when the dataset was part of training, due to disrupted memorization patterns. Experiments show that CoDeC produces interpretable contamination scores that clearly separate seen and unseen datasets, and reveals strong evidence of memorization in open-weight models with undisclosed training corpora. The method is simple, automated, and both model- and dataset-agnostic, making it easy to integrate with benchmark evaluations.

Abstract--Enterprise AI agents must continuously adapt to maintain accuracy, reduce latency, and remain aligned with user needs. We present a practical implementation of a data flywheel in NVInfo AI, NVIDIA's Mixture-of-Experts (MoE) Knowledge Assistant serving over 30,000 employees. By operationalizing a MAPE-driven data flywheel, we built a closed-loop system that systematically addresses failures in retrieval-augmented generation (RAG) pipelines and enables continuous learning. Over a 3-month post-deployment period, we monitored feedback and collected 495 negative samples. Analysis revealed two major failure modes: routing errors (5.25%) and query rephrasal errors (3.2%). Using NVIDIA NeMo microservices, we implemented targeted improvements through fine-tuning. For routing, we replaced a Llama 3.1 70B model with a fine-tuned 8B variant, achieving 96% accuracy, a 10 reduction in model size, and 70% latency improvement. For query rephrasal, fine-tuning yielded a 3.7% gain in accuracy and a 40% latency reduction. Our approach demonstrates how human-in-the-loop (HITL) feedback, when structured within a data flywheel, transforms enterprise AI agents into self-improving systems. Key learnings include approaches to ensure agent robustness despite limited user feedback, navigating privacy constraints, and executing staged rollouts in production. This work offers a repeatable blueprint for building robust, adaptive enterprise AI agents capable of learning from real-world usage at scale. Enterprise adoption of generative AI (GenAI) agents has accelerated rapidly, with applications ranging from knowledge retrieval to workflow automation.

Submitted to ApJ ABSTRACT Radiative transfer calculations are essential for modeling planetary atmospheres. However, standard methods are computationally demanding and impose accuracy-speed trade-offs. High computational costs force numerical simplifications in large models (e.g., General Circulation Models) that degrade the accuracy of the simulation. Radiative transfer calculations are an ideal candidate for machine learning emulation: fundamentally, it is a well-defined physical mapping from a static atmospheric profile to the resulting fluxes, and high-fidelity training data can be created from first principles calculations. We developed a radiative transfer emulator using an encoder-only transformer neural network architecture, trained on 1D profiles representative of solar-composition hot Jupiter atmospheres. Our emulator reproduced bolometric two-stream layer fluxes with mean test set errors of 1% compared to the traditional method and achieved speedups of more than 100x. Emulating radiative transfer with machine learning opens up the possibility for faster and more accurate routines within planetary atmospheric models such as GCMs. INTRODUCTION At the heart of almost every computational model of an exoplanet atmosphere lies a radiative transfer routine that determines how radiation is scattered, absorbed, and emitted as it propagates through the atmosphere. These methods are computationally demanding, as they require solutions to integro-differential equations in many distinct wavelength bins.