diff --git a/agent-evolution/data/real-fit-report.json b/agent-evolution/data/real-fit-report.json index f3b3287..4647826 100644 --- a/agent-evolution/data/real-fit-report.json +++ b/agent-evolution/data/real-fit-report.json @@ -1,7 +1,7 @@ { - "generated": "2026-05-28T12:17:26.011791+00:00", - "source": "real-fit-engine", - "total_evaluations": 150, + "generated": "2026-05-28T12:57:20Z", + "source": "real-fit-engine-db-filtered", + "total_evaluations": 144, "agents": { "agent-architect": { "name": "agent-architect", @@ -9,7 +9,6 @@ "deepseek-v4-pro": 48.3, "glm-5.1": 48.3, "kimi-k2.6": 53.5, - "minimax-m2.5": 30.9, "qwen3-coder:480b": 48.3 }, "info": [ @@ -31,7 +30,7 @@ "info": [ "Indexes and maps project codebase architecture into .architect/ directory. Creates and maintains structured documentation of entities, APIs, DB schema, file graphs, and conventions. (GNS-2 Tier 0)", "core", - "ollama-cloud/glm-5.1" + "ollama-cloud/qwen3-coder:480b" ], "best_model": "qwen3-coder:480b", "best_score": 54.0 @@ -47,7 +46,7 @@ "info": [ "Backend specialist for Node.js, Express, APIs, and database integration (GNS-2 Tier 1)", "core", - "ollama-cloud/qwen3-coder:480b" + "ollama-cloud/deepseek-v4-pro" ], "best_model": "deepseek-v4-pro", "best_score": 53.5 @@ -63,7 +62,7 @@ "info": [ "Browser automation agent using Playwright MCP for E2E testing, form filling, navigation, and web interaction (GNS-2 Tier 0)", "testing", - "ollama-cloud/deepseek-v4-flash" + "ollama-cloud/kimi-k2.6" ], "best_model": "kimi-k2.6", "best_score": 63.8 @@ -90,13 +89,12 @@ "deepseek-v4-pro": 22.8, "glm-5.1": 89.1, "kimi-k2.6": 91.2, - "minimax-m2.5": 45.0, "qwen3-coder:480b": 90.6 }, "info": [ "Adversarial code reviewer. Finds problems and issues. Does NOT suggest implementations (GNS-2 Tier 0)", "quality", - "ollama-cloud/minimax-m2.5" + "ollama-cloud/kimi-k2.6" ], "best_model": "kimi-k2.6", "best_score": 91.2 @@ -112,7 +110,7 @@ "info": [ "DevOps specialist for Docker, Kubernetes, CI/CD pipeline automation, and infrastructure management (GNS-2 Tier 1)", "core", - "ollama-cloud/kimi-k2.6" + "ollama-cloud/glm-5.1" ], "best_model": "glm-5.1", "best_score": 96.2 @@ -128,43 +126,11 @@ "info": [ "Scores agent effectiveness after task completion for continuous improvement. Tier 2 meta-agent with self-cascade enabled.", "meta", - "ollama-cloud/qwen3.5-122b" + "ollama-cloud/glm-5.1" ], "best_model": "glm-5.1", "best_score": 58.7 }, - "evolution-prompt": { - "name": "evolution-prompt", - "evaluations": { - "deepseek-v4-pro": 52.6, - "glm-5.1": 44.7, - "kimi-k2.6": 53.5, - "qwen3-coder:480b": 21.3 - }, - "info": [ - "Generates role-specific stress-test prompts by analyzing agent definitions", - "meta", - "ollama-cloud/deepseek-v4-pro" - ], - "best_model": "kimi-k2.6", - "best_score": 53.5 - }, - "evolution-skeptic": { - "name": "evolution-skeptic", - "evaluations": { - "deepseek-v4-pro": 33.1, - "glm-5.1": 31.6, - "kimi-k2.6": 37.3, - "qwen3-coder:480b": 42.9 - }, - "info": [ - "Evaluates model responses against role-specific rubrics with detailed scoring and commentary", - "meta", - "ollama-cloud/deepseek-v4-pro" - ], - "best_model": "qwen3-coder:480b", - "best_score": 42.9 - }, "flutter-developer": { "name": "flutter-developer", "evaluations": { @@ -176,7 +142,7 @@ "info": [ "Flutter mobile specialist for cross-platform apps, state management, and UI components (GNS-2 Tier 1)", "core", - "ollama-cloud/qwen3-coder:480b" + "ollama-cloud/glm-5.1" ], "best_model": "glm-5.1", "best_score": 54.9 @@ -187,13 +153,12 @@ "deepseek-v4-pro": 31.6, "glm-5.1": 53.2, "kimi-k2.6": 