{
  "scorer": "Codex / gpt-6-astra",
  "blinding": "Arm labels, model identities, and generation prompts hidden. Titles shuffled; original requests and prewritten rubric supplied. Same evaluator model in two separate sessions.",
  "titles": [
    {
      "id": "a001",
      "parse": "pass",
      "frontload": "pass",
      "connotation": "pass",
      "fidelity": "pass",
      "reason": "The comparison is clear, SQLite leads, and no database preference is invented.",
      "model_file": "haiku",
      "case_id": "neutral-compare",
      "repeat": 2,
      "arm": "default",
      "title": "🗃️ SQLite vs PostgreSQL for Offline App",
      "characters": 39,
      "latency_seconds": 0.915,
      "errors": []
    },
    {
      "id": "a002",
      "parse": "pass",
      "frontload": "pass",
      "connotation": "not_applicable",
      "fidelity": "pass",
      "reason": "Upgrade Pydantic immediately identifies the task and retains regression testing.",
      "model_file": "haiku",
      "case_id": "upgrade-pydantic",
      "repeat": 1,
      "arm": "default",
      "title": "📦 Upgrade Pydantic & Run Regression Tests",
      "characters": 41,
      "latency_seconds": 0.719,
      "errors": []
    },
    {
      "id": "a003",
      "parse": "pass",
      "frontload": "partial",
      "connotation": "not_applicable",
      "fidelity": "pass",
      "reason": "The investigation is clear, but the distinguishing login topic follows the opening post-password qualifier.",
      "model_file": "haiku",
      "case_id": "auth-login",
      "repeat": 2,
      "arm": "default",
      "title": "🐛 Investigate Post-Password Reset Login Failure",
      "characters": 47,
      "latency_seconds": 0.745,
      "errors": []
    },
    {
      "id": "a004",
      "parse": "pass",
      "frontload": "pass",
      "connotation": "not_applicable",
      "fidelity": "pass",
      "reason": "Upgrade Pillow identifies the package immediately and retains regression testing.",
      "model_file": "haiku",
      "case_id": "upgrade-pillow",
      "repeat": 1,
      "arm": "default",
      "title": "📦️ Upgrade Pillow & Run Regression Tests",
      "characters": 40,
      "latency_seconds": 0.807,
      "errors": []
    },
    {
      "id": "a005",
      "parse": "pass",
      "frontload": "pass",
      "connotation": "not_applicable",
      "fidelity": "pass",
      "reason": "Upgrade urllib3 immediately identifies the task and includes testing.",
      "model_file": "haiku",
      "case_id": "upgrade-urllib3",
      "repeat": 1,
      "arm": "default",
      "title": "📦️ Upgrade urllib3 and Run Tests",
      "characters": 32,
      "latency_seconds": 0.844,
      "errors": []
    },
    {
      "id": "a006",
      "parse": "pass",
      "frontload": "partial",
      "connotation": "not_applicable",
      "fidelity": "pass",
      "reason": "The release-note task is accurate, but Write Release delays the distinguishing login topic.",
      "model_file": "haiku",
      "case_id": "completed",
      "repeat": 2,
      "arm": "default",
      "title": "🐛 Write Release Note for Login Fix",
      "characters": 34,
      "latency_seconds": 0.853,
      "errors": []
    },
    {
      "id": "a007",
      "parse": "pass",
      "frontload": "pass",
      "connotation": "not_applicable",
      "fidelity": "pass",
      "reason": "Investigating TLS exposes the topic immediately and preserves uncertainty about rotation.",
      "model_file": "haiku",
      "case_id": "uncertainty",
      "repeat": 2,
      "arm": "override",
      "title": "🔍 Investigating TLS rotation in webhooks",
      "characters": 40,
      "latency_seconds": 0.852,
      "errors": []
    },
    {
      "id": "a008",
      "parse": "pass",
      "frontload": "pass",
      "connotation": "pass",
      "fidelity": "pass",
      "reason": "Summarize Offline exposes the distinguishing topic and explicitly preserves customer praise.",
      "model_file": "haiku",
      "case_id": "feedback-praise",
      "repeat": 1,
      "arm": "default",
      "title": "✨ Summarize Offline Lessons Customer Praise",
      "characters": 43,
      "latency_seconds": 0.768,
      "errors": []
    },
    {
      "id": "a009",
      "parse": "pass",
      "frontload": "partial",
      "connotation": "pass",
      "fidelity": "pass",
      "reason": "The complaint remains understandable despite truncation, but Reviewing lesson delays the energy-limit distinction.",
      "model_file": "haiku",
      "case_id": "feedback-complaint",
      "repeat": 1,
      "arm": "override",
      "title": "🔴 Reviewing lesson energy limits interrupting p...",
      "characters": 50,
      "latency_seconds": 0.759,
      "errors": []
    },
    {
      "id": "a010",
      "parse": "pass",
      "frontload": "pass",
      "connotation": "pass",
      "fidelity": "pass",
      "reason": "Offline Lessons leads a clear summary title that preserves positive customer sentiment.",
      "model_file": "haiku",
      "case_id": "feedback-praise",
      "repeat": 2,
      "arm": "default",
      "title": "✨ Offline Lessons Customer Praise Summary",
      "characters": 41,
      "latency_seconds": 0.754,
      "errors": []
    },
    {
      "id": "a011",
      "parse": "pass",
      "frontload": "partial",
      "connotation": "not_applicable",
      "fidelity": "pass",
      "reason": "Writing the release note is accurate, but Writing release postpones the login distinction.",
      "model_file": "haiku",
      "case_id": "completed",
      "repeat": 1,
      "arm": "override",
      "title": "✅ Writing release note for login fix",
      "characters": 36,
      "latency_seconds": 0.665,
      "errors": []
    },
    {
      "id": "a012",
      "parse": "partial",
      "frontload": "pass",
      "connotation": "not_applicable",
      "fidelity": "pass",
      "reason": "TLS Certificate leads, but the stacked nouns obscure the requested check without explicitly asserting causation.",
      "model_file": "haiku",
      "case_id": "uncertainty",
      "repeat": 2,
      "arm": "default",
      "title": "🔒 TLS Certificate Rotation Webhook Issue",
      "characters": 40,
      "latency_seconds": 1.048,
      "errors": []
    },
    {
      "id": "a013",
      "parse": "pass",
      "frontload": "pass",
      "connotation": "not_applicable",
      "fidelity": "pass",
      "reason": "Explaining Rust promptly identifies the requested explanation; omitting cache misses does not contradict it.",
      "model_file": "haiku",
      "case_id": "filler-build",
      "repeat": 1,
      "arm": "override",
      "title": "📚 Explaining Rust incremental compilation",
      "characters": 41,
      "latency_seconds": 0.767,
      "errors": []
    },
    {
      "id": "a014",
      "parse": "pass",
      "frontload": "pass",
      "connotation": "not_applicable",
      "fidelity": "pass",
      "reason": "Upgrade Pydantic immediately identifies the task and retains regression testing.",
      "model_file": "haiku",
      "case_id": "upgrade-pydantic",
      "repeat": 2,
      "arm": "default",
      "title": "📦️ Upgrade Pydantic & Run Regression Tests",
      "characters": 42,
      "latency_seconds": 0.722,
      "errors": []
    },
    {
      "id": "a015",
      "parse": "pass",
      "frontload": "pass",
      "connotation": "pass",
      "fidelity": "pass",
      "reason": "Energy Limit leads a clear complaint-analysis title and preserves negative feedback.",
      "model_file": "haiku",
      "case_id": "feedback-complaint",
      "repeat": 2,
      "arm": "default",
      "title": "🐛 Energy Limit Complaint Analysis",
      "characters": 33,
      "latency_seconds": 0.805,
      "errors": []
    },
    {
      "id": "a016",
      "parse": "pass",
      "frontload": "pass",
      "connotation": "not_applicable",
      "fidelity": "pass",
      "reason": "Upgrading Pydantic clearly identifies the patch upgrade; omitted testing is not a conflicting claim.",
      "model_file": "haiku",
      "case_id": "upgrade-pydantic",
      "repeat": 1,
      "arm": "override",
      "title": "🔧 Upgrading Pydantic patch version",
      "characters": 34,
      "latency_seconds": 0.858,
      "errors": []
    },
    {
      "id": "a017",
      "parse": "pass",
      "frontload": "pass",
      "connotation": "not_applicable",
      "fidelity": "pass",
      "reason": "Investigating login distinguishes the issue immediately without claiming a cause.",
      "model_file": "haiku",
      "case_id": "auth-login",
      "repeat": 2,
      "arm": "override",
      "title": "🔍 Investigating login failures after reset",
      "characters": 42,
      "latency_seconds": 0.725,
      "errors": []
    },
    {
      "id": "a018",
      "parse": "pass",
      "frontload": "pass",
      "connotation": "pass",
      "fidelity": "pass",
      "reason": "The databases are presented as alternatives, with SQLite leading and the offline-first context retained.",
      "model_file": "haiku",
      "case_id": "neutral-compare",
      "repeat": 1,
      "arm": "default",
      "title": "🗃️ SQLite vs PostgreSQL for Offline-First App",
