DeepSignal-4B (GGUF)

This repository provides a GGUF model file for local inference (e.g., llama.cpp / LM Studio). It is intended for traffic-signal-control analysis and related text-generation workflows.

Files

• DeepSignal-4B_V1.F16.gguf
• config.json

Quickstart (llama.cpp)

llama-cli -m DeepSignal-4B_V1.F16.gguf -p "Summarize the traffic state and suggest a signal timing adjustment."

Evaluation (Traffic Simulation)

Performance Metrics Comparison by Model

Model	Avg Saturation	Avg Queue Length	Max Saturation	Max Queue Length	Avg Throughput	Avg Congestion Index
Qwen3-30B-A3B	0.1663	5.8604	0.1663	5.8604	9288.6633	0.1625
DeepSignal-4B	0.1657