If you ever had pastries at breakfast, drank soy milk, used soaps at home, or built yourself a nice flat-pack piece of furniture, you may have contributed to deforestation and climate change. Every item has a price--but the cost isn't felt only in our pockets. Hidden in that price is a complex chain of production, encompassing economic, social, and environmental relations that sustain livelihoods and, unfortunately, contribute to habitat destruction, deforestation, and the warming of our planet. Approximately 4 billion hectares of forest around the world act as a carbon sink which, over the past two decades, has annually absorbed a net 7.6 billion metric tons of CO2. Conversely, a cleared forest becomes a carbon source.

The piece seems to conjure a prehistoric avant-garde musical workshop, a sonic analogue of the visual culture that can be glimpsed in the cave. Fully notated passages--scampering runs, precisely hammering chords, ghostly arpeggios--are interspersed with opportunities for improvisation. The first twenty-four bars indicate rhythms, dynamics, and registers but not precise pitches. The ending, too, is left open. Cory Smythe, himself a composer and improviser of note, proved an ideal conduit, making the distinction between Lewis's ideas and his own elaborations inconsequential.

We introduce a launch device, called the remotely-enabled modular release mechanism, to augment rapid testing and prototyping of cooperative autonomy maritime applications by facilitating autonomous deployment of an autonomous underwater vehicle (AUV) from an autonomous surface vessel (ASV). While we focus our development on a specific application of deploying an AUV from a catamaran style ASV, the release mechanism can be adapted to different deployable objects and towing vehicles, such as buoys and sensors for oceanographic surveys or mono-hull ASVs. In this paper we explore a number of hardware and software design considerations to facilitate ease of integration with existing maritime autonomy systems. We expound on bench tests and in-water tests used to explore the utility of the release system and diagnose system issues. Additionally, we make a first-principles argument, based on a hydrodynamics physics model, for assured deployment that is virtually independent of sea state, making the release system a suitable alternative for different maritime applications in varying environmental conditions.

Yiwen Lin, University of California, Irvine Lauren Snow, University of California, Irvine Acknowledgments: This work was partially supported by the National Science Foundation (Grant Number 1535300), and National Institutes of Health (Grant Number 5UC2NS128361-02). Abstract Collaboration is key to STEM, where multidisciplinary team research can solve complex problems. However, inequality in STEM fields hinders their full potential, due to persistent psychological barriers in underrepresented students' experience. This paper documents teamwork in STEM and explores the transformative potential of computational modeling and generative AI in promoting STEM-team diversity and inclusion. Leveraging generative AI, this paper outlines two primary areas for advancing diversity, equity, and inclusion. First, formalizing collaboration assessment with inclusive analytics can capture fine-grained learner behavior. Second, adaptive, personalized AI systems can support diversity and inclusion in STEM teams. Four policy recommendations highlight AI's capacity: formalized collaborative skill assessment, inclusive analytics, funding for socio-cognitive research, human-AI teaming for inclusion training.

Fine-tuning large pre-trained language models for downstream tasks remains a critical challenge in natural language processing. This paper presents an empirical analysis comparing two efficient fine-tuning methods - BitFit and adapter modules - to standard full model fine-tuning. Experiments conducted on GLUE benchmark datasets (MRPC, COLA, STS-B) reveal several key insights. The BitFit approach, which trains only bias terms and task heads, matches full fine-tuning performance across varying amounts of training data and time constraints. It demonstrates remarkable stability even with only 30\% of data, outperforming full fine-tuning at intermediate data levels. Adapter modules exhibit high variability, with inconsistent gains over default models. The findings indicate BitFit offers an attractive balance between performance and parameter efficiency. Our work provides valuable perspectives on model tuning, emphasizing robustness and highlighting BitFit as a promising alternative for resource-constrained or streaming task settings. The analysis offers actionable guidelines for efficient adaptation of large pre-trained models, while illustrating open challenges in stabilizing techniques like adapter modules.

Generative AI and transfer learning fields have experienced remarkable advancements in recent years especially in the domain of Natural Language Processing (NLP). Transformers were at the heart of these advancements where the cutting-edge transformer-based Language Models (LMs) enabled new state-of-the-art results in a wide spectrum of applications. While the number of research works involving neural LMs is exponentially increasing, their vast majority are high-level and far from self-contained. Consequently, a deep understanding of the literature in this area is a tough task especially at the absence of a unified mathematical framework explaining the main types of neural LMs. We address the aforementioned problem in this tutorial where the objective is to explain neural LMs in a detailed, simplified and unambiguous mathematical framework accompanied with clear graphical illustrations. Concrete examples on widely used models like BERT and GPT2 are explored. Finally, since transformers pretrained on language-modeling-like tasks have been widely adopted in computer vision and time series applications, we briefly explore some examples of such solutions in order to enable readers understand how transformers work in the aforementioned domains and compare this use with the original one in NLP.