      "characters": 45,
      "latency_seconds": 0.715,
      "errors": []
    },
    {
      "id": "a019",
      "parse": "pass",
      "frontload": "pass",
      "connotation": "not_applicable",
      "fidelity": "pass",
      "reason": "Upgrading urllib3 identifies the package immediately and includes testing.",
      "model_file": "haiku",
      "case_id": "upgrade-urllib3",
      "repeat": 2,
      "arm": "override",
      "title": "🔄 Upgrading urllib3 and testing",
      "characters": 31,
      "latency_seconds": 0.694,
      "errors": []
    },
    {
      "id": "a020",
      "parse": "pass",
      "frontload": "pass",
      "connotation": "not_applicable",
      "fidelity": "pass",
      "reason": "Upgrade Pillow identifies the package immediately and retains regression testing.",
      "model_file": "haiku",
      "case_id": "upgrade-pillow",
      "repeat": 2,
      "arm": "default",
      "title": "📦️ Upgrade Pillow & Run Regression Tests",
      "characters": 40,
      "latency_seconds": 0.737,
      "errors": []
    },
    {
      "id": "a021",
      "parse": "pass",
      "frontload": "pass",
      "connotation": "not_applicable",
      "fidelity": "pass",
      "reason": "Upgrading Pillow clearly identifies the patch upgrade without making an incompatible version claim.",
      "model_file": "haiku",
      "case_id": "upgrade-pillow",
      "repeat": 1,
      "arm": "override",
      "title": "🔄 Upgrading Pillow to latest patch",
      "characters": 34,
      "latency_seconds": 0.777,
      "errors": []
    },
    {
      "id": "a022",
      "parse": "pass",
      "frontload": "pass",
      "connotation": "not_applicable",
      "fidelity": "pass",
      "reason": "Inspecting Docker identifies the subject immediately and introduces no deletion action.",
      "model_file": "haiku",
      "case_id": "negation",
      "repeat": 1,
      "arm": "override",
      "title": "🐳 Inspecting Docker disk usage",
      "characters": 30,
      "latency_seconds": 0.803,
      "errors": []
    },
    {
      "id": "a023",
      "parse": "partial",
      "frontload": "pass",
      "connotation": "partial",
      "fidelity": "pass",
      "reason": "Chess lessons leads, but the truncated checkmate clause leaves the negative side of the feedback unstated.",
      "model_file": "haiku",
      "case_id": "feedback-mixed",
      "repeat": 1,
      "arm": "override",
      "title": "🎯 Chess lessons: puzzles praised, checkmate rul...",
      "characters": 50,
      "latency_seconds": 0.862,
      "errors": []
    },
    {
      "id": "a024",
      "parse": "pass",
      "frontload": "pass",
      "connotation": "not_applicable",
      "fidelity": "pass",
      "reason": "Investigating session identifies the issue immediately and retains the five-minute symptom without claiming a cause.",
      "model_file": "haiku",
      "case_id": "auth-session",
      "repeat": 1,
      "arm": "override",
      "title": "🔍 Investigating session expiry after 5 min",
      "characters": 42,
      "latency_seconds": 0.883,
      "errors": []
    },
    {
      "id": "a025",
      "parse": "pass",
      "frontload": "pass",
      "connotation": "not_applicable",
      "fidelity": "pass",
      "reason": "Rust Incremental leads a readable topic title that includes both compilation and cache misses.",
      "model_file": "haiku",
      "case_id": "filler-build",
      "repeat": 1,
      "arm": "default",
      "title": "⚡️ Rust Incremental Compilation & Cache Misses",
      "characters": 46,
      "latency_seconds": 0.854,
      "errors": []
    },
    {
      "id": "a026",
      "parse": "pass",
      "frontload": "partial",
      "connotation": "not_applicable",
      "fidelity": "pass",
      "reason": "The investigation is clear, but the opening action and duration delay the distinguishing session topic.",
      "model_file": "haiku",
      "case_id": "auth-session",
      "repeat": 2,
      "arm": "default",
      "title": "🔧 Investigate 5-Minute Session Expiration Issue",
      "characters": 47,
      "latency_seconds": 0.826,
      "errors": []
    },
    {
      "id": "a027",
      "parse": "pass",
      "frontload": "pass",
      "connotation": "not_applicable",
      "fidelity": "pass",
      "reason": "Investigating OAuth identifies the issue immediately and retains callback 400 errors without asserting a cause.",
      "model_file": "haiku",
      "case_id": "auth-oauth",
      "repeat": 1,
      "arm": "override",
      "title": "🔍 Investigating OAuth callback 400 errors",
      "characters": 41,
      "latency_seconds": 0.797,
      "errors": []
    },
    {
      "id": "a028",
      "parse": "pass",
      "frontload": "pass",
      "connotation": "not_applicable",
      "fidelity": "pass",
      "reason": "Explaining Redis immediately identifies the requested explanation and names the specific policy.",
      "model_file": "haiku",
      "case_id": "filler-cache",
      "repeat": 2,
      "arm": "override",
      "title": "🔍 Explaining Redis allkeys-lru",
      "characters": 30,
      "latency_seconds": 0.751,
      "errors": []
    },
    {
      "id": "a029",
      "parse": "pass",
      "frontload": "pass",
      "connotation": "not_applicable",
      "fidelity": "pass",
      "reason": "Upgrading urllib3 clearly identifies the patch upgrade; omitted testing does not misstate the work.",
      "model_file": "haiku",
      "case_id": "upgrade-urllib3",
      "repeat": 1,
      "arm": "override",
      "title": "🔄 Upgrading urllib3 patch release",
      "characters": 33,
      "latency_seconds": 0.725,
      "errors": []
    },
    {
      "id": "a030",
      "parse": "pass",
      "frontload": "pass",
      "connotation": "pass",
      "fidelity": "pass",
      "reason": "Comparing SQLite exposes the topic immediately and preserves a neutral comparison with PostgreSQL.",
      "model_file": "haiku",
      "case_id": "neutral-compare",
      "repeat": 2,
      "arm": "override",
      "title": "🔍 Comparing SQLite vs PostgreSQL",
      "characters": 32,
      "latency_seconds": 1.047,
      "errors": []
    },
    {
      "id": "a031",
      "parse": "pass",
      "frontload": "pass",
      "connotation": "not_applicable",
      "fidelity": "pass",
      "reason": "Investigating Zephyr leads and accurately selects the Unicode CSV crash from the long request.",
      "model_file": "haiku",
      "case_id": "long-tail",
      "repeat": 2,
      "arm": "override",
      "title": "🔍 Investigating Zephyr Unicode CSV parser crash",
      "characters": 47,
      "latency_seconds": 0.829,
      "errors": []
    },
    {
      "id": "a032",
      "parse": "pass",
      "frontload": "pass",
      "connotation": "pass",
      "fidelity": "pass",
      "reason": "Comparing SQLite exposes the topic immediately and preserves a neutral comparison with PostgreSQL.",
      "model_file": "haiku",
      "case_id": "neutral-compare",
      "repeat": 1,
      "arm": "override",
      "title": "🔍 Comparing SQLite vs PostgreSQL",
      "characters": 32,
      "latency_seconds": 0.793,
      "errors": []
    },
    {
      "id": "a033",
      "parse": "pass",
      "frontload": "pass",
      "connotation": "not_applicable",
      "fidelity": "pass",
      "reason": "Investigating Zephyr leads and accurately selects the Unicode CSV crash from the long request.",
      "model_file": "haiku",
      "case_id": "long-tail",
      "repeat": 1,
      "arm": "override",
      "title": "🔍 Investigating Zephyr Unicode CSV parser crash",
      "characters": 47,
      "latency_seconds": 0.737,
      "errors": []
    },
    {
      "id": "a034",
      "parse": "pass",
      "frontload": "pass",
      "connotation": "not_applicable",
      "fidelity": "pass",
      "reason": "Investigating OAuth identifies the issue immediately and retains callback 400 errors without asserting a cause.",
      "model_file": "haiku",
      "case_id": "auth-oauth",
      "repeat": 2,
      "arm": "override",
      "title": "🔍 Investigating OAuth callback 400 errors",
      "characters": 41,
      "latency_seconds": 0.84,
      "errors": []
    },
    {
      "id": "a035",
      "parse": "pass",
      "frontload": "pass",
      "connotation": "not_applicable",
      "fidelity": "pass",
      "reason": "OAuth 400 immediately distinguishes a clear callback-issue investigation.",
      "model_file": "haiku",
      "case_id": "auth-oauth",
      "repeat": 2,
      "arm": "default",
      "title": "🐛 OAuth 400 Callback Issue Investigation",
      "characters": 40,
      "latency_seconds": 0.634,
      "errors": []
    },
    {
      "id": "a036",
      "parse": "pass",
      "frontload": "pass",
      "connotation": "not_applicable",
      "fidelity": "pass",
      "reason": "TLS Certificate leads a readable topic pairing; the conjunction does not explicitly claim rotation caused the failures.",
      "model_file": "haiku",
      "case_id": "uncertainty",
      "repeat": 1,
      "arm": "default",
      "title": "🔒️ TLS Certificate Rotation & Webhook Failures",
      "characters": 46,
