BankBench

Cetalabs · LLM safety & capability evals for financial / commerce agents

🛍️ Stripe Link CLI — Sandbox

Interactive character-driven walkthrough: Shopper Agent, Sandbox Clerk, and Judge. Replay real Gemini vs. Llama runs, or try a scenario yourself.

stripe-link-cli/stripe_link_cli_sandbox.html

📊 Stripe Link CLI — Eval Dashboard

Cetalabs dashboard for the Safe Commerce Benchmark: KPI cards, charts, transaction log, run history.

stripe-link-cli/stripe_link_dashboard.html

🏦 BankBench Sinar

Sandbox bank environment dashboard.

bankbench-sinar.html

📈 3–4 LLMs — Eval Dashboard

Fast BankBench eval results across multiple LLMs.

3-4 LLMs/bankbench_dashboard.html

🔬 3–4 LLMs — Mechinterp Dashboard

Mechanistic interpretability probes and code-switching analysis.

3-4 LLMs/bankbench_mechinterp_dashboard.html

🧩 Gemma 2 — BankBench Dashboard

Gemma 2 capability + mechinterp eval dashboard (20-sample run).

Gemma/gemma2_bankbench_dashboard.html