added reward function normalization, hidden eval set

added canonical evaluation harness, unified DIME index, deterministic replay guarantees

Update codebase and ignore DIME.pdf and image_gen binaries

chore: update codebase and fix gitignore

refined the reward algos

updating reward function from step to gradient

added alibaba trace data

adding ruberics,updating tasks,validators

debugging

fixing task score reset

refactoring file structure