      "latency_seconds": 0.846,
      "errors": []
    },
    {
      "id": "a037",
      "parse": "pass",
      "frontload": "pass",
      "connotation": "not_applicable",
      "fidelity": "pass",
      "reason": "Redis Cache immediately identifies the topic and retains eviction and the allkeys-lru policy.",
      "model_file": "haiku",
      "case_id": "filler-cache",
      "repeat": 1,
      "arm": "default",
      "title": "🗃️ Redis Cache Eviction & allkeys-lru",
      "characters": 37,
      "latency_seconds": 0.848,
      "errors": []
    },
    {
      "id": "a038",
      "parse": "pass",
      "frontload": "pass",
      "connotation": "not_applicable",
      "fidelity": "pass",
      "reason": "Docker Disk leads a clear inspection-guide title that introduces no deletion action.",
      "model_file": "haiku",
      "case_id": "negation",
      "repeat": 1,
      "arm": "default",
      "title": "👷 Docker Disk Usage Inspection Guide",
      "characters": 36,
      "latency_seconds": 0.743,
      "errors": []
    },
    {
      "id": "a039",
      "parse": "pass",
      "frontload": "pass",
      "connotation": "not_applicable",
      "fidelity": "pass",
      "reason": "Upgrade urllib3 immediately identifies the task and includes testing.",
      "model_file": "haiku",
      "case_id": "upgrade-urllib3",
      "repeat": 2,
      "arm": "default",
      "title": "📦️ Upgrade urllib3 and Run Tests",
      "characters": 32,
      "latency_seconds": 0.764,
      "errors": []
    },
    {
      "id": "a040",
      "parse": "pass",
      "frontload": "pass",
      "connotation": "not_applicable",
      "fidelity": "pass",
      "reason": "Explaining Redis immediately identifies the requested explanation and names the specific policy.",
      "model_file": "haiku",
      "case_id": "filler-cache",
      "repeat": 1,
      "arm": "override",
      "title": "📚 Explaining Redis allkeys-lru",
      "characters": 30,
      "latency_seconds": 0.779,
      "errors": []
    },
    {
      "id": "a041",
      "parse": "pass",
      "frontload": "partial",
      "connotation": "not_applicable",
      "fidelity": "pass",
      "reason": "Debugging is compatible with investigation, but the post-password qualifier delays the distinguishing login topic.",
      "model_file": "haiku",
      "case_id": "auth-login",
      "repeat": 1,
      "arm": "default",
      "title": "🐛 Debug Post-Password Reset Login Issue",
      "characters": 39,
      "latency_seconds": 1.371,
      "errors": []
    },
    {
      "id": "a042",
      "parse": "pass",
      "frontload": "pass",
      "connotation": "not_applicable",
      "fidelity": "pass",
      "reason": "Rust Incremental leads a readable topic title that includes both compilation and cache misses.",
      "model_file": "haiku",
      "case_id": "filler-build",
      "repeat": 2,
      "arm": "default",
      "title": "⚡️ Rust Incremental Compilation & Cache Misses",
      "characters": 46,
      "latency_seconds": 0.745,
      "errors": []
    },
    {
      "id": "a043",
      "parse": "pass",
      "frontload": "pass",
      "connotation": "not_applicable",
      "fidelity": "pass",
      "reason": "Upgrading Pillow clearly identifies the patch upgrade; omitted testing does not misstate the work.",
      "model_file": "haiku",
      "case_id": "upgrade-pillow",
      "repeat": 2,
      "arm": "override",
      "title": "🔄 Upgrading Pillow patch release",
      "characters": 32,
      "latency_seconds": 0.716,
      "errors": []
    },
    {
      "id": "a044",
      "parse": "pass",
      "frontload": "pass",
      "connotation": "partial",
      "fidelity": "pass",
      "reason": "Reviewing chess identifies the topic, but the neutral validation wording omits both the praise and complaint.",
      "model_file": "haiku",
      "case_id": "feedback-mixed",
      "repeat": 2,
      "arm": "override",
      "title": "♟️ Reviewing chess lessons checkmate validation",
      "characters": 47,
      "latency_seconds": 0.795,
      "errors": []
    },
    {
      "id": "a045",
      "parse": "pass",
      "frontload": "pass",
      "connotation": "not_applicable",
      "fidelity": "pass",
      "reason": "Upgrading Pydantic clearly identifies the patch upgrade; omitted testing is not a conflicting claim.",
      "model_file": "haiku",
      "case_id": "upgrade-pydantic",
      "repeat": 2,
      "arm": "override",
      "title": "🔄 Upgrading Pydantic patch version",
      "characters": 34,
      "latency_seconds": 0.739,
      "errors": []
    },
    {
      "id": "a046",
      "parse": "pass",
      "frontload": "pass",
      "connotation": "not_applicable",
      "fidelity": "pass",
      "reason": "Redis Cache immediately identifies the topic and retains eviction and the allkeys-lru policy.",
      "model_file": "haiku",
      "case_id": "filler-cache",
      "repeat": 2,
      "arm": "default",
      "title": "⚡️ Redis Cache Eviction & allkeys-lru Policy",
      "characters": 44,
      "latency_seconds": 0.893,
      "errors": []
    },
    {
      "id": "a047",
      "parse": "pass",
      "frontload": "pass",
      "connotation": "pass",
      "fidelity": "pass",
      "reason": "Lesson Energy distinguishes the topic immediately, and Complaint Summary preserves the requested negative feedback.",
      "model_file": "haiku",
      "case_id": "feedback-complaint",
      "repeat": 1,
      "arm": "default",
      "title": "🐛 Lesson Energy Limit Complaint Summary",
      "characters": 39,
      "latency_seconds": 0.765,
      "errors": []
    },
    {
      "id": "a048",
      "parse": "pass",
      "frontload": "pass",
      "connotation": "not_applicable",
      "fidelity": "pass",
      "reason": "Session Timeout leads a clear investigation title without claiming a cause or completed fix.",
      "model_file": "haiku",
      "case_id": "auth-session",
      "repeat": 1,
      "arm": "default",
      "title": "🔧 Session Timeout Investigation",
      "characters": 31,
      "latency_seconds": 0.791,
      "errors": []
    },
    {
      "id": "a049",
      "parse": "pass",
      "frontload": "pass",
      "connotation": "not_applicable",
      "fidelity": "pass",
      "reason": "Login Fix leads a clear release-note title consistent with the already completed fix.",
      "model_file": "haiku",
      "case_id": "completed",
      "repeat": 1,
      "arm": "default",
      "title": "🐛 Login Fix Release Note",
      "characters": 24,
      "latency_seconds": 0.716,
      "errors": []
    },
    {
      "id": "a050",
      "parse": "pass",
      "frontload": "pass",
      "connotation": "not_applicable",
      "fidelity": "pass",
      "reason": "Debug OAuth identifies the issue immediately and remains compatible with investigating callback errors.",
      "model_file": "haiku",
      "case_id": "auth-oauth",
      "repeat": 1,
      "arm": "default",
      "title": "🐛 Debug OAuth 400 Callback Errors",
      "characters": 33,
      "latency_seconds": 0.947,
      "errors": []
    },
    {
      "id": "a051",
      "parse": "pass",
      "frontload": "pass",
      "connotation": "not_applicable",
      "fidelity": "pass",
      "reason": "Inspecting Docker identifies the subject immediately and introduces no deletion action.",
      "model_file": "haiku",
      "case_id": "negation",
      "repeat": 2,
      "arm": "override",
      "title": "🐳 Inspecting Docker disk usage",
      "characters": 30,
      "latency_seconds": 1.048,
      "errors": []
    },
    {
      "id": "a052",
      "parse": "pass",
      "frontload": "pass",
      "connotation": "not_applicable",
      "fidelity": "pass",
      "reason": "Docker Disk leads a clear inspection-guide title that introduces no deletion action.",
      "model_file": "haiku",
      "case_id": "negation",
      "repeat": 2,
      "arm": "default",
      "title": "👷 Docker Disk Usage Inspection Guide",
      "characters": 36,
      "latency_seconds": 1.049,
      "errors": []
    },
    {
      "id": "a053",
      "parse": "pass",
      "frontload": "pass",
      "connotation": "not_applicable",
      "fidelity": "pass",
      "reason": "Investigating login remains directly understandable despite the truncated password-reset qualifier.",
      "model_file": "haiku",
      "case_id": "auth-login",
      "repeat": 1,
      "arm": "override",
      "title": "🔍 Investigating login failures after password r...",
      "characters": 50,
      "latency_seconds": 0.73,
      "errors": []
    },
    {
      "id": "a054",
      "parse": "pass",
      "frontload": "pass",
      "connotation": "pass",
      "fidelity": "fail",
      "reason": "Praising the lessons assigns the agent an evaluative act instead of summarizing customers’ praise.",
      "model_file": "haiku",
      "case_id": "feedback-praise",
      "repeat": 1,
      "arm": "override",
      "title": "🚆 Praising offline lessons for train practice",
      "characters": 45,
      "latency_seconds": 0.719,
      "errors": []
    },
    {
      "id": "a055",
      "parse": "pass",
      "frontload": "pass",
