benchflow-ai / awesome-evals

What it does

• Asymmetry of Verification and Verifier's Law. Jason Wei — · *blog* — Trainability tracks verifiability; verifying = creating an RL environment.
• A Taxonomy of RL Environments for LLM Agents. Han-Chung Lee — · *blog* — A benchmark is a frozen RL environment; the E = {T,H,V,S,C} decomposition; "verifiable beats judgeable."
• The Life Cycle of an RL Environment. Kanav Garg (Core Automation; ex-DeepMind) — talk; summary at · *talk* — Difficulty calibration (the 1–4/16 Goldilocks band), RL as variance reduction, reward hacking under training pressure. *(local not
• Welcome to the Era of Experience. David Silver & Richard Sutton — · *paper* — Human-data value approaching its ceiling; the frontier is agents learning from experience / synthetic environments.
• RLHF Book, Ch. 16. Evaluation — Nathan Lambert — · *book* — Evaluation as a reflection of training goals; prompt-format sensitivity (60%→~0%).