Remove training notebook for ChargebackOps Merchant Agent, consolidating code and documentation for outcome-based RL on Qwen2.5-3B.

fix(notebook): rewrite diag-code as self-contained loader

fix(eval): sequential per-checkpoint base load + product-grade docs

perf(training): v5 - shorter SFT, longer GRPO, hard/nightmare oversample

fix(eval): pass offload_folder to from_pretrained on T4 OOM path

fix(grpo): revert num_generations to 8 (TRL gen_batch divisibility)

perf(notebook): shrink SFT+GRPO budget for Colab free T4

fix(eval): preload bases, swap adapters, fix peft offload_dir crash

fix(grpo): unblock learning - widen sampling + raise lr + keep dropout

fix(notebook): pin accelerate==1.0.1 to keep huggingface-hub at 0.26.x

feat(training): outcome-based RLVR reward + clean Colab T4 notebook

Implement code changes to enhance functionality and improve performance

feat: Implement wait_for_updates action for handling delayed cases and evidence

feat: enhance completion parsing to handle truncated JSON and `<think>` blocks

feat: add per-family evaluation and plotting for training curves

feat: Add training curve evaluation and plotting utilities with unit tests

Add training notebook and benchmark runner for ChargebackOps