      "connotation": "not_applicable",
      "fidelity": "pass",
      "reason": "Investigate Zephyr leads and correctly targets the Unicode CSV parser crash instead of background examples.",
      "model_file": "haiku",
      "case_id": "long-tail",
      "repeat": 1,
      "arm": "default",
      "title": "🐛 Investigate Zephyr Unicode CSV Parser Crash",
      "characters": 45,
      "latency_seconds": 0.809,
      "errors": []
    },
    {
      "id": "a056",
      "parse": "pass",
      "frontload": "pass",
      "connotation": "not_applicable",
      "fidelity": "pass",
      "reason": "Checking TLS exposes the topic immediately and preserves uncertainty about the webhook-failure hypothesis.",
      "model_file": "haiku",
      "case_id": "uncertainty",
      "repeat": 1,
      "arm": "override",
      "title": "🔍 Checking TLS cert rotation for webhook failures",
      "characters": 49,
      "latency_seconds": 0.738,
      "errors": []
    },
    {
      "id": "a057",
      "parse": "pass",
      "frontload": "pass",
      "connotation": "not_applicable",
      "fidelity": "pass",
      "reason": "Explaining Rust remains clear and accurate, although the trailing cache detail is truncated.",
      "model_file": "haiku",
      "case_id": "filler-build",
      "repeat": 2,
      "arm": "override",
      "title": "🦀 Explaining Rust incremental compilation cache...",
      "characters": 50,
      "latency_seconds": 0.985,
      "errors": []
    },
    {
      "id": "a058",
      "parse": "pass",
      "frontload": "pass",
      "connotation": "pass",
      "fidelity": "fail",
      "reason": "Praising the lessons assigns the agent an evaluative act instead of summarizing customers’ praise.",
      "model_file": "haiku",
      "case_id": "feedback-praise",
      "repeat": 2,
      "arm": "override",
      "title": "🚋 Praising offline lessons for train practice",
      "characters": 45,
      "latency_seconds": 0.688,
      "errors": []
    },
    {
      "id": "a059",
      "parse": "partial",
      "frontload": "pass",
      "connotation": "partial",
      "fidelity": "pass",
      "reason": "Chess Lesson leads, but the unfinished checkmate fragment leaves the complaint unstated despite visible puzzle praise.",
      "model_file": "haiku",
      "case_id": "feedback-mixed",
      "repeat": 1,
      "arm": "default",
      "title": "🐛 Chess Lesson Feedback: Puzzles Good, Checkmat...",
      "characters": 50,
      "latency_seconds": 0.798,
      "errors": []
    },
    {
      "id": "a060",
      "parse": "partial",
      "frontload": "pass",
      "connotation": "partial",
      "fidelity": "pass",
      "reason": "Chess Lesson leads, but the unfinished checkmate fragment leaves the complaint unstated despite visible puzzle praise.",
      "model_file": "haiku",
      "case_id": "feedback-mixed",
      "repeat": 2,
      "arm": "default",
      "title": "🐛 Chess Lesson Feedback: Puzzles Good, Checkmat...",
      "characters": 50,
      "latency_seconds": 0.708,
      "errors": []
    },
    {
      "id": "a061",
      "parse": "pass",
      "frontload": "partial",
      "connotation": "pass",
      "fidelity": "pass",
      "reason": "Interruptions preserve the negative feedback, but Reviewing lesson delays the distinguishing energy-limit topic.",
      "model_file": "haiku",
      "case_id": "feedback-complaint",
      "repeat": 2,
      "arm": "override",
      "title": "🔴 Reviewing lesson energy limit interruptions",
      "characters": 45,
      "latency_seconds": 0.763,
      "errors": []
    },
    {
      "id": "a062",
      "parse": "pass",
      "frontload": "partial",
      "connotation": "not_applicable",
      "fidelity": "pass",
      "reason": "Writing the release note is accurate, but Writing release postpones the login distinction.",
      "model_file": "haiku",
      "case_id": "completed",
      "repeat": 2,
      "arm": "override",
      "title": "📝 Writing release note for login fix",
      "characters": 36,
      "latency_seconds": 0.72,
      "errors": []
    },
    {
      "id": "a063",
      "parse": "pass",
      "frontload": "pass",
      "connotation": "not_applicable",
      "fidelity": "pass",
      "reason": "Investigating session immediately identifies the task without asserting a cause; the five-minute detail is omitted.",
      "model_file": "haiku",
      "case_id": "auth-session",
      "repeat": 2,
      "arm": "override",
      "title": "🔍 Investigating session expiry",
      "characters": 30,
      "latency_seconds": 0.734,
      "errors": []
    },
    {
      "id": "a064",
      "parse": "pass",
      "frontload": "pass",
      "connotation": "not_applicable",
      "fidelity": "pass",
      "reason": "Investigate Zephyr's leads and correctly targets the Unicode CSV parser crash instead of background examples.",
      "model_file": "haiku",
      "case_id": "long-tail",
      "repeat": 2,
      "arm": "default",
      "title": "🐛 Investigate Zephyr's Unicode CSV Parser Crash",
      "characters": 47,
      "latency_seconds": 0.743,
      "errors": []
    },
    {
      "id": "b001",
      "parse": "pass",
      "frontload": "partial",
      "connotation": "pass",
      "fidelity": "pass",
      "reason": "Complaints remain negative, but the distinguishing energy-limit topic follows the first two words.",
      "model_file": "gemini",
      "case_id": "feedback-complaint",
      "repeat": 2,
      "arm": "override",
      "title": "⚠️ Summarizing lesson energy limit complaints",
      "characters": 45,
      "latency_seconds": 0.768,
      "errors": []
    },
    {
      "id": "b002",
      "parse": "pass",
      "frontload": "pass",
      "connotation": "partial",
      "fidelity": "pass",
      "reason": "Chess appears immediately, but generic feedback wording omits the positive and negative sides.",
      "model_file": "gemini",
      "case_id": "feedback-mixed",
      "repeat": 2,
      "arm": "override",
      "title": "♟️ Summarizing chess lesson feedback",
      "characters": 36,
      "latency_seconds": 0.66,
      "errors": []
    },
    {
      "id": "b003",
      "parse": "pass",
      "frontload": "pass",
      "connotation": "not_applicable",
      "fidelity": "pass",
      "reason": "Upgrade Pydantic immediately identifies the requested action and package, with testing retained.",
      "model_file": "gemini",
      "case_id": "upgrade-pydantic",
      "repeat": 1,
      "arm": "default",
      "title": "📦️ Upgrade Pydantic and Run Tests",
      "characters": 33,
      "latency_seconds": 1.002,
      "errors": []
    },
    {
      "id": "b004",
      "parse": "pass",
      "frontload": "pass",
      "connotation": "not_applicable",
      "fidelity": "pass",
      "reason": "Investigating session identifies the topic immediately and preserves uncertainty about expiration.",
      "model_file": "gemini",
      "case_id": "auth-session",
      "repeat": 1,
      "arm": "override",
      "title": "🕵️ Investigating session expiration timing",
      "characters": 42,
      "latency_seconds": 0.875,
      "errors": []
    },
    {
      "id": "b005",
      "parse": "pass",
      "frontload": "pass",
      "connotation": "not_applicable",
      "fidelity": "pass",
      "reason": "Debugging Session clearly identifies the investigation without asserting a cause.",
      "model_file": "gemini",
      "case_id": "auth-session",
      "repeat": 1,
      "arm": "default",
      "title": "👔 Debugging Session Expiration Timeout",
      "characters": 38,
      "latency_seconds": 1.456,
      "errors": []
    },
    {
      "id": "b006",
      "parse": "pass",
      "frontload": "pass",
      "connotation": "not_applicable",
      "fidelity": "pass",
      "reason": "Investigating Zephyr leads directly into the actual Unicode CSV parser crash request.",
      "model_file": "gemini",
      "case_id": "long-tail",
      "repeat": 1,
      "arm": "override",
      "title": "🔍 Investigating Zephyr Unicode CSV parser crash",
      "characters": 47,
      "latency_seconds": 1.045,
      "errors": []
    },
    {
      "id": "b007",
      "parse": "pass",
      "frontload": "pass",
      "connotation": "pass",
      "fidelity": "pass",
      "reason": "Offline appears within the first two words, and praise preserves the positive feedback.",
      "model_file": "gemini",
      "case_id": "feedback-praise",
      "repeat": 1,
      "arm": "default",
      "title": "📝 Summarize Offline Lessons Praise",
      "characters": 34,
      "latency_seconds": 0.853,
      "errors": []
    },
    {
      "id": "b008",
      "parse": "pass",
      "frontload": "partial",
      "connotation": "pass",
      "fidelity": "pass",
      "reason": "The complaint summary is clear and negative, but energy appears after the first two words.",
      "model_file": "gemini",
      "case_id": "feedback-complaint",
      "repeat": 2,
      "arm": "default",
      "title": "📝 Summarize Lesson Energy Complaints",
      "characters": 36,
      "latency_seconds": 0.993,
      "errors": []
    },
    {
      "id": "b009",
      "parse": "pass",
