Global ETD Search
Search theses and dissertations gathered from participating repositories worldwide. Every result links back to the library that holds it. No account is needed.
Results
Showing 1 to 12 of 12 for “"LLM Training"”.
-
Systematic Explorations for Data-Efficient LLM Training
Large language models (LLMs) have demonstrated remarkable performance across many downstream domains. When training these models, the first phase of pretraining involves learning to predict the next token across massive amounts of web-scale text data. While the ever-increasing scale of pretraining …
-
Efficient LLM training and inference with contextual sparsity
Submission published under a 24 month embargo labeled 'U of I Access', the embargo will last until 2027-05-01
-
AUDITING LARGE LANGUAGE MODELS: ACCOUNTABILITY, INTEGRITY, AND COPYRIGHT
… thesis addresses critical Large Language Model (LLM) governance challenges by proposing three technical solutions. First, to enhance content traceability, we introduce a novel watermarking method based on pseudo-random segment assignment, achieving superior extraction accuracy and time efficiency …
-
Improving Collective Aggregation Strategies for HPC and AI Workloads
… domains: large-scale HPC checkpointing, LLM training, and LLM inference. We first show that collective I/O techniques designed for synchronous checkpointing are poorly suited to asynchronous execution due to their reliance on globally coordinated synchronization. Through an extensive …
-
Scaling the Efficiency of Large-Scale Learning with Structured Representations
… data. In Part 3, I consider the problem of LLM quantization. I first show that by taking advantage of both the geometric structure of the model weights as well as the model loss landscape, we can compress models to as little as 1/8 of their original size without catastrophic loss of …
-
Evaluating and Enhancing Large Language Models (LLMs) in the Clinical Domain
Recent advancements in large language models (LLMs) have demonstrated human-level performance on many specialized medical tasks, even without annotated training data. However, three main challenges remain: (1) due to the sensitive and highly specialized nature of clinical narratives, as well as the …
-
Multi-Layer OTN Simulation and LLM-Driven Root Cause Analysis: From Alarm Propagation to Reinforcement-Optimized Diagnostic Agents
… framework for OTN fault simulation and LLM-based root-cause analysis. The electrical-layer–centric simulator decomposes each network element into typed functional boards (Tributary, XCON, Line, OA, OD, OM, FIU) connected by directed dependency edges along the Service Function Chain. An …
-
STOCHASTIC OPTIMIZATION FOR DISTRIBUTED TRAINING AND MULTI-AGENT INTERACTION IN LARGE LANGUAGE MODELS
… develops stochastic optimization frameworks for training and coordinating Large Language Models (LLMs) in distributed settings. We study two interconnected problems: (i) controlling the training dynamics of LLMs across distributed datasets, and (ii) modeling and controlling emergent behaviors …
-
Transitioning to Inverter-based Resources: Stability of AC Power Systems considering Current Constraints, Small-signal Sensitivities, and Large Loads
… caused by the load variations typical of LLM training tasks. Collectively, this work provides an overview of a range of stability-related challenges that must be considered when integrating IBRs into modern, large-scale power systems.
-
Navigating the Ocean of Language Model Training Data
… step toward understanding large language models (LLMs) is to understand their training data. Modern LLMs are trained on text corpora with trillions of tokens, hindering them from being easily analyzed. In this thesis, I discuss my research on making these massive text corpora efficiently …
-
Methodologies, Architectures, and Prototypes for Scaling On- and Off-Chip Interconnects
… for machine learning workloads. I first present LLMCompass-E2E, a comprehensive framework for modeling large-scale distributed LLM training performance. I then explore the potential of emerging co-packaged silicon photonic interconnects by proposing an optically connected multi-stack HBM module …
-
Poisoning Attacks in Deep Learning Security
… such as data poisoning, which compromises training data, and model poisoning, which manipulates the training process, are among the most pressing threats, aiming either to degrade performance or insert hidden backdoors. Backdoored models behave normally under typical conditions but produce …