38.8, - "minimax-m2.5": 48.3, "qwen3-coder:480b": 56.0 }, "info": [ "Handles UI implementation with multimodal capabilities. Accepts visual references like screenshots and mockups (GNS-2 Tier 1)", "core", - "ollama-cloud/minimax-m2.5" + "ollama-cloud/qwen3-coder:480b" ], "best_model": "qwen3-coder:480b", "best_score": 56.0 @@ -209,7 +174,7 @@ "info": [ "Go backend specialist for Gin, Echo, APIs, and database integration (GNS-2 Tier 1)", "core", - "ollama-cloud/deepseek-v4-pro" + "ollama-cloud/qwen3-coder:480b" ], "best_model": "qwen3-coder:480b", "best_score": 58.7 @@ -225,7 +190,7 @@ "info": [ "Analyzes git history to find duplicates and past solutions, preventing regression and duplicate work (GNS-2 Tier 0)", "core", - "ollama-cloud/qwen3.5-122b" + "ollama-cloud/kimi-k2.6" ], "best_model": "kimi-k2.6", "best_score": 46.9 @@ -241,7 +206,7 @@ "info": [ "Server incident response and system hardening specialist. Handles live forensics, malware removal, persistence hunting, SSH-based server cleanup, and post-incident hardening. Works with any OS and panel.", "core", - "ollama-cloud/kimi-k2.6" + "ollama-cloud/glm-5.1" ], "best_model": "glm-5.1", "best_score": 65.6 @@ -257,7 +222,7 @@ "info": [ "Primary code writer for backend and core logic. Writes implementation to pass tests (GNS-2 Tier 1)", "core", - "ollama-cloud/qwen3-coder:480b" + "ollama-cloud/kimi-k2.6" ], "best_model": "kimi-k2.6", "best_score": 72.5 @@ -273,7 +238,7 @@ "info": [ "Validates and corrects Markdown descriptions for Gitea issues (GNS-2 Tier 0)", "meta", - "ollama-cloud/nemotron-3-nano" + "ollama-cloud/qwen3-coder:480b" ], "best_model": "qwen3-coder:480b", "best_score": 47.4 @@ -289,7 +254,7 @@ "info": [ "Manages agent memory systems - short-term (context), long-term (vector store), and episodic (experiences) (GNS-2 Tier 0)", "cognitive", - "ollama-cloud/deepseek-v4-pro" + "ollama-cloud/glm-5.1" ], "best_model": "glm-5.1", "best_score": 48.3 @@ -297,11 +262,9 @@ "orchestrator": { "name": "orchestrator", "evaluations": { - "deepseek-v4-flash": 27.0, "deepseek-v4-pro": 19.6, "glm-5.1": 36.2, "kimi-k2.6": 40.0, - "minimax-m2.5": 36.3, "qwen3-coder:480b": 39.1 }, "info": [ @@ -323,7 +286,7 @@ "info": [ "Reviews code for performance issues. Focuses on efficiency, N+1 queries, memory leaks, and algorithmic complexity (GNS-2 Tier 0)", "quality", - "ollama-cloud/deepseek-v4-pro" + "ollama-cloud/glm-5.1" ], "best_model": "glm-5.1", "best_score": 63.8 @@ -339,7 +302,7 @@ "info": [ "PHP backend specialist for Laravel, Symfony, WordPress, and full-stack web applications (GNS-2 Tier 1)", "core", - "ollama-cloud/qwen3-coder:480b" + "ollama-cloud/deepseek-v4-pro" ], "best_model": "deepseek-v4-pro", "best_score": 53.5 @@ -355,7 +318,7 @@ "info": [ "Automated pipeline judge. Evaluates workflow execution by running tests, measuring token cost and wall-clock time. Produces objective fitness scores. Never writes code - only measures and scores. (GNS-2 Tier 0)", "meta", - "ollama-cloud/kimi-k2.6" + "ollama-cloud/qwen3-coder:480b" ], "best_model": "qwen3-coder:480b", "best_score": 52.9 @@ -387,7 +350,7 @@ "info": [ "Manages issue checklists, status labels, tracks progress and coordinates with human users (GNS-2 Tier 1)", "meta", - "ollama-cloud/glm-5.1" + "ollama-cloud/kimi-k2.6" ], "best_model": "kimi-k2.6", "best_score": 34.6 @@ -403,7 +366,7 @@ "info": [ "Improves agent system prompts based on performance failures. Meta-learner for prompt optimization (GNS-2 Tier 1)", "meta", - "ollama-cloud/qwen3.5-122b" + "ollama-cloud/glm-5.1" ], "best_model": "glm-5.1", "best_score": 48.3 @@ -419,7 +382,7 @@ "info": [ "Python backend specialist for Django, FastAPI, data science, and API development (GNS-2 Tier 1)", "core", - "ollama-cloud/qwen3-coder:480b" + "ollama-cloud/deepseek-v4-pro" ], "best_model": "deepseek-v4-pro", "best_score": 48.3 @@ -435,7 +398,7 @@ "info": [ "Self-reflection agent using Reflexion pattern - learns from mistakes (GNS-2 Tier 0)", "cognitive", - "ollama-cloud/deepseek-v4-pro" + "ollama-cloud/kimi-k2.6" ], "best_model": "kimi-k2.6", "best_score": 58.7 @@ -467,7 +430,7 @@ "info": [ "Converts vague ideas and bug reports into strict User Stories with acceptance criteria checklists (GNS-2 Tier 1)", "core", - "ollama-cloud/kimi-k2-thinking" + "ollama-cloud/qwen3-coder:480b" ], "best_model": "qwen3-coder:480b", "best_score": 45.3 @@ -478,13 +441,12 @@ "deepseek-v4-pro": 58.7, "glm-5.1": 86.0, "kimi-k2.6": 97.0, - "minimax-m2.5": 53.5, "qwen3-coder:480b": 97.0 }, "info": [ "Writes tests following TDD methodology. Tests MUST fail initially (Red phase) (GNS-2 Tier 1)", "core", - "ollama-cloud/qwen3-coder:480b" + "ollama-cloud/kimi-k2.6" ], "best_model": "kimi-k2.6", "best_score": 97.0 @@ -500,7 +462,7 @@ "info": [ "Scans for security vulnerabilities, OWASP Top 10, dependency CVEs, and hardcoded secrets (GNS-2 Tier 0)", "quality", - "ollama-cloud/deepseek-v4-pro" + "ollama-cloud/kimi-k2.6" ], "best_model": "kimi-k2.6", "best_score": 63.8 @@ -516,7 +478,7 @@ "info": [ "Designs technical specifications, data schemas, and API contracts before implementation (GNS-2 Tier 1)", "core", - "ollama-cloud/deepseek-v4-pro" + "ollama-cloud/kimi-k2.6" ], "best_model": "kimi-k2.6", "best_score": 92.0 @@ -532,7 +494,7 @@ "info": [ "Iteratively fixes bugs based on specific error reports and test failures (GNS-2 Tier 1)", "quality", - "ollama-cloud/kimi-k2.6" + "ollama-cloud/glm-5.1" ], "best_model": "glm-5.1", "best_score": 46.6 @@ -548,7 +510,7 @@ "info": [ "Visual regression testing agent that compares screenshots and detects UI differences using pixelmatch and image diff (GNS-2 Tier 0)", "quality", - "ollama-cloud/qwen3-coder:480b" + "ollama-cloud/glm-5.1" ], "best_model": "glm-5.1", "best_score": 58.7 @@ -564,7 +526,7 @@ "info": [ "Creates and maintains workflow definitions with complete architecture, Gitea integration, and quality gates (GNS-2 Tier 1)", "meta", - "ollama-cloud/qwen3.5-122b" + "ollama-cloud/glm-5.1" ], "best_model": "glm-5.1", "best_score": 48.3 @@ -580,192 +542,224 @@ "info": [ "Workflow cross-checker and process inspector. Analyzes inter-agent interaction logic, prevents conflicting tasks between agents, validates conformance to project architecture, tracks current state, and asks uncomfortable but important questions before expensive work begins.", "meta", - "ollama-cloud/kimi-k2.6" + "ollama-cloud/qwen3-coder:480b" ], "best_model": "qwen3-coder:480b", "best_score": 65.6 + }, + "evolution-skeptic": { + "name": "evolution-skeptic", + "evaluations": { + "deepseek-v4-pro": 33.1, + "glm-5.1": 31.6, + "kimi-k2.6": 37.3, + "qwen3-coder:480b": 42.9 + }, + "info": [ + "Evaluates model responses against role-specific rubrics with detailed scoring and commentary", + "meta", + "ollama-cloud/qwen3-coder:480b" + ], + "best_model": "qwen3-coder:480b", + "best_score": 42.9 + }, + "evolution-prompt": { + "name": "evolution-prompt", + "evaluations": { + "deepseek-v4-pro": 52.6, + "glm-5.1": 44.7, + "kimi-k2.6": 53.5, + "qwen3-coder:480b": 21.3 + }, + "info": [ + "Generates role-specific stress-test prompts by analyzing agent definitions", + "meta", + "ollama-cloud/kimi-k2.6" + ], + "best_model": "kimi-k2.6", + "best_score": 