Orientation week. Met my advisor and fellow REU participants. Read background papers on tensor decomposition, LoRA and began to set up the baseline model. Attended seminar talks on quantum matrix states and fair division.

Presented to fellow REU participants on project - slides linked here! Established a benchmark baseline to find the validation metrics of which we want to compare our findings with.

Post-trained models using LoRA and established baseline accuracy results to compare future fine-tuning tensor compression methods to. Began a literature review on existing implementations of tensor-rank adapatations.

Narrowed existing methodology on tensor-based LoRA adaptations. Reviewed hueristics to consider and did background reading on architectural elements of transformers. Further solidified understanding of adapter and reparameterization-based PEFT methods.

Solidified two potential paths to further explore with regards to improving a specific tensor-based algorithm through LSR and federated learning. Developed a general argument for low seperation rank implementation in varying aspects, and the improvements theorized to occur from them.

Proposed framework for LSR and general BTD approaches to tensor decomposition in PEFTs. Explored various ways of forming tensors from various-dimension matrices, and outlined generalizations to consider when implementing adapatations on hybrid architectural models.