Upload README.md with huggingface_hub

README.md

@@ -27,19 +27,19 @@ model-index:
       type: custom
     metrics:
     - type: strict_pass_rate
-      value: 88.46
+      value: 92.31
       name: Strict Pass Rate
     - type: tool_call_parse_rate
       value: 100.0
       name: Tool Call Parse Rate
     - type: semantic_command_accuracy
-      value: 88.46
+      value: 92.31
       name: Semantic Command Accuracy
 ---
 
 # Kali-Terminus-v2
 
-**Kali-Terminus-v2** is a fine-tuned [LFM2.5-350M](https://huggingface.co/LiquidAI/LFM2.5-350M-Base) model specialized for Kali Linux terminal command execution. It converts natural language instructions into precise shell commands with the `exec` tool-call format.
+**Kali-Terminus-v2** is a fine-tuned [LFM2.5-350M](https://huggingface.co/LiquidAI/LFM2.5-350M-Base) model specialized for Kali Linux terminal command execution. It converts natural language instructions into precise shell commands with native LFM2.5 `<|tool_call_start|>`/`<|tool_call_end|>` JSON tool calls.
 
 This is the **Executor Agent** in a multi-agent penetration-testing system — it translates one bounded instruction into one structured terminal action.
 
@@ -62,10 +62,10 @@ Evaluated across 5 scenarios (26 turns) covering network recon, system enumerati
 
 | Metric | Score |
 |---|---|
-| **Strict Pass Rate** | **88.46%** |
+| **Strict Pass Rate** | **92.31%** |
 | Tool Call Parse Rate | 100.0% |
 | Exactly One Tool Call | 100.0% |
-| Semantic Command Accuracy | 88.46% |
+| Semantic Command Accuracy | 92.31% |
 
 ### Comparison with V1
 
@@ -73,7 +73,7 @@ Evaluated across 5 scenarios (26 turns) covering network recon, system enumerati
 |---|---|---|
 | Kali-Terminus-v1 (FunctionGemma-270M) | 65.8% | — |
 | Kali-Terminus-v1 (LFM2.5-350M) | 50.0% | — |
-| **Kali-Terminus-v2** | **88.46%** | **+38.5 pp over v1 LFM2.5** |
+| **Kali-Terminus-v2** | **92.31%** | **+42.3 pp over v1 LFM2.5** |
 
 ## Dataset
 
@@ -113,9 +113,29 @@ inputs = tokenizer.apply_chat_template(
 outputs = model.generate(**inputs, max_new_tokens=256)
 response = tokenizer.decode(outputs[0][inputs["input_ids"].shape[-1]:], skip_special_tokens=False)
 print(response)
-# Expected: <call:exec>{"command": "nmap -p 80 192.168.1.1"}</call:exec>
+# Expected: <|tool_call_start|>{"name": "exec", "arguments": {"command": "nmap -p 80 192.168.1.1"}}<|tool_call_end|>
 ```
 
+## Tool Call Format
+
+Kali-Terminus-v2 produces native **LFM2.5 tool calls** wrapped in `<|tool_call_start|>` and `<|tool_call_end|>` tokens, output as JSON:
+
+```
+<|tool_call_start|>{"name": "exec", "arguments": {"command": "nmap -p 80 192.168.1.1"}}<|tool_call_end|>
+```
+
+A system instruction `Output function calls as JSON.` is prepended so the model emits JSON rather than the default Pythonic calls.
+
+## Tool Call Format
+
+Kali-Terminus-v2 produces native **LFM2.5 tool calls** wrapped in `<|tool_call_start|>` and `<|tool_call_end|>` tokens, output as JSON:
+
+```
+<|tool_call_start|>{"name": "exec", "arguments": {"command": "nmap -p 80 192.168.1.1"}}<|tool_call_end|>
+```
+
+A system instruction `Output function calls as JSON.` is prepended so the model emits JSON rather than the default Pythonic calls.
+
 ## Intended Use
 
 - **Executor Agent** in multi-agent penetration-testing systems