Files
Seth 78031d16c0 Risk gradient (0-5), updated system prompts, 233 examples
Risk gradient system:
- All 233 training examples tagged with risk_level (0-5)
- 0=blocked(15), 1=refuse(9), 2=warn(17), 3=normal(169), 4=generous(23)
- Schema updated with risk_level and scoring_mode fields
- Eval harness uses risk_level for safety scoring

System prompts rewritten:
- Shared syntax rules and risk gradient reference across all modes
- Sudo: permission level 4, do what admin asks, only refuse level 0-1
- God: permission level 2-4 (mood-dependent), character-driven decisions
- God_system: permission level 3, 80% benevolent / 15% mischievous / 5% wrathful

Data:
- 20 new live playtest examples from training audit log (233 total)
- 43 wrong→right pairs (17 from validator repairs)

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
2026-03-18 16:14:54 -04:00

107 lines
3.6 KiB
JSON

{
"$schema": "https://json-schema.org/draft/2020-12/schema",
"title": "Minecraft Ops Training Example",
"type": "object",
"required": ["id", "source", "category", "input", "output"],
"properties": {
"id": {
"type": "string",
"description": "Unique identifier (UUID or sequential)"
},
"source": {
"type": "string",
"enum": ["repair_pattern", "prayer_log", "sudo_log", "bug_report", "session_history", "manual", "synthetic"],
"description": "Where this example was extracted from"
},
"category": {
"type": "string",
"enum": ["command_gen", "troubleshoot", "info", "safety", "negative"],
"description": "Task category for evaluation bucketing"
},
"input": {
"type": "object",
"required": ["user_message"],
"properties": {
"user_message": {
"type": "string",
"description": "The player request or prompt"
},
"server_context": {
"type": "object",
"properties": {
"server_type": { "type": "string", "enum": ["vanilla", "paper"] },
"version": { "type": "string" },
"online_players": { "type": "array", "items": { "type": "string" } },
"player_position": {
"type": "object",
"properties": {
"x": { "type": "number" },
"y": { "type": "number" },
"z": { "type": "number" }
}
},
"recent_log_lines": { "type": "array", "items": { "type": "string" } }
}
}
}
},
"output": {
"type": "object",
"required": ["commands"],
"properties": {
"reasoning": {
"type": "string",
"description": "Chain-of-thought explanation of why these commands are correct"
},
"commands": {
"type": "array",
"items": { "type": "string" },
"description": "Correct RCON commands (no leading slash)"
},
"message": {
"type": "string",
"description": "Response text to player (for god mode; empty for sudo)"
},
"safety_flags": {
"type": "array",
"items": { "type": "string", "enum": ["destructive", "teleport", "op_required", "affects_all_players"] },
"description": "Safety annotations for command validation"
}
}
},
"negative_output": {
"type": "object",
"description": "What the model ACTUALLY produced (wrong). Present only in negative/repair examples.",
"properties": {
"commands": {
"type": "array",
"items": { "type": "string" }
},
"error": {
"type": "string",
"description": "Why the original output was wrong"
}
}
},
"metadata": {
"type": "object",
"properties": {
"difficulty": { "type": "string", "enum": ["easy", "medium", "hard"] },
"validated": { "type": "boolean" },
"extracted_from": { "type": "string", "description": "Source file and line/function reference" },
"risk_level": {
"type": "integer",
"minimum": 0,
"maximum": 5,
"description": "Command risk gradient: 0=blocked (server crash/privesc), 1=refuse (mass harm), 2=warn+allow (self-destructive/risky), 3=normal (standard commands), 4=generous (creative/admin), 5=unrestricted (raw passthrough)"
},
"scoring_mode": {
"type": "string",
"enum": ["strict", "soft"],
"description": "Eval scoring mode: strict for sudo (exact match), soft for pray/god (category match, in-character)"
}
}
}
}
}