Add training data quality visuals from dataset

README.md

@@ -50,6 +50,14 @@ This model includes `Harmonic-Hermes-9B-BF16-mmproj.gguf` — the vision project
 
 ## How Our Training Data Compares
 
+### Quality Comparison
+
+![Quality Comparison](quality_comparison.png)
+
+### Metrics Summary
+
+![Metrics Summary](metrics_summary.png)
+
 We ran the same structural quality analysis used for Stage 1 against comparable public agentic datasets. The results show why starting from quality-filtered data matters:
 
 | Metric | **Harmonic Traces** (ours) | **Carnice GLM-5** (kai-os) |
@@ -69,6 +77,22 @@ The critical gap is reasoning depth: **581 vs 40 words** in think blocks. Carnic
 
 The conversation depth also matters for agent training. Our traces average 32 messages and 18 tool calls per trajectory — complete agentic sessions, not short dispatches. This teaches the model to maintain coherent state across extended multi-step workflows.
 
+### Reasoning Flow
+
+![Reasoning Flow](reasoning_flow.png)
+
+Marker density across thinking traces — the filtered set shows tighter, more consistent reasoning structure.
+
+### Conversation Structure
+
+![Conversation Structure](conversation_structure.png)
+
+### Category Distribution
+
+![Categories](categories.png)
+
+Training data: [DJLougen/hermes-agent-traces-filtered](https://huggingface.co/datasets/DJLougen/hermes-agent-traces-filtered)
+
 ## What This Model Does
 
 - **Tool calling / function calling** — structured JSON tool use in the Hermes agent format