docs: update image links in BLOG.md to point to raw GitHub URLs for better accessibility

feat: add canonical demo URLs and update root endpoint response

docs: update previous training run link in README for accuracy

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

docs: replace mermaid diagrams with PNG images for architecture and dispute lifecycle

Implement code changes to enhance functionality and improve performance

docs: enhance README with additional resources and demo links

chore: remove ChargebackOps project paper source file

Add Overleaf research paper draft

docs: correct HF Space URL to mitudrudutta/ChargeBackOps

Add Hugging Face blog writeup

Enhance documentation and address specification gaming in ChargebackOps

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)

chore(deps): add matplotlib to dev extras + sync uv.lock

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(training): SFT dataset + stall detection in eval rollout

fix(training): per-action reward scoring vs heuristic oracle

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

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

feat: tighten EscalationROI, add ambiguous medium case, LLM note judge wrapper

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

Add training notebook and benchmark runner for ChargebackOps

feat: Add LLM softening support to IssuerAgent and implement related tests

feat: Implement multi-round dispute lifecycle with arbitration scoring and related tests

feat: Implement Issuer agent for multi-round dispute lifecycle

refactor: tighten rubric discrimination + LLM path + add running doc

refactor: update difficulty levels and enhance scoring rubrics in documentation and code

Add documentation for core modules, data assets, evaluation components, runners, and tests

Refactor evidence building and improve code readability in iso_adapter.py

refactor: build grading on OpenEnv Rubric system

fix: correct list_reports ordering and track ISO dataset in git

fix: match structured output to hackathon spec format

fix: add [START]/[STEP]/[END] structured output to inference.py

fix: squash inflated evidence scores for wrongly contested concedable cases

feat: redesign Gradio demo with visual dashboard and update OPENENV.md

feat: add card network metadata, trim README, add limitations section

docs: sync AGENT.md and README with current grading values, add LICENSE

fix: replace silent suppress with logged warning for Gradio mount failure

fix: address Codex adversarial review findings

feat: add Gradio demo UI, realistic metadata, info dict, and metrics