fix: scrub API keys from evaluation scripts for security

chore: remove hardcoded API keys and replace with environment variable loading

feat: add batched evaluation script for Gemini to bypass rate limits

chore: switch default Gemini model to gemini-1.5-flash for better free-tier limits

chore: switch default Gemini evaluation model to gemini-3.1-flash-lite

fix: add smart rate limit handling parsing wait_time from 429 errors for Gemini eval

feat: add eval_gemini.py to evaluate Gemini models via API

feat: rewrite eval_gonyai.py as full Kaggle-ready script matching colab_eval.py harness

fix: address ACL reviewer issues — non-monotonic scaling, OOD label, Gonyai/Gemma/Aya paragraphs, Section ref

feat: scatter 31 handcrafted items across 4 modules → 3,031 total items

feat: backfill 37 handcrafted items to reach exactly 3,000 items and update paper

feat: add comprehensive statistical, validation, and diagnostic scripts

fix: implement conditional probability scoring

fix: shuffle MCQ candidates, fix length bias, add difficulty tracking

feat: adapt eval script for Kaggle environment and 19GB disk limits

fix: enforce GPU 0 device_map and enable CPU offload for 8-bit models to bypass conservative memory dispatcher

fix: use BitsAndBytesConfig for newer transformers and clear disk cache to prevent IO Error

fix: total_mem -> total_memory (correct PyTorch API)

feat: v4.0 crash-safe eval suite — checkpoint/resume, aggressive VRAM cleanup, 12 mainstream models

docs: rewrite paper to professional 2-column format with expanded analysis; feat: add Google Colab evaluation script

docs: fix pdflatex fontspec crash by using romanized Konkani

fix: removed all duplicate candidates across the benchmark using rigorous multi-pass review

fix: scrubbed all english translation hints from json files