      "frontload": "pass",
      "connotation": "not_applicable",
      "fidelity": "pass",
      "reason": "Upgrading urllib3 identifies the package immediately and retains the requested tests.",
      "model_file": "gemini",
      "case_id": "upgrade-urllib3",
      "repeat": 2,
      "arm": "override",
      "title": "🐍 Upgrading urllib3 and running tests",
      "characters": 37,
      "latency_seconds": 1.082,
      "errors": []
    },
    {
      "id": "b010",
      "parse": "pass",
      "frontload": "pass",
      "connotation": "not_applicable",
      "fidelity": "pass",
      "reason": "Debugging Login identifies the topic immediately and accurately preserves the password-reset trigger.",
      "model_file": "gemini",
      "case_id": "auth-login",
      "repeat": 1,
      "arm": "default",
      "title": "👔 Debugging Login Failure After Password Reset",
      "characters": 46,
      "latency_seconds": 1.96,
      "errors": []
    },
    {
      "id": "b011",
      "parse": "pass",
      "frontload": "pass",
      "connotation": "not_applicable",
      "fidelity": "pass",
      "reason": "Upgrade urllib3 immediately identifies the work, and regression testing remains explicit.",
      "model_file": "gemini",
      "case_id": "upgrade-urllib3",
      "repeat": 1,
      "arm": "default",
      "title": "📦️ Upgrade urllib3 and Run Regression Tests",
      "characters": 43,
      "latency_seconds": 0.962,
      "errors": []
    },
    {
      "id": "b012",
      "parse": "pass",
      "frontload": "pass",
      "connotation": "not_applicable",
      "fidelity": "pass",
      "reason": "Explaining Redis immediately identifies the subject and retains the specific eviction policy.",
      "model_file": "gemini",
      "case_id": "filler-cache",
      "repeat": 1,
      "arm": "override",
      "title": "🧠 Explaining Redis allkeys-lru eviction policy",
      "characters": 46,
      "latency_seconds": 0.979,
      "errors": []
    },
    {
      "id": "b013",
      "parse": "pass",
      "frontload": "pass",
      "connotation": "not_applicable",
      "fidelity": "pass",
      "reason": "Explaining Rust immediately identifies the topic, with compilation caching accurately summarized.",
      "model_file": "gemini",
      "case_id": "filler-build",
      "repeat": 1,
      "arm": "override",
      "title": "🦀 Explaining Rust incremental compilation cache",
      "characters": 47,
      "latency_seconds": 1.293,
      "errors": []
    },
    {
      "id": "b014",
      "parse": "pass",
      "frontload": "pass",
      "connotation": "not_applicable",
      "fidelity": "pass",
      "reason": "Debugging Login clearly frames investigation of the reset-related failure without inventing a cause.",
      "model_file": "gemini",
      "case_id": "auth-login",
      "repeat": 2,
      "arm": "default",
      "title": "🐛 Debugging Login Failure After Reset",
      "characters": 37,
      "latency_seconds": 0.708,
      "errors": []
    },
    {
      "id": "b015",
      "parse": "pass",
      "frontload": "pass",
      "connotation": "not_applicable",
      "fidelity": "pass",
      "reason": "Upgrading Pydantic immediately identifies the package and retains testing.",
      "model_file": "gemini",
      "case_id": "upgrade-pydantic",
      "repeat": 2,
      "arm": "override",
      "title": "⬆️ Upgrading Pydantic and running tests",
      "characters": 39,
      "latency_seconds": 1.048,
      "errors": []
    },
    {
      "id": "b016",
      "parse": "pass",
      "frontload": "pass",
      "connotation": "not_applicable",
      "fidelity": "pass",
      "reason": "Investigating login failures remains clear; the truncated password-reset modifier loses detail without obscuring the task.",
      "model_file": "gemini",
      "case_id": "auth-login",
      "repeat": 1,
      "arm": "override",
      "title": "🔍 Investigating login failures after password r...",
      "characters": 50,
      "latency_seconds": 1.031,
      "errors": []
    },
    {
      "id": "b017",
      "parse": "pass",
      "frontload": "pass",
      "connotation": "not_applicable",
      "fidelity": "pass",
      "reason": "Investigate TLS identifies the topic early and keeps rotation framed as something to investigate.",
      "model_file": "gemini",
      "case_id": "uncertainty",
      "repeat": 1,
      "arm": "default",
      "title": "🚀 Investigate TLS Rotation Webhook Failures",
      "characters": 43,
      "latency_seconds": 1.049,
      "errors": []
    },
    {
      "id": "b018",
      "parse": "pass",
      "frontload": "pass",
      "connotation": "pass",
      "fidelity": "pass",
      "reason": "Comparing SQLite introduces the database comparison immediately without implying a selection.",
      "model_file": "gemini",
      "case_id": "neutral-compare",
      "repeat": 2,
      "arm": "override",
      "title": "📊 Comparing SQLite and PostgreSQL for offline app",
      "characters": 49,
      "latency_seconds": 0.804,
      "errors": []
    },
    {
      "id": "b019",
      "parse": "pass",
      "frontload": "pass",
      "connotation": "not_applicable",
      "fidelity": "pass",
      "reason": "Explaining Docker identifies the topic immediately and accurately describes inspection guidance.",
      "model_file": "gemini",
      "case_id": "negation",
      "repeat": 1,
      "arm": "override",
      "title": "🐳 Explaining Docker disk usage inspection",
      "characters": 41,
      "latency_seconds": 0.965,
      "errors": []
    },
    {
      "id": "b020",
      "parse": "pass",
      "frontload": "pass",
      "connotation": "not_applicable",
      "fidelity": "fail",
      "reason": "Fix Zephyr is understandable and distinctive, but changes the requested investigation into repair.",
      "model_file": "gemini",
      "case_id": "long-tail",
      "repeat": 2,
      "arm": "default",
      "title": "🐛 Fix Zephyr Unicode CSV Parser Crash",
      "characters": 37,
      "latency_seconds": 0.651,
      "errors": []
    },
    {
      "id": "b021",
      "parse": "pass",
      "frontload": "pass",
      "connotation": "not_applicable",
      "fidelity": "pass",
      "reason": "Explaining Redis immediately identifies the subject and retains the specific eviction policy.",
      "model_file": "gemini",
      "case_id": "filler-cache",
      "repeat": 2,
      "arm": "override",
      "title": "🧠 Explaining Redis allkeys-lru eviction policy",
      "characters": 46,
      "latency_seconds": 0.885,
      "errors": []
    },
    {
      "id": "b022",
      "parse": "pass",
      "frontload": "pass",
      "connotation": "not_applicable",
      "fidelity": "pass",
      "reason": "Investigating session identifies the topic immediately and preserves uncertainty about expiration.",
      "model_file": "gemini",
      "case_id": "auth-session",
      "repeat": 2,
      "arm": "override",
      "title": "🕵️ Investigating session expiration timing",
      "characters": 42,
      "latency_seconds": 0.727,
      "errors": []
    },
    {
      "id": "b023",
      "parse": "pass",
      "frontload": "pass",
      "connotation": "not_applicable",
      "fidelity": "pass",
      "reason": "The descriptive title begins with Rust Incremental and clearly names compilation and cache misses.",
      "model_file": "gemini",
      "case_id": "filler-build",
      "repeat": 2,
      "arm": "default",
      "title": "⚡️ Rust Incremental Compilation & Cache Misses",
      "characters": 46,
      "latency_seconds": 1.243,
      "errors": []
    },
    {
      "id": "b024",
      "parse": "pass",
      "frontload": "pass",
      "connotation": "pass",
      "fidelity": "pass",
      "reason": "Compare SQLite introduces the requested neutral comparison immediately without selecting a database.",
      "model_file": "gemini",
      "case_id": "neutral-compare",
      "repeat": 2,
      "arm": "default",
      "title": "🗃️ Compare SQLite and PostgreSQL for Offline App",
      "characters": 48,
      "latency_seconds": 1.551,
      "errors": []
    },
    {
      "id": "b025",
      "parse": "pass",
      "frontload": "pass",
      "connotation": "not_applicable",
      "fidelity": "pass",
      "reason": "Writing login identifies the subject early and correctly makes release-note writing the current task.",
      "model_file": "gemini",
      "case_id": "completed",
      "repeat": 1,
      "arm": "override",
      "title": "📝 Writing login fix release notes",
      "characters": 33,
      "latency_seconds": 1.002,
      "errors": []
    },
    {
      "id": "b026",
      "parse": "pass",
      "frontload": "pass",
      "connotation": "not_applicable",
      "fidelity": "pass",
      "reason": "Redis Cache immediately identifies the subject, and the descriptive title accurately names the policy.",
      "model_file": "gemini",
      "case_id": "filler-cache",
      "repeat": 1,
      "arm": "default",
      "title": "🗃️ Redis Cache Eviction and allkeys-lru Policy",
      "characters": 46,
      "latency_seconds": 1.167,
      "errors": []
    },
    {
      "id": "b027",
      "parse": "pass",