53.5 } }, "fit_scores": { "agent-architect": { - "model": "kimi-k2.6", - "fit": 53.5, - "explanation": "Best model for agent-architect is kimi-k2.6 with avg score 53.5. Strongest dimension: code_presence." + "model": "qwen3-coder:480b", + "fit": 48.3, + "explanation": "Best model for agent-architect is qwen3-coder:480b with avg score 48.3. Evaluator preference: evolution-skeptic > rubric_v2 > rubric_v1 (ignored HTTP errors)." }, "architect-indexer": { "model": "qwen3-coder:480b", "fit": 54.0, - "explanation": "Best model for architect-indexer is qwen3-coder:480b with avg score 54.0. Strongest dimension: code_presence." + "explanation": "Best model for architect-indexer is qwen3-coder:480b with avg score 54.0. Evaluator preference: evolution-skeptic > rubric_v2 > rubric_v1 (ignored HTTP errors)." }, "backend-developer": { - "model": "deepseek-v4-pro", - "fit": 53.5, - "explanation": "Best model for backend-developer is deepseek-v4-pro with avg score 53.5. Strongest dimension: code_presence." + "model": "qwen3-coder:480b", + "fit": 43.2, + "explanation": "Best model for backend-developer is qwen3-coder:480b with avg score 43.2. Evaluator preference: evolution-skeptic > rubric_v2 > rubric_v1 (ignored HTTP errors)." }, "browser-automation": { - "model": "kimi-k2.6", - "fit": 63.8, - "explanation": "Best model for browser-automation is kimi-k2.6 with avg score 63.8. Strongest dimension: code_presence." + "model": "qwen3-coder:480b", + "fit": 48.9, + "explanation": "Best model for browser-automation is qwen3-coder:480b with avg score 48.9. Evaluator preference: evolution-skeptic > rubric_v2 > rubric_v1 (ignored HTTP errors)." }, "capability-analyst": { - "model": "deepseek-v4-pro", - "fit": 58.7, - "explanation": "Best model for capability-analyst is deepseek-v4-pro with avg score 58.7. Strongest dimension: code_presence." + "model": "qwen3-coder:480b", + "fit": 52.3, + "explanation": "Best model for capability-analyst is qwen3-coder:480b with avg score 52.3. Evaluator preference: evolution-skeptic > rubric_v2 > rubric_v1 (ignored HTTP errors)." }, "code-skeptic": { - "model": "kimi-k2.6", - "fit": 91.2, - "explanation": "Best model for code-skeptic is kimi-k2.6 with avg score 91.2. Strongest dimension: code_presence." + "model": "qwen3-coder:480b", + "fit": 90.6, + "explanation": "Best model for code-skeptic is qwen3-coder:480b with avg score 90.6. Evaluator preference: evolution-skeptic > rubric_v2 > rubric_v1 (ignored HTTP errors)." }, "devops-engineer": { - "model": "glm-5.1", - "fit": 96.2, - "explanation": "Best model for devops-engineer is glm-5.1 with avg score 96.2. Strongest dimension: keyword_coverage." + "model": "qwen3-coder:480b", + "fit": 87.2, + "explanation": "Best model for devops-engineer is qwen3-coder:480b with avg score 87.2. Evaluator preference: evolution-skeptic > rubric_v2 > rubric_v1 (ignored HTTP errors)." }, "evaluator": { - "model": "glm-5.1", - "fit": 58.7, - "explanation": "Best model for evaluator is glm-5.1 with avg score 58.7. Strongest dimension: code_presence." + "model": "qwen3-coder:480b", + "fit": 43.8, + "explanation": "Best model for evaluator is qwen3-coder:480b with avg score 43.8. Evaluator preference: evolution-skeptic > rubric_v2 > rubric_v1 (ignored HTTP errors)." }, "evolution-prompt": { - "model": "kimi-k2.6", - "fit": 53.5, - "explanation": "Best model for evolution-prompt is kimi-k2.6 with avg score 53.5. Strongest dimension: code_presence." + "model": "qwen3-coder:480b", + "fit": 21.3, + "explanation": "Best model for evolution-prompt is qwen3-coder:480b with avg score 21.3. Evaluator