sync: pull latest from main (model_server.py, captain LLM toggle in ui.py, 0.6B configs, SUBMISSION + RUNTIME_DURABILITY docs)

sync: today's source updates (XML-only prompt, reward unclip, neg-reward on loss, pinned versions, configs reorg)

illus: all 3 opponent modes verified across 5-over inference + train-smoke

docs: update README, slides, illustrations with format mapper, tool budget, player rosters, 5/20-over baselines

feat: all-12-tool UI, auto-play, timestamped run folders, model fix

feat: add field-aware hybrid ball simulation and timing logs

fix: tighten cricket rules and reward visibility

test: add 5-over training rollout smoke

fix: preserve match outcome reward objective

docs: clarify HF Gemma runs and OpenEnv traces

docs: add experiment workflow guide

docs: explain adaptive benchmark architecture