Add humanoid instruction validator model

Initial release of a model that evaluates and classifies natural language instructions before execution for humanoid and agent systems.

README.md

@@ -0,0 +1,34 @@
+---
+language:
+- en
+- id
+license: mit
+tags:
+- humanoid
+- instruction-validation
+- safety
+- reasoning
+- llm
+---
+
+# humanoid-instruction-validator
+
+## Model Description
+`humanoid-instruction-validator` is a language model designed to evaluate natural language instructions before execution.
+
+The model does not perform the requested task. Instead, it analyzes the instruction and determines whether it is valid, ambiguous, contradictory, incomplete, or unsafe. This enables safer decision-making for humanoid and agent systems.
+
+## Intended Use
+- Pre-execution instruction validation
+- Humanoid and robotics command filtering
+- AI agent safety layers
+- Reasoning and alignment systems
+
+## Output Format
+The model outputs **JSON only** with the following structure:
+
+```json
+{
+  "label": "VALID | AMBIGUOUS | CONTRADICTORY | INCOMPLETE | UNSAFE",
+  "confidence": 0.0
+}

adapter.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c847e4a9a806f42f190f96d27a10a6172491e003bef10ffc051d8a2ef3859977
+size 20

config.json

@@ -0,0 +1,41 @@
+
+---
+
+# 📄 2️⃣ config.json (MODEL CONFIGURATION)
+
+**Upload this as `config.json`**
+
+```json
+{
+  "model_type": "instruction_validator",
+  "task": "instruction_validation",
+  "languages": ["en", "id"],
+  "output_format": "json",
+  "output_schema": {
+    "label": {
+      "type": "string",
+      "enum": [
+        "VALID",
+        "AMBIGUOUS",
+        "CONTRADICTORY",
+        "INCOMPLETE",
+        "UNSAFE"
+      ]
+    },
+    "confidence": {
+      "type": "number",
+      "minimum": 0.0,
+      "maximum": 1.0
+    }
+  },
+  "inference_settings": {
+    "json_only": true,
+    "temperature": 0.0,
+    "max_tokens": 64
+  },
+  "intended_use": [
+    "humanoid_pre_execution_validation",
+    "agent_safety_filter"
+  ],
+  "license": "mit"
+}