The traditional approach to morphological inflection (the task of modifying a base word (lemma) to express grammatical categories) has been, for decades, to consider lexical entries of lemma-tag-form triples uniformly, lacking any information about their frequency distribution. However, in production deployment, one might expect the user inputs to reflect a real-world distribution of frequencies in natural texts. With future deployment in mind, we explore the incorporation of corpus frequency information into the task of morphological inflection along three key dimensions during system development: (i) for train-dev-test split, we combine a lemma-disjoint approach, which evaluates the model's generalization capabilities, with a frequency-weighted strategy to better reflect the realistic distribution of items across different frequency bands in training and test sets; (ii) for evaluation, we complement the standard type accuracy (often referred to simply as accuracy), which treats all items equally regardless of frequency, with token accuracy, which assigns greater weight to frequent words and better approximates performance on running text; (iii) for training data sampling, we introduce a method novel in the context of inflection, frequency-aware training, which explicitly incorporates word frequency into the sampling process. We show that frequency-aware training outperforms uniform sampling in 26 out of 43 languages.

Recent advances in large language models (LLMs) have significantly impacted data science workflows, giving rise to specialized data science agents designed to automate analytical tasks. Despite rapid adoption, systematic benchmarks evaluating the efficacy and limitations of these agents remain scarce. In this paper, we introduce a comprehensive benchmark specifically crafted to reflect real-world user interactions with data science agents by observing usage of our commercial applications. We evaluate three LLMs: Claude-4.0-Sonnet, Gemini-2.5-Flash, and OpenAI-o4-Mini across three approaches: zero-shot with context engineering, multi-step with context engineering, and with SmolAgent. Our benchmark assesses performance across a diverse set of eight data science task categories, additionally exploring the sensitivity of models to common prompting issues, such as data leakage and slightly ambiguous instructions. We further investigate the influence of temperature parameters on overall and task-specific outcomes for each model and approach. Our findings reveal distinct performance disparities among the evaluated models and methodologies, highlighting critical factors that affect practical deployment. The benchmark dataset and evaluation framework introduced herein aim to provide a foundation for future research of more robust and effective data science agents.

The fashion retail business is centered around the capacity to comprehend products. Product attribution helps in comprehending products depending on the business process. Quality attribution improves the customer experience as they navigate through millions of products offered by a retail website. It leads to well-organized product catalogs. In the end, product attribution directly impacts the 'discovery experience' of the customer. Although large language models (LLMs) have shown remarkable capabilities in understanding multimodal data, their performance on fine-grained fashion attribute recognition remains under-explored. This paper presents a zero-shot evaluation of state-of-the-art LLMs that balance performance with speed and cost efficiency, mainly GPT-4o-mini and Gemini 2.0 Flash. We have used the dataset DeepFashion-MultiModal (https://github.com/yumingj/DeepFashion-MultiModal) to evaluate these models in the attribution tasks of fashion products. Our study evaluates these models across 18 categories of fashion attributes, offering insight into where these models excel. We only use images as the sole input for product information to create a constrained environment. Our analysis shows that Gemini 2.0 Flash demonstrates the strongest overall performance with a macro F1 score of 56.79% across all attributes, while GPT-4o-mini scored a macro F1 score of 43.28%. Through detailed error analysis, our findings provide practical insights for deploying these LLMs in production e-commerce product attribution-related tasks and highlight the need for domain-specific fine-tuning approaches. This work also lays the groundwork for future research in fashion AI and multimodal attribute extraction.

Recently, large-scale pre-trained speech encoders and Large Language Models (LLMs) have been released, which show state-of-the-art performance on a range of spoken language processing tasks including Automatic Speech Recognition (ASR). To effectively combine both models for better performance, continuous speech prompts, and ASR error correction have been adopted. However, these methods are prone to suboptimal performance or are inflexible. In this paper, we propose a new paradigm, LegoSLM, that bridges speech encoders and LLMs using the ASR posterior matrices. The speech encoder is trained to generate Connectionist Temporal Classification (CTC) posteriors over the LLM vocabulary, which are used to reconstruct pseudo-audio embeddings by computing a weighted sum of the LLM input embeddings. These embeddings are concatenated with text embeddings in the LLM input space. Using the well-performing USM and Gemma models as an example, we demonstrate that our proposed LegoSLM method yields good performance on both ASR and speech translation tasks. By connecting USM with Gemma models, we can get an average of 49% WERR over the USM-CTC baseline on 8 MLS testsets. The trained model also exhibits modularity in a range of settings -- after fine-tuning the Gemma model weights, the speech encoder can be switched and combined with the LLM in a zero-shot fashion. Additionally, we propose to control the decode-time influence of the USM and LLM using a softmax temperature, which shows effectiveness in domain adaptation.

Uncertainty quantification (UQ) in Large Language Models (LLMs) is essential for their safe and reliable deployment, particularly in critical applications where incorrect outputs can have serious consequences. Current UQ methods typically rely on querying the model multiple times using non-zero temperature sampling to generate diverse outputs for uncertainty estimation. However, the impact of selecting a given temperature parameter is understudied, and our analysis reveals that temperature plays a fundamental role in the quality of uncertainty estimates. The conventional approach of identifying optimal temperature values requires expensive hyperparameter optimization (HPO) that must be repeated for each new model-dataset combination. We propose Monte Carlo Temperature (MCT), a robust sampling strategy that eliminates the need for temperature calibration. Our analysis reveals that: 1) MCT provides more robust uncertainty estimates across a wide range of temperatures, 2) MCT improves the performance of UQ methods by replacing fixed-temperature strategies that do not rely on HPO, and 3) MCT achieves statistical parity with oracle temperatures, which represent the ideal outcome of a well-tuned but computationally expensive HPO process. These findings demonstrate that effective UQ can be achieved without the computational burden of temperature parameter calibration.

