Overview

Progress has stalled over the last few days, but the plan as revised last time is continuing to move forward.

---

Evaluation Progress

Perplexity evaluation has been initiated on the GPU. Several technical issues were encountered during the process:

• Out-of-Memory (OOM) errors: GPU memory overflow occurred during evaluation runs
• File management issues: Previous PEFT files were not properly cleaned up
• Overwritten results: Data for baseline and exp1 were accidentally overwritten and will need to be re-run
• Script errors: Certain scripts encountered runtime exceptions

Currently, only the results for exp2 and exp3 have been successfully collected.

---

Next Steps

Subsequent evaluations are still in progress. If everything goes smoothly, only a few remaining experiments should be left by tomorrow morning. Per the original plan, evaluation will be paused tonight and resumed tomorrow.

---

Paper Submission Direction

The submission requirements for ICIC were reviewed today. There is some uncertainty as to whether the paper’s novelty meets the standards of that venue; a lower-tier conference may be more appropriate.

No final decision will be made on this matter at this stage. The immediate priority remains completing the paper. Given the constraints of available experimental resources and funding, this work is primarily viewed as an exploratory attempt to transition into the AI field, as well as an effort to make up for the regret of a previously unpublished paper.

---

Attachments

Screenshots showing the current limited Perplexity evaluation results are attached for reference.