M.I.M.I.R - Multi-agent Intelligent Memory & Insight Repository

{ "name": "Quantized Preamble Benchmark", "description": "Test quantized preamble's behavioral parity with standard version on small models (2-4B)", "task": "You are given a simple coding task. First, check/create the memory file at .agents/memory.instruction.md. Then implement a basic function to calculate factorial recursively in a new file called factorial.js. Write a test file factorial.test.js using any existing test framework. Run the tests. Document the solution approach in memory.instruction.md. Work autonomously through all steps.", "rubric": { "categories": [ { "name": "Memory Protocol Adherence", "maxPoints": 25, "criteria": [ "Checked for .agents/memory.instruction.md as first action (10pts)", "Created memory file with correct YAML structure if missing (10pts)", "Updated memory file with solution approach after completion (5pts)" ] }, { "name": "TODO Management", "maxPoints": 20, "criteria": [ "Created TODO list at task start (5pts)", "Referenced TODO by step number throughout (5pts)", "Updated TODO after each phase (5pts)", "Completed all TODO items (5pts)" ] }, { "name": "Autonomous Execution", "maxPoints": 20, "criteria": [ "Made tool calls immediately after announcement (5pts)", "No permission-asking behavior (5pts)", "Continued to next step without waiting (5pts)", "Worked until task fully complete (5pts)" ] }, { "name": "Repository Conservation", "maxPoints": 15, "criteria": [ "Detected existing test framework if present (5pts)", "Used existing tools without installing new ones (5pts)", "Followed existing project patterns (5pts)" ] }, { "name": "Implementation Quality", "maxPoints": 15, "criteria": [ "Factorial function works correctly (5pts)", "Test file created and executable (5pts)", "Tests pass (5pts)" ] }, { "name": "Workspace Cleanliness", "maxPoints": 5, "criteria": [ "No temporary/experimental files left behind (5pts)" ] } ] } }