Vision-Language Models trained on massive collections of human-generated data often reproduce and amplify societal stereotypes. One critical form of stereotyping reproduced by these models is homogeneity bias-the tendency to represent certain groups as more homogeneous than others. We investigate how this bias responds to hyperparameter adjustments in GPT-4, specifically examining sampling temperature and top p which control the randomness of model outputs. By generating stories about individuals from different racial and gender groups and comparing their similarities using vector representations, we assess both bias robustness and its relationship with hyperparameter values. We find that (1) homogeneity bias persists across most hyperparameter configurations, with Black Americans and women being represented more homogeneously than White Americans and men, (2) the relationship between hyperparameters and group representations shows unexpected non-linear patterns, particularly at extreme values, and (3) hyperparameter adjustments affect racial and gender homogeneity bias differently-while increasing temperature or decreasing top p can reduce racial homogeneity bias, these changes show different effects on gender homogeneity bias. Our findings suggest that while hyperparameter tuning may mitigate certain biases to some extent, it cannot serve as a universal solution for addressing homogeneity bias across different social group dimensions.

During language model decoding, it is known that using higher temperature sampling gives more creative responses, while lower temperatures are more factually accurate. However, such models are commonly applied to general instruction following, which involves both creative and fact seeking tasks, using a single fixed temperature across all examples and tokens. In this work, we introduce Adaptive Decoding, a layer added to the model to select the sampling temperature dynamically at inference time, at either the token or example level, in order to optimize performance. To learn its parameters we introduce Latent Preference Optimization (LPO) a general approach to train discrete latent variables such as choices of temperature. Our method outperforms all fixed decoding temperatures across a range of tasks that require different temperatures, including UltraFeedback, Creative Story Writing, and GSM8K.

State-of-the-art space science missions increasingly rely on automation due to spacecraft complexity and the costs of human oversight. The high volume of data, including scientific and telemetry data, makes manual inspection challenging. Machine learning offers significant potential to meet these demands. The Euclid space telescope, in its survey phase since February 2024, exemplifies this shift. Euclid's success depends on accurate monitoring and interpretation of housekeeping telemetry and science-derived data. Thousands of telemetry parameters, monitored as time series, may or may not impact the quality of scientific data. These parameters have complex interdependencies, often due to physical relationships (e.g., proximity of temperature sensors). Optimising science operations requires careful anomaly detection and identification of hidden parameter states. Moreover, understanding the interactions between known anomalies and physical quantities is crucial yet complex, as related parameters may display anomalies with varied timing and intensity. We address these challenges by analysing temperature anomalies in Euclid's telemetry from February to August 2024, focusing on eleven temperature parameters and 35 covariates. We use a predictive XGBoost model to forecast temperatures based on historical values, detecting anomalies as deviations from predictions. A second XGBoost model predicts anomalies from covariates, capturing their relationships to temperature anomalies. We identify the top three anomalies per parameter and analyse their interactions with covariates using SHAP (Shapley Additive Explanations), enabling rapid, automated analysis of complex parameter relationships. Our method demonstrates how machine learning can enhance telemetry monitoring, offering scalable solutions for other missions with similar data challenges.

Large Language Models (LLMs) are powerful computational models trained on extensive corpora of human-readable text, enabling them to perform general-purpose language understanding and generation. LLMs have garnered significant attention in both industry and academia due to their exceptional performance across various natural language processing (NLP) tasks. Despite these successes, LLMs often produce inaccuracies, commonly referred to as hallucinations. Prompt engineering, the process of designing and formulating instructions for LLMs to perform specific tasks, has emerged as a key approach to mitigating hallucinations. This paper provides a comprehensive empirical evaluation of different prompting strategies and frameworks aimed at reducing hallucinations in LLMs. Various prompting techniques are applied to a broad set of benchmark datasets to assess the accuracy and hallucination rate of each method. Additionally, the paper investigates the influence of tool-calling agents (LLMs augmented with external tools to enhance their capabilities beyond language generation) on hallucination rates in the same benchmarks. The findings demonstrate that the optimal prompting technique depends on the type of problem, and that simpler techniques often outperform more complex methods in reducing hallucinations. Furthermore, it is shown that LLM agents can exhibit significantly higher hallucination rates due to the added complexity of external tool usage.

Around 5-10\% of newborns need assistance to start breathing. Currently, there is a lack of evidence-based research, objective data collection, and opportunities for learning from real newborn resuscitation emergency events. Generating and evaluating automated newborn resuscitation algorithm activity timelines relative to the Time of Birth (ToB) offers a promising opportunity to enhance newborn care practices. Given the importance of prompt resuscitation interventions within the "golden minute" after birth, having an accurate ToB with second precision is essential for effective subsequent analysis of newborn resuscitation episodes. Instead, ToB is generally registered manually, often with minute precision, making the process inefficient and susceptible to error and imprecision. In this work, we explore the fusion of Artificial Intelligence (AI) and thermal imaging to develop the first AI-driven ToB detector. The use of temperature information offers a promising alternative to detect the newborn while respecting the privacy of healthcare providers and mothers. However, the frequent inconsistencies in thermal measurements, especially in a multi-camera setup, make normalization strategies critical. Our methodology involves a three-step process: first, we propose an adaptive normalization method based on Gaussian mixture models (GMM) to mitigate issues related to temperature variations; second, we implement and deploy an AI model to detect the presence of the newborn within the thermal video frames; and third, we evaluate and post-process the model's predictions to estimate the ToB. A precision of 88.1\% and a recall of 89.3\% are reported in the detection of the newborn within thermal frames during performance evaluation. Our approach achieves an absolute median deviation of 2.7 seconds in estimating the ToB relative to the manual annotations.