Auto-Arch Tournament vs evalmonkey

Auto-Arch Tournament is an autonomous research system where an AI agent iteratively proposes, implements, and validates microarchitectural improvements to a RISC-V CPU. Starting from a standard 5-stage pipeline, the loop runs hypotheses in parallel, each going through formal verification (53 symbolic checks), cycle-accurate simulation, multi-seed FPGA place-and-route, and CoreMark CRC validation. Only hypotheses that beat the current champion get merged; everything else gets discarded. Starting from 301 iterations/second, the system hit 577 iter/s (+92%) across 73 attempts in 9.8 hours — producing a design 26% faster and 40% smaller in LUTs than the baseline. The insight the author drives home is that the real innovation isn't the AI agent — it's the verifier. The orchestrator is hardcoded to prevent agents from manipulating their own evaluation gates, a simple but critical design constraint that turns a creative process into a trustworthy one. Without a rigorous verification harness, agent-driven optimization becomes a confidence trick. This is early but fascinating proof that AI-driven hardware design loops can produce commercially meaningful gains. The repo uses Claude Code or Codex as the coding agent, SystemVerilog for the RTL, and standard open-source EDA tooling (Yosys, nextpnr, Verilator). It's a compelling template for anyone building agentic optimization loops where correctness matters.

evalmonkey is an open-source framework for testing how LLM agents degrade under adversarial conditions. You run your agent against 10 standard datasets (GSM8K, ARC, HellaSwag, etc.) pulled automatically from HuggingFace, then apply chaos profiles that introduce realistic failure modes: malformed JSON schemas, artificial latency spikes, 429 rate-limit errors, context-window overflow, and prompt injection payloads. The key output is a degradation delta — evalmonkey shows you exactly how much your agent's accuracy drops under each failure type versus clean inputs. A model that scores 78% on GSM8K normally but drops to 31% when it gets a 429 mid-chain tells you something crucial about its error-recovery behavior that standard benchmarks completely miss. It supports OpenAI, Anthropic (via Bedrock and direct), Azure, GCP, and any Ollama-hosted model. Corbell-AI published this with a clear thesis: agents break in production for infrastructure reasons, not model reasons — and no existing benchmark tests that. evalmonkey was created today (April 17, 2026) and is still at 3 stars, but the core idea is genuinely novel in the evals space.