To better understand the behavior of image classifiers, it is useful to visualize the contribution of individual pixels to the model prediction. In this study, we propose a method, MoXI~($\textbf{Mo}$del e$\textbf{X}$planation by $\textbf{I}$nteractions), that efficiently and accurately identifies a group of pixels with high prediction confidence. The proposed method employs game-theoretic concepts, Shapley values and interactions, taking into account the effects of individual pixels and the cooperative influence of pixels on model confidence. Theoretical analysis and experiments demonstrate that our method better identifies the pixels that are highly contributing to the model outputs than widely-used visualization methods using Grad-CAM, Attention rollout, and Shapley value. While prior studies have suffered from the exponential computational cost in the computation of Shapley value and interactions, we show that this can be reduced to linear cost for our task.

Given the nonlinearity of the interaction between weather and soil variables, a novel deep neural network regressor (DNNR) was carefully designed with considerations to the depth, number of neurons of the hidden layers, and the hyperparameters with their optimizations. Additionally, a new metric, the average of absolute root squared error (ARSE) was proposed to address the shortcomings of root mean square error (RMSE) and mean absolute error (MAE) while combining their strengths. Using the ARSE metric, the random forest regressor (RFR) and the extreme gradient boosting regressor (XGBR), were compared with DNNR. The RFR and XGBR achieved yield errors of 0.0000294 t/ha, and 0.000792 t/ha, respectively, compared to the DNNR(s) which achieved 0.0146 t/ha and 0.0209 t/ha, respectively. All errors were impressively small. However, with changes to the explanatory variables to ensure generalizability to unforeseen data, DNNR(s) performed best. The unforeseen data, different from unseen data, is coined to represent sudden and unexplainable change to weather and soil variables due to climate change. Further analysis reveals that a strong interaction does exist between weather and soil variables. Using precipitation and silt, which are strong-negatively and strong-positively correlated with yield, respectively, yield was observed to increase when precipitation was reduced and silt increased, and vice-versa.

For workload-based optimization, the input workload plays a crucial role and needs to be a good representation of the expected Database systems often rely on historical query traces to perform workload. Traditionally, historical query traces have been used as workload-based performance tuning. However, real production input workloads with the assumption that workloads are mostly workloads are time-evolving, making historical queries ineffective static. However, as we discuss in 2, many real workloads exhibit for optimizing future workloads. To address this challenge, we propose highly recurring query structures with changing patterns in both Sibyl, an end-to-end machine learning-based framework that their arrival intervals and data accesses. For instance, query templates accurately forecasts a sequence of future queries, with the entire are often shared across users, teams, and applications, but query statements, in various prediction windows. Drawing insights may be customized with different parameter values to access varying from real-workloads, we propose template-based featurization techniques data at different points in time. Consider a log analysis query and develop a stacked-LSTM with an encoder-decoder architecture that reports errors for different devices and error types: "SELECT for accurate forecasting of query workloads. We also * FROM T WHERE deviceType =? AND errorType =? AND develop techniques to improve forecasting accuracy over large prediction eventDate BETWEEN?

As instruction-tuned large language models (LLMs) gain global adoption, their ability to follow instructions in multiple languages becomes increasingly crucial. One promising approach is cross-lingual transfer, where a model acquires specific functionality on some language by finetuning on another language. In this work, we investigate how multilinguality during instruction tuning of a multilingual LLM affects instruction-following across languages. We first show that many languages transfer some instruction-following capabilities to other languages from even monolingual tuning. Furthermore, we find that only 40 multilingual examples in an English tuning set substantially improve multilingual instruction-following, both in seen and unseen languages during tuning. In general, we observe that models tuned on multilingual mixtures exhibit comparable or superior performance in several languages compared to monolingually tuned models, despite training on 10x fewer examples in those languages. Finally, we find that increasing the number of languages in the instruction tuning set from 1 to only 2, 3, or 4 increases cross-lingual generalization. Our results suggest that building massively multilingual instruction-tuned models can be done with only a very small set of multilingual instruction-responses.