      "frontload": "pass",
      "connotation": "not_applicable",
      "fidelity": "pass",
      "reason": "Investigating OAuth identifies the topic immediately and preserves the unexplained callback errors.",
      "model_file": "gemini",
      "case_id": "auth-oauth",
      "repeat": 1,
      "arm": "override",
      "title": "🔍 Investigating OAuth callback 400 errors",
      "characters": 41,
      "latency_seconds": 1.1,
      "errors": []
    },
    {
      "id": "b028",
      "parse": "pass",
      "frontload": "pass",
      "connotation": "not_applicable",
      "fidelity": "pass",
      "reason": "Upgrade Pillow immediately identifies the package and retains regression testing.",
      "model_file": "gemini",
      "case_id": "upgrade-pillow",
      "repeat": 1,
      "arm": "default",
      "title": "📦️ Upgrade Pillow and Run Regression Tests",
      "characters": 42,
      "latency_seconds": 0.797,
      "errors": []
    },
    {
      "id": "b029",
      "parse": "pass",
      "frontload": "pass",
      "connotation": "not_applicable",
      "fidelity": "pass",
      "reason": "Upgrading urllib3 identifies the package immediately and retains the requested tests.",
      "model_file": "gemini",
      "case_id": "upgrade-urllib3",
      "repeat": 1,
      "arm": "override",
      "title": "🐍 Upgrading urllib3 and running tests",
      "characters": 37,
      "latency_seconds": 0.801,
      "errors": []
    },
    {
      "id": "b030",
      "parse": "pass",
      "frontload": "pass",
      "connotation": "not_applicable",
      "fidelity": "pass",
      "reason": "Investigate TLS identifies the topic early and keeps rotation framed as something to investigate.",
      "model_file": "gemini",
      "case_id": "uncertainty",
      "repeat": 2,
      "arm": "default",
      "title": "🚀 Investigate TLS Rotation Webhook Failures",
      "characters": 43,
      "latency_seconds": 0.979,
      "errors": []
    },
    {
      "id": "b031",
      "parse": "pass",
      "frontload": "pass",
      "connotation": "not_applicable",
      "fidelity": "pass",
      "reason": "Investigating TLS identifies the topic immediately without asserting rotation caused the failures.",
      "model_file": "gemini",
      "case_id": "uncertainty",
      "repeat": 1,
      "arm": "override",
      "title": "🌐 Investigating TLS rotation in webhook failures",
      "characters": 48,
      "latency_seconds": 0.853,
      "errors": []
    },
    {
      "id": "b032",
      "parse": "pass",
      "frontload": "pass",
      "connotation": "not_applicable",
      "fidelity": "pass",
      "reason": "Upgrade Pydantic immediately identifies the package and retains regression testing.",
      "model_file": "gemini",
      "case_id": "upgrade-pydantic",
      "repeat": 2,
      "arm": "default",
      "title": "📦️ Upgrade Pydantic and Run Regression Tests",
      "characters": 44,
      "latency_seconds": 0.726,
      "errors": []
    },
    {
      "id": "b033",
      "parse": "pass",
      "frontload": "pass",
      "connotation": "not_applicable",
      "fidelity": "pass",
      "reason": "Debugging OAuth identifies the topic immediately and accurately frames investigation of callback errors.",
      "model_file": "gemini",
      "case_id": "auth-oauth",
      "repeat": 2,
      "arm": "override",
      "title": "🔍 Debugging OAuth callback 400 errors",
      "characters": 37,
      "latency_seconds": 1.227,
      "errors": []
    },
    {
      "id": "b034",
      "parse": "partial",
      "frontload": "pass",
      "connotation": "fail",
      "fidelity": "fail",
      "reason": "Bugfix conflicts with Feedback Summary, adds repair work, and recasts mixed feedback as solely a defect.",
      "model_file": "gemini",
      "case_id": "feedback-mixed",
      "repeat": 2,
      "arm": "default",
      "title": "🐛 Bugfix: Chess Checkmate Rule Feedback Summary",
      "characters": 47,
      "latency_seconds": 1.414,
      "errors": []
    },
    {
      "id": "b035",
      "parse": "pass",
      "frontload": "partial",
      "connotation": "pass",
      "fidelity": "pass",
      "reason": "Praise remains positive, but Summarizing customer delays the distinguishing offline-lessons topic.",
      "model_file": "gemini",
      "case_id": "feedback-praise",
      "repeat": 2,
      "arm": "override",
      "title": "📈 Summarizing customer praise for offline lessons",
      "characters": 49,
      "latency_seconds": 0.865,
      "errors": []
    },
    {
      "id": "b036",
      "parse": "pass",
      "frontload": "pass",
      "connotation": "not_applicable",
      "fidelity": "pass",
      "reason": "Inspect Docker clearly names the requested inspection topic without implying deletion.",
      "model_file": "gemini",
      "case_id": "negation",
      "repeat": 1,
      "arm": "default",
      "title": "🔧 Inspect Docker Disk Usage",
      "characters": 27,
      "latency_seconds": 0.707,
      "errors": []
    },
    {
      "id": "b037",
      "parse": "pass",
      "frontload": "pass",
      "connotation": "not_applicable",
      "fidelity": "pass",
      "reason": "Inspect Docker clearly names the requested inspection topic without implying deletion.",
      "model_file": "gemini",
      "case_id": "negation",
      "repeat": 2,
      "arm": "default",
      "title": "🔧 Inspect Docker Disk Usage",
      "characters": 27,
      "latency_seconds": 1.322,
      "errors": []
    },
    {
      "id": "b038",
      "parse": "pass",
      "frontload": "pass",
      "connotation": "fail",
      "fidelity": "fail",
      "reason": "Fix Chess identifies the topic early but adds unrequested repair and frames mixed feedback around a defect.",
      "model_file": "gemini",
      "case_id": "feedback-mixed",
      "repeat": 1,
      "arm": "default",
      "title": "🐛 Fix Chess Checkmate Logic & Summarize Feedback",
      "characters": 48,
      "latency_seconds": 1.745,
      "errors": []
    },
    {
      "id": "b039",
      "parse": "pass",
      "frontload": "pass",
      "connotation": "not_applicable",
      "fidelity": "pass",
      "reason": "Upgrading Pydantic immediately identifies the package and retains testing.",
      "model_file": "gemini",
      "case_id": "upgrade-pydantic",
      "repeat": 1,
      "arm": "override",
      "title": "⬆️ Upgrading Pydantic and running tests",
      "characters": 39,
      "latency_seconds": 1.428,
      "errors": []
    },
    {
      "id": "b040",
      "parse": "pass",
      "frontload": "pass",
      "connotation": "not_applicable",
      "fidelity": "pass",
      "reason": "Explaining Rust identifies the topic immediately, and compilation misses remain understandable in context.",
      "model_file": "gemini",
      "case_id": "filler-build",
      "repeat": 2,
      "arm": "override",
      "title": "🦀 Explaining Rust incremental compilation misses",
      "characters": 48,
      "latency_seconds": 0.849,
      "errors": []
    },
    {
      "id": "b041",
      "parse": "pass",
      "frontload": "partial",
      "connotation": "pass",
      "fidelity": "pass",
      "reason": "Praise remains positive, but Summarizing customer delays the distinguishing offline-lessons topic.",
      "model_file": "gemini",
      "case_id": "feedback-praise",
      "repeat": 1,
      "arm": "override",
      "title": "📝 Summarizing customer praise for offline lessons",
      "characters": 49,
      "latency_seconds": 1.133,
      "errors": []
    },
    {
      "id": "b042",
      "parse": "pass",
      "frontload": "pass",
      "connotation": "not_applicable",
      "fidelity": "pass",
      "reason": "Writing login identifies the subject early and correctly makes release-note writing the current task.",
      "model_file": "gemini",
      "case_id": "completed",
      "repeat": 2,
      "arm": "override",
      "title": "📝 Writing login fix release note",
      "characters": 32,
      "latency_seconds": 0.823,
      "errors": []
    },
    {
      "id": "b043",
      "parse": "pass",
      "frontload": "pass",
      "connotation": "not_applicable",
      "fidelity": "pass",
      "reason": "Write Login identifies the subject early and accurately requests a release note about the fix.",
      "model_file": "gemini",
      "case_id": "completed",
      "repeat": 2,
      "arm": "default",
      "title": "📝 Write Login Fix Release Note",
      "characters": 30,
      "latency_seconds": 0.921,
      "errors": []
    },
    {
      "id": "b044",
      "parse": "pass",
      "frontload": "pass",
      "connotation": "not_applicable",
      "fidelity": "pass",
      "reason": "Upgrade Pillow immediately identifies the package and retains the requested tests.",
      "model_file": "gemini",
      "case_id": "upgrade-pillow",
      "repeat": 2,
      "arm": "default",
      "title": "📦️ Upgrade Pillow and Run Tests",
      "characters": 31,
      "latency_seconds": 1.371,
      "errors": []
    },
    {
      "id": "b045",
      "parse": "pass",
      "frontload": "pass",
      "connotation": "not_applicable",
      "fidelity": "pass",
