Seth
78031d16c0
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>
107 lines
3.6 KiB
JSON
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)"
|
|
}
|
|
}
|
|
}
|
|
}
|
|
}
|