preference: evolution-skeptic > rubric_v2 > rubric_v1 (ignored HTTP errors)." }, "evolution-skeptic": { "model": "qwen3-coder:480b", "fit": 42.9, - "explanation": "Best model for evolution-skeptic is qwen3-coder:480b with avg score 42.9. Strongest dimension: structure." + "explanation": "Best model for evolution-skeptic is qwen3-coder:480b with avg score 42.9. Evaluator preference: evolution-skeptic > rubric_v2 > rubric_v1 (ignored HTTP errors)." }, "flutter-developer": { - "model": "glm-5.1", + "model": "qwen3-coder:480b", "fit": 54.9, - "explanation": "Best model for flutter-developer is glm-5.1 with avg score 54.9. Strongest dimension: code_presence." + "explanation": "Best model for flutter-developer is qwen3-coder:480b with avg score 54.9. Evaluator preference: evolution-skeptic > rubric_v2 > rubric_v1 (ignored HTTP errors)." }, "frontend-developer": { "model": "qwen3-coder:480b", "fit": 56.0, - "explanation": "Best model for frontend-developer is qwen3-coder:480b with avg score 56.0. Strongest dimension: code_presence." + "explanation": "Best model for frontend-developer is qwen3-coder:480b with avg score 56.0. Evaluator preference: evolution-skeptic > rubric_v2 > rubric_v1 (ignored HTTP errors)." }, "go-developer": { "model": "qwen3-coder:480b", "fit": 58.7, - "explanation": "Best model for go-developer is qwen3-coder:480b with avg score 58.7. Strongest dimension: code_presence." + "explanation": "Best model for go-developer is qwen3-coder:480b with avg score 58.7. Evaluator preference: evolution-skeptic > rubric_v2 > rubric_v1 (ignored HTTP errors)." }, "history-miner": { - "model": "kimi-k2.6", - "fit": 46.9, - "explanation": "Best model for history-miner is kimi-k2.6 with avg score 46.9. Strongest dimension: code_presence." + "model": "qwen3-coder:480b", + "fit": 44.8, + "explanation": "Best model for history-miner is qwen3-coder:480b with avg score 44.8. Evaluator preference: evolution-skeptic > rubric_v2 > rubric_v1 (ignored HTTP errors)." }, "incident-responder": { - "model": "glm-5.1", - "fit": 65.6, - "explanation": "Best model for incident-responder is glm-5.1 with avg score 65.6. Strongest dimension: code_presence." + "model": "qwen3-coder:480b", + "fit": 56.4, + "explanation": "Best model for incident-responder is qwen3-coder:480b with avg score 56.4. Evaluator preference: evolution-skeptic > rubric_v2 > rubric_v1 (ignored HTTP errors)." }, "lead-developer": { - "model": "kimi-k2.6", + "model": "qwen3-coder:480b", "fit": 72.5, - "explanation": "Best model for lead-developer is kimi-k2.6 with avg score 72.5. Strongest dimension: keyword_coverage." + "explanation": "Best model for lead-developer is qwen3-coder:480b with avg score 72.5. Evaluator preference: evolution-skeptic > rubric_v2 > rubric_v1 (ignored HTTP errors)." }, "markdown-validator": { "model": "qwen3-coder:480b", "fit": 47.4, - "explanation": "Best model for markdown-validator is qwen3-coder:480b with avg score 47.4. Strongest dimension: code_presence." + "explanation": "Best model for markdown-validator is qwen3-coder:480b with avg score 47.4. Evaluator preference: evolution-skeptic > rubric_v2 > rubric_v1 (ignored HTTP errors)." }, "memory-manager": { - "model": "glm-5.1", - "fit": 48.3, - "explanation": "Best model for memory-manager is glm-5.1 with avg score 48.3. Strongest dimension: code_presence." + "model": "qwen3-coder:480b", + "fit": 46.8, + "explanation": "Best model for memory-manager is qwen3-coder:480b with avg score 46.8. Evaluator preference: evolution-skeptic > rubric_v2 > rubric_v1 (ignored HTTP errors)." }, "orchestrator": { - "model": "kimi-k2.6", - "fit": 40.0, - "explanation": "Best model for orchestrator is kimi-k2.6 with avg score 40.0. Strongest dimension: code_presence." + "model": "qwen3-coder:480b", + "fit": 39.1, + "explanation": "Best model for orchestrator is qwen3-coder:480b with avg score 39.1. Evaluator preference: evolution-skeptic > rubric_v2 > rubric_v1 (ignored HTTP errors)." }, "performance-engineer": { - "model": "glm-5.1", - "fit": 63.8, - "explanation": "Best model for performance-engineer is glm-5.1 with avg score 63.8. Strongest dimension: code_presence." + "model": "qwen3-coder:480b", + "fit": 36.3, + "explanation": "Best model for performance-engineer is qwen3-coder:480b with avg score 36.3. Evaluator preference: evolution-skeptic > rubric_v2 > rubric_v1 (ignored HTTP errors)." }, "php-developer": { - "model": "deepseek-v4-pro", - "fit": 53.5, - "explanation": "Best model for php-developer is deepseek-v4-pro with avg score 53.5. Strongest dimension: code_presence." + "model": "qwen3-coder:480b", + "fit": 48.3, + "explanation": "Best model for php-developer is qwen3-coder:480b with avg score 48.3. Evaluator preference: evolution-skeptic > rubric_v2 > rubric_v1 (ignored HTTP errors)." }, "pipeline-judge": { "model": "qwen3-coder:480b", "fit": 52.9, - "explanation": "Best model for pipeline-judge is qwen3-coder:480b with avg score 52.9. Strongest dimension: code_presence." + "explanation": "Best model for pipeline-judge is qwen3-coder:480b with avg score 52.9. Evaluator preference: evolution-skeptic > rubric_v2 > rubric_v1 (ignored HTTP errors)." }, "planner": { - "model": "deepseek-v4-pro", - "fit": 41.7, - "explanation": "Best model for planner is deepseek-v4-pro with avg score 41.7. Strongest dimension: code_presence." + "model": "qwen3-coder:480b", + "fit": 33.7, + "explanation": "Best model for planner is qwen3-coder:480b with avg score 33.7. Evaluator preference: evolution-skeptic > rubric_v2 > rubric_v1 (ignored HTTP errors)." }, "product-owner": { - "model": "kimi-k2.6", - "fit": 34.6, - "explanation": "Best model for product-owner is kimi-k2.6 with avg score 34.6. Strongest dimension: actionability." + "model": "qwen3-coder:480b", + "fit": 27.0, + "explanation": "Best model for product-owner is qwen3-coder:480b with avg score 27.0. Evaluator preference: evolution-skeptic > rubric_v2 > rubric_v1 (ignored HTTP errors)." }, "prompt-optimizer": { - "model": "glm-5.1", - "fit": 48.3, - "explanation": "Best model for prompt-optimizer is glm-5.1 with avg score 48.3. Strongest dimension: code_presence." + "model": "qwen3-coder:480b", + "fit": 31.8, + "explanation": "Best model for prompt-optimizer is qwen3-coder:480b with avg score 31.8. Evaluator preference: evolution-skeptic > rubric_v2 > rubric_v1 (ignored HTTP errors)." }, "python-developer": { - "model": "deepseek-v4-pro", + "model": "qwen3-coder:480b", "fit": 48.3, - "explanation": "Best model for python-developer is deepseek-v4-pro with avg score 48.3. Strongest dimension: code_presence." + "explanation": "Best model for python-developer is qwen3-coder:480b with avg score 48.3. Evaluator preference: evolution-skeptic > rubric_v2 > rubric_v1 (ignored HTTP errors)." }, "reflector": { - "model": "kimi-k2.6", - "fit": 58.7, - "explanation": "Best model for reflector is kimi-k2.6 with avg score 58.7. Strongest dimension: code_presence." + "model": "qwen3-coder:480b", + "fit": 20.9, + "explanation": "Best model for reflector is qwen3-coder:480b with avg score 20.9. Evaluator preference: evolution-skeptic > rubric_v2 > rubric_v1 (ignored HTTP errors)." }, "release-manager": { - "model": "kimi-k2.6", - "fit": 50.2, - "explanation": "Best model for release-manager is kimi-k2.6 with avg score 50.2. Strongest dimension: code_presence." + "model": "qwen3-coder:480b", + "fit": 41.7, + "explanation": "Best model for release-manager is qwen3-coder:480b with avg score 41.7. Evaluator preference: evolution-skeptic > rubric_v2 > rubric_v1 (ignored HTTP errors)." }, "requirement-refiner": { "model": "qwen3-coder:480b", "fit": 45.3, - "explanation": "Best model for requirement-refiner is qwen3-coder:480b with avg score 45.3. Strongest dimension: code_presence." + "explanation": "Best model for requirement-refiner is qwen3-coder:480b with avg score 45.3. Evaluator preference: evolution-skeptic > rubric_v2 > rubric_v1 (ignored HTTP errors)." }, "sdet-engineer": { - "model": "kimi-k2.6", + "model": "qwen3-coder:480b", "fit": 97.0, - "explanation": "Best model for sdet-engineer is kimi-k2.6 with avg score 97.0. Strongest dimension: keyword_coverage." + "explanation": "Best model for sdet-engineer is qwen3-coder:480b with avg score 97.0. Evaluator preference: evolution-skeptic > rubric_v2 > rubric_v1 (ignored HTTP errors)." }, "security-auditor": { - "model": "kimi-k2.6", - "fit": 63.8, - "explanation": "Best model for security-auditor is kimi-k2.6 with avg score 63.8. Strongest dimension: code_presence." + "model": "qwen3-coder:480b", + "fit": 41.5, + "explanation": "Best model for security-auditor is qwen3-coder:480b with avg score 41.5. Evaluator preference: evolution-skeptic > rubric_v2 > rubric_v1 (ignored HTTP errors)." }, "system-analyst": { - "model": "kimi-k2.6", - "fit": 92.0, - "explanation": "Best model for system-analyst is kimi-k2.6 with avg score 92.0. Strongest dimension: keyword_coverage." + "model": "qwen3-coder:480b", + "fit": 77.0, + "explanation": "Best model for system-analyst is qwen3-coder:480b with avg score 77.0. Evaluator preference: evolution-skeptic > rubric_v2 > rubric_v1 (ignored HTTP errors)." }, "the-fixer": { - "model": "glm-5.1", - "fit": 46.6, - "explanation": "Best model for the-fixer is glm-5.1 with avg score 46.6. Strongest dimension: code_presence." + "model": "qwen3-coder:480b", + "fit": 42.9, + "explanation": "Best model for the-fixer is qwen3-coder:480b with avg score 42.9. Evaluator preference: evolution-skeptic > rubric_v2 > rubric_v1 (ignored HTTP errors)." }, "visual-tester": { - "model": "glm-5.1", - "fit": 58.7, - "explanation": "Best model for visual-tester is glm-5.1 with avg score 58.7. Strongest dimension: code_presence." + "model": "qwen3-coder:480b", + "fit": 53.5, + "explanation": "Best model for visual-tester is qwen3-coder:480b with avg score 53.5. Evaluator preference: evolution-skeptic > rubric_v2 > rubric_v1 (ignored HTTP errors)." }, "workflow-architect": { - "model": "glm-5.1", - "fit": 48.3, - "explanation": "Best model for workflow-architect is glm-5.1 with avg score 48.3. Strongest dimension: code_presence." + "model": "qwen3-coder:480b", + "fit": 36.3, + "explanation": "Best model for workflow-architect is qwen3-coder:480b with avg score 36.3. Evaluator preference: evolution-skeptic > rubric_v2 > rubric_v1 (ignored HTTP errors)." }, "workflow-cross-checker": { "model": "qwen3-coder:480b", "fit": 65.6, - "explanation": "Best model for workflow-cross-checker is qwen3-coder:480b with avg score 65.6. Strongest dimension: code_presence." + "explanation": "Best model for workflow-cross-checker is qwen3-coder:480b with avg score 65.6. Evaluator preference: evolution-skeptic > rubric_v2 > rubric_v1 (ignored HTTP errors)." } } } \ No newline at end of file