Week-by-Week Progress

Week	Focus	Key Outcomes
1–2	Setup & Intro to VLMs	Ran first baseline experiments
3–4	Prompt Engineering Basics	Learned few-shot & CoT prompting
5–6	Float Value Stability	Designed parsing + clamping logic
7–8	Regional Consistency	Built region-specific scoring prompts
9–10	Determinism & Repeatability	Logged nondeterminism patterns
11	Human-like Grading	Mental mapping + region independence
12	Benchmarking with Trackers	W&B, Weave, Trackio dashboards integrated
13+	Fine-tuning & Feedback Loop	Starting dataset prep and calibration

---

Each week has been a step toward making AI grading stable, interpretable, and human-aligned.