      "reason": "Investigating login identifies the topic immediately and preserves the password-reset trigger without claiming a cause.",
      "model_file": "gemini",
      "case_id": "auth-login",
      "repeat": 2,
      "arm": "override",
      "title": "🔑 Investigating login failure after password reset",
      "characters": 50,
      "latency_seconds": 1.245,
      "errors": []
    },
    {
      "id": "b046",
      "parse": "pass",
      "frontload": "pass",
      "connotation": "pass",
      "fidelity": "pass",
      "reason": "Comparing SQLite introduces the database comparison immediately without implying a selection.",
      "model_file": "gemini",
      "case_id": "neutral-compare",
      "repeat": 1,
      "arm": "override",
      "title": "📊 Comparing SQLite and PostgreSQL for offline app",
      "characters": 49,
      "latency_seconds": 1.392,
      "errors": []
    },
    {
      "id": "b047",
      "parse": "pass",
      "frontload": "pass",
      "connotation": "not_applicable",
      "fidelity": "pass",
      "reason": "Explaining Docker identifies the topic immediately and accurately describes inspection guidance.",
      "model_file": "gemini",
      "case_id": "negation",
      "repeat": 2,
      "arm": "override",
      "title": "🐳 Explaining Docker disk usage inspection",
      "characters": 41,
      "latency_seconds": 0.752,
      "errors": []
    },
    {
      "id": "b048",
      "parse": "pass",
      "frontload": "partial",
      "connotation": "pass",
      "fidelity": "pass",
      "reason": "Mixed preserves the two-sided sentiment, but Summarizing mixed delays the chess topic beyond two words.",
      "model_file": "gemini",
      "case_id": "feedback-mixed",
      "repeat": 1,
      "arm": "override",
      "title": "♟️ Summarizing mixed chess lesson feedback",
      "characters": 42,
      "latency_seconds": 1.23,
      "errors": []
    },
    {
      "id": "b049",
      "parse": "pass",
      "frontload": "pass",
      "connotation": "pass",
      "fidelity": "pass",
      "reason": "Offline appears within the first two words, and praise preserves the positive feedback.",
      "model_file": "gemini",
      "case_id": "feedback-praise",
      "repeat": 2,
      "arm": "default",
      "title": "✨ Summarize Offline Lessons Praise",
      "characters": 34,
      "latency_seconds": 1.617,
      "errors": []
    },
    {
      "id": "b050",
      "parse": "pass",
      "frontload": "partial",
      "connotation": "pass",
      "fidelity": "pass",
      "reason": "Complaints remain negative, but the distinguishing energy-limit topic follows the first two words.",
      "model_file": "gemini",
      "case_id": "feedback-complaint",
      "repeat": 1,
      "arm": "default",
      "title": "📝 Summarize Lesson Energy Limit Complaints",
      "characters": 42,
      "latency_seconds": 1.055,
      "errors": []
    },
    {
      "id": "b051",
      "parse": "pass",
      "frontload": "pass",
      "connotation": "not_applicable",
      "fidelity": "pass",
      "reason": "Investigating TLS identifies the topic immediately without asserting rotation caused the failures.",
      "model_file": "gemini",
      "case_id": "uncertainty",
      "repeat": 2,
      "arm": "override",
      "title": "🌐 Investigating TLS rotation in webhook failures",
      "characters": 48,
      "latency_seconds": 0.734,
      "errors": []
    },
    {
      "id": "b052",
      "parse": "pass",
      "frontload": "pass",
      "connotation": "not_applicable",
      "fidelity": "pass",
      "reason": "Investigating Zephyr identifies the actual Unicode CSV crash request despite the lengthy background.",
      "model_file": "gemini",
      "case_id": "long-tail",
      "repeat": 2,
      "arm": "override",
      "title": "🔍 Investigating Zephyr Unicode CSV crash",
      "characters": 40,
      "latency_seconds": 0.849,
      "errors": []
    },
    {
      "id": "b053",
      "parse": "pass",
      "frontload": "pass",
      "connotation": "pass",
      "fidelity": "pass",
      "reason": "SQLite leads a clear, neutral database comparison that retains the offline-first context.",
      "model_file": "gemini",
      "case_id": "neutral-compare",
      "repeat": 1,
      "arm": "default",
      "title": "🗃️ SQLite vs PostgreSQL for Offline-First App",
      "characters": 45,
      "latency_seconds": 0.762,
      "errors": []
    },
    {
      "id": "b054",
      "parse": "pass",
      "frontload": "pass",
      "connotation": "not_applicable",
      "fidelity": "pass",
      "reason": "The descriptive title begins with Rust Incremental and clearly names compilation and cache misses.",
      "model_file": "gemini",
      "case_id": "filler-build",
      "repeat": 1,
      "arm": "default",
      "title": "⚡️ Rust Incremental Compilation & Cache Misses",
      "characters": 46,
      "latency_seconds": 0.748,
      "errors": []
    },
    {
      "id": "b055",
      "parse": "pass",
      "frontload": "pass",
      "connotation": "not_applicable",
      "fidelity": "pass",
      "reason": "Upgrading Pillow immediately identifies the package and retains testing.",
      "model_file": "gemini",
      "case_id": "upgrade-pillow",
      "repeat": 2,
      "arm": "override",
      "title": "🖼️ Upgrading Pillow and running tests",
      "characters": 37,
      "latency_seconds": 0.684,
      "errors": []
    },
    {
      "id": "b056",
      "parse": "pass",
      "frontload": "pass",
      "connotation": "not_applicable",
      "fidelity": "pass",
      "reason": "Upgrading Pillow immediately identifies the package and retains testing.",
      "model_file": "gemini",
      "case_id": "upgrade-pillow",
      "repeat": 1,
      "arm": "override",
      "title": "🖼️ Upgrading Pillow and running tests",
      "characters": 37,
      "latency_seconds": 1.66,
      "errors": []
    },
    {
      "id": "b057",
      "parse": "pass",
      "frontload": "pass",
      "connotation": "not_applicable",
      "fidelity": "pass",
      "reason": "Write Login identifies the subject early and accurately requests a release note about the fix.",
      "model_file": "gemini",
      "case_id": "completed",
      "repeat": 1,
      "arm": "default",
      "title": "📝 Write Login Fix Release Note",
      "characters": 30,
      "latency_seconds": 0.677,
      "errors": []
    },
    {
      "id": "b058",
      "parse": "pass",
      "frontload": "partial",
      "connotation": "pass",
      "fidelity": "pass",
      "reason": "Complaints remain negative, but the distinguishing energy-limit topic follows the first two words.",
      "model_file": "gemini",
      "case_id": "feedback-complaint",
      "repeat": 1,
      "arm": "override",
      "title": "📉 Summarizing lesson energy limit complaints",
      "characters": 44,
      "latency_seconds": 0.844,
      "errors": []
    },
    {
      "id": "b059",
      "parse": "pass",
      "frontload": "pass",
      "connotation": "not_applicable",
      "fidelity": "fail",
      "reason": "Fix Zephyr is understandable and distinctive, but changes the requested investigation into repair.",
      "model_file": "gemini",
      "case_id": "long-tail",
      "repeat": 1,
      "arm": "default",
      "title": "🐛 Fix Zephyr Unicode CSV Parser Crash",
      "characters": 37,
      "latency_seconds": 0.732,
      "errors": []
    },
    {
      "id": "b060",
      "parse": "pass",
      "frontload": "pass",
      "connotation": "not_applicable",
      "fidelity": "pass",
      "reason": "Redis Cache immediately identifies the subject, and the descriptive title accurately names the policy.",
      "model_file": "gemini",
      "case_id": "filler-cache",
      "repeat": 2,
      "arm": "default",
      "title": "🗃️ Redis Cache Eviction and allkeys-lru Policy",
      "characters": 46,
      "latency_seconds": 0.852,
      "errors": []
    },
    {
      "id": "b061",
      "parse": "pass",
      "frontload": "pass",
      "connotation": "not_applicable",
      "fidelity": "pass",
      "reason": "Debugging OAuth identifies the topic immediately and accurately frames investigation of callback errors.",
      "model_file": "gemini",
      "case_id": "auth-oauth",
      "repeat": 2,
      "arm": "default",
      "title": "👔 Debugging OAuth Callback 400 Errors",
      "characters": 37,
      "latency_seconds": 0.739,
      "errors": []
    },
    {
      "id": "b062",
      "parse": "pass",
      "frontload": "partial",
      "connotation": "not_applicable",
      "fidelity": "pass",
      "reason": "The five-minute expiration symptom is accurate, but the session topic follows Debugging 5-Minute.",
      "model_file": "gemini",
      "case_id": "auth-session",
      "repeat": 2,
      "arm": "default",
      "title": "👔 Debugging 5-Minute Session Expiration",
      "characters": 39,
      "latency_seconds": 0.698,
      "errors": []
    },
    {
      "id": "b063",
      "parse": "pass",
      "frontload": "pass",
      "connotation": "not_applicable",
      "fidelity": "pass",
