Abstract: A computer-implemented method, a computer program product, and a computer system for tuning large language models. A computer receives pairs of textual prompts and ground truth labels. A computer creates a data selection scoring function, by repurposing one or more reward functions to compute similarity between the textual prompts and the ground truth labels, where the one or more reward functions measure similarity between textual outputs produced by a large language model and the ground truth labels. A computer selects a training dataset from the pairs of the textual prompts and the ground truth labels, by using the data selection scoring function. A computer tunes the large language model using the training dataset and reinforcement learning with the one or more reward functions.

Abstract: A method, computer program product, and computer system for developing a fine-tuned generative model (GM) that receives metadata and multiple choices as input and is configured to select a best choice, from the multiple choices, for describing the metadata. The GM is a neural network of interconnected nodes with each node having an associated weight. Developing the fine-tuned GM includes: (i) generating training data as input for training a pre-trained GM to become the fine-tuned GM, the training data including P input prompts; and (ii) training the pre-trained GM to become the fine-tuned GM, using reinforcement learning and using the training data as input. The training includes: performing a nested iterative process in which N input prompts are sampled randomly from the P input prompts subject to N<P, followed by iteratively using the N input prompts to dynamically update the weights for a maximum of K iterations.

Abstract: A computer-implemented method, a computer program product, and a computer system for tuning large language models. A computer receives pairs of textual prompts and ground truth labels. A computer creates a data selection scoring function, by repurposing one or more reward functions to compute similarity between the textual prompts and the ground truth labels, where the one or more reward functions measure similarity between textual outputs produced by a large language model and the ground truth labels. A computer selects a training dataset from the pairs of the textual prompts and the ground truth labels, by using the data selection scoring function. A computer tunes the large language model using the training dataset and reinforcement learning with the one or more reward functions.

Abstract: Automated or autonomic techniques for managing deployment of one or more resources in a computing environment based on varying workload levels. The automated techniques may comprise predicting a future workload level based on data associated with the computing environment. Then, an estimation is performed to determine whether a current resource deployment is insufficient, sufficient, or overly sufficient to satisfy the future workload level. Then, one or more actions are caused to be taken when the current resource deployment is estimated to be insufficient or overly sufficient to satisfy the future workload level. Actions may comprise resource provisioning, resource tuning and/or admission control.

Abstract: Automated or autonomic techniques for managing deployment of one or more resources in a computing environment based on varying workload levels. The automated techniques may comprise predicting a future workload level based on data associated with the computing environment. Then, an estimation is performed to determine whether a current resource deployment is insufficient, sufficient, or overly sufficient to satisfy the future workload level. Then, one or more actions are caused to be taken when the current resource deployment is estimated to be insufficient or overly sufficient to satisfy the future workload level. Actions may comprise resource provisioning, resource tuning and/or admission control.

Abstract: Techniques for performing adaptive and robust prediction. Prediction techniques are adaptive in that they use a minimal amount of historical data to make predictions, the amount of data being selectable. The techniques are able to learn quickly about changes in the workload traffic pattern and make predictions, based on such learning, that are useful for proactive response to workload changes. To counter the increased variability in the prediction as a result of using minimal history, robustness is improved by checking model stability at every time interval and revising the model structure as needed to meet designated stability criteria. Furthermore, the short term prediction techniques can be used in conjunction with a long term forecaster.

Abstract: Automated or autonomic techniques for managing deployment of one or more resources in a computing environment based on varying workload levels. The automated techniques may comprise predicting a future workload level based on data associated with the computing environment. Then, an estimation is performed to determine whether a current resource deployment is insufficient, sufficient, or overly sufficient to satisfy the future workload level. Then, one or more actions are caused to be taken when the current resource deployment is estimated to be insufficient or overly sufficient to satisfy the future workload level. Actions may comprise resource provisioning, resource tuning and/or admission control.

Abstract: Techniques for performing adaptive and robust prediction. Prediction techniques are adaptive in that they use a minimal amount of historical data to make predictions, the amount of data being selectable. The techniques are able to learn quickly about changes in the workload traffic pattern and make predictions, based on such learning, that are useful for proactive response to workload changes. To counter the increased variability in the prediction as a result of using minimal history, robustness is improved by checking model stability at every time interval and revising the model structure as needed to meet designated stability criteria. Furthermore, the short term prediction techniques can be used in conjunction with a long term forecaster.

Abstract: The present invention provides methods, devices and systems for dynamically adjusting transcoding parameters so as to increase the benefits of transcoding. Methods of adaptation are designed to cope with the variability of network characteristics and of the size of transcoded images. The invention also provides a method and apparatus to enable the transcoding proxy to adjust a quality-size tradeoff on a per-image and/or a per-client basis. The adaptive transcoder chooses different parameters for each object, and provides performance improvements. The invention further provides a general framework for making policy decisions taking into account available bandwidth, content and type of image, and user preferences. The invention also includes methods for generating feedback about the choice of optimal transcoding parameters to the user.