      "reason": "Debugging OAuth identifies the topic immediately and accurately frames investigation of callback errors.",
      "model_file": "gemini",
      "case_id": "auth-oauth",
      "repeat": 1,
      "arm": "default",
      "title": "🐛 Debugging OAuth Callback 400 Errors",
      "characters": 37,
      "latency_seconds": 1.564,
      "errors": []
    },
    {
      "id": "b064",
      "parse": "pass",
      "frontload": "pass",
      "connotation": "not_applicable",
      "fidelity": "pass",
      "reason": "Upgrade urllib3 immediately identifies the package and retains the requested tests.",
      "model_file": "gemini",
      "case_id": "upgrade-urllib3",
      "repeat": 2,
      "arm": "default",
      "title": "📦️ Upgrade urllib3 and Run Tests",
      "characters": 32,
      "latency_seconds": 0.776,
      "errors": []
    }
  ],
  "groups": [
    {
      "id": "a-g01",
      "consistent": "fail",
      "variation": "partial",
      "reason": "Two Debug imperatives alternate with a Session Timeout Investigation noun title, and only the session title omits its specific trigger.",
      "model_file": "haiku",
      "arm": "default",
      "repeat": 1,
      "family": "auth"
    },
    {
      "id": "a-g02",
      "consistent": "pass",
      "variation": "pass",
      "reason": "All three use Upgrade plus package and testing; the package names remain distinct and Tests is a compatible shortening.",
      "model_file": "haiku",
      "arm": "default",
      "repeat": 1,
      "family": "upgrade"
    },
    {
      "id": "a-g03",
      "consistent": "not_applicable",
      "variation": "partial",
      "reason": "Complaint, praise, and chess feedback remain distinguishable, but truncation hides the mixed feedback's negative side.",
      "model_file": "haiku",
      "arm": "default",
      "repeat": 1,
      "family": "feedback"
    },
    {
      "id": "a-g04",
      "consistent": "fail",
      "variation": "partial",
      "reason": "Two Investigate imperatives alternate with a topic-first Investigation noun title, although all three issue distinctions remain visible.",
      "model_file": "haiku",
      "arm": "default",
      "repeat": 2,
      "family": "auth"
    },
    {
      "id": "a-g05",
      "consistent": "pass",
      "variation": "pass",
      "reason": "All three use Upgrade plus package and testing; the package names remain distinct and Tests is a compatible shortening.",
      "model_file": "haiku",
      "arm": "default",
      "repeat": 2,
      "family": "upgrade"
    },
    {
      "id": "a-g06",
      "consistent": "not_applicable",
      "variation": "partial",
      "reason": "Energy complaints and offline praise are distinct, but the truncated chess title does not expose its negative feedback.",
      "model_file": "haiku",
      "arm": "default",
      "repeat": 2,
      "family": "feedback"
    },
    {
      "id": "a-g07",
      "consistent": "pass",
      "variation": "pass",
      "reason": "All three use Investigating plus the issue and retain distinct login, session, and OAuth symptoms despite the clipped login qualifier.",
      "model_file": "haiku",
      "arm": "override",
      "repeat": 1,
      "family": "auth"
    },
    {
      "id": "a-g08",
      "consistent": "partial",
      "variation": "partial",
      "reason": "The Upgrading pattern is shared, but patch release, patch version, and to latest patch introduce uneven wording and specificity.",
      "model_file": "haiku",
      "arm": "override",
      "repeat": 1,
      "family": "upgrade"
    },
    {
      "id": "a-g09",
      "consistent": "not_applicable",
      "variation": "partial",
      "reason": "Energy interruptions, offline praise, and chess lessons are distinct, but the clipped chess clause hides the complaint.",
      "model_file": "haiku",
      "arm": "override",
      "repeat": 1,
      "family": "feedback"
    },
    {
      "id": "a-g10",
      "consistent": "pass",
      "variation": "partial",
      "reason": "All three use Investigating plus the issue, but session expiry omits its timing while login and OAuth retain specific circumstances.",
      "model_file": "haiku",
      "arm": "override",
      "repeat": 2,
      "family": "auth"
    },
    {
      "id": "a-g11",
      "consistent": "partial",
      "variation": "partial",
      "reason": "All three begin Upgrading plus package, but only urllib3 mentions testing and the others vary between patch version and patch release.",
      "model_file": "haiku",
      "arm": "override",
      "repeat": 2,
      "family": "upgrade"
    },
    {
      "id": "a-g12",
      "consistent": "not_applicable",
      "variation": "partial",
      "reason": "The topics remain distinct, but chess validation loses the mixed sentiment that differentiates it from straightforward praise or complaint.",
      "model_file": "haiku",
      "arm": "override",
      "repeat": 2,
      "family": "feedback"
    },
    {
      "id": "b-g09",
      "consistent": "not_applicable",
      "variation": "pass",
      "reason": "The summaries preserve distinct topics and explicitly distinguish complaints, praise, and mixed feedback.",
      "model_file": "gemini",
      "arm": "override",
      "repeat": 1,
      "family": "feedback"
    },
    {
      "id": "b-g12",
      "consistent": "not_applicable",
      "variation": "partial",
      "reason": "Topics remain distinct, but generic chess feedback loses the mixed sentiment retained by the other sentiment labels.",
      "model_file": "gemini",
      "arm": "override",
      "repeat": 2,
      "family": "feedback"
    },
    {
      "id": "b-g06",
      "consistent": "not_applicable",
      "variation": "fail",
      "reason": "The chess title invents a bugfix task, creating a false action distinction among feedback summaries.",
      "model_file": "gemini",
      "arm": "default",
      "repeat": 2,
      "family": "feedback"
    },
    {
      "id": "b-g02",
      "consistent": "pass",
      "variation": "pass",
      "reason": "All use Upgrade plus package and testing; Tests is a compatible shortening of Regression Tests under the same rule used for Haiku.",
      "model_file": "gemini",
      "arm": "default",
      "repeat": 1,
      "family": "upgrade"
    },
    {
      "id": "b-g10",
      "consistent": "partial",
      "variation": "pass",
      "reason": "OAuth switches from Investigating to Debugging, while all three distinct authentication symptoms remain visible.",
      "model_file": "gemini",
      "arm": "override",
      "repeat": 2,
      "family": "auth"
    },
    {
      "id": "b-g04",
      "consistent": "pass",
      "variation": "pass",
      "reason": "All titles use Debugging and retain meaningful distinctions among reset-related login, timed expiration, and OAuth errors.",
      "model_file": "gemini",
      "arm": "default",
      "repeat": 2,
      "family": "auth"
    },
    {
      "id": "b-g05",
      "consistent": "pass",
      "variation": "pass",
      "reason": "All use Upgrade plus package and testing; Tests is a compatible shortening of Regression Tests under the same rule used for Haiku.",
      "model_file": "gemini",
      "arm": "default",
      "repeat": 2,
      "family": "upgrade"
    },
    {
      "id": "b-g03",
      "consistent": "not_applicable",
      "variation": "fail",
      "reason": "The chess title adds unrequested repair, creating a false distinction from the other feedback-summary tasks.",
      "model_file": "gemini",
      "arm": "default",
      "repeat": 1,
      "family": "feedback"
    },
    {
      "id": "b-g01",
      "consistent": "pass",
      "variation": "pass",
      "reason": "All titles use Debugging with parallel symptom descriptions that distinguish login, session expiration, and OAuth.",
      "model_file": "gemini",
      "arm": "default",
      "repeat": 1,
      "family": "auth"
    },
    {
      "id": "b-g08",
      "consistent": "pass",
      "variation": "pass",
      "reason": "All titles share Upgrading followed by the package and running tests, with meaningful package differences preserved.",
      "model_file": "gemini",
      "arm": "override",
      "repeat": 1,
      "family": "upgrade"
    },
    {
      "id": "b-g11",
      "consistent": "pass",
      "variation": "pass",
      "reason": "All titles share Upgrading followed by the package and running tests, with meaningful package differences preserved.",
      "model_file": "gemini",
      "arm": "override",
      "repeat": 2,
      "family": "upgrade"
    },
    {
      "id": "b-g07",
      "consistent": "pass",
      "variation": "partial",
      "reason": "All titles use Investigating and distinguish their subjects, but truncation leaves the login trigger incompletely expressed.",
      "model_file": "gemini",
      "arm": "override",
      "repeat": 1,
      "family": "auth"
    }
  ],
  "adjudication": "Post-blind cross-session consistency audit by SmolPaws. Original blind judgments preserved in judgments-original.json. Published scorecard uses these adjudicated judgments; changes and rule clarifications are explicit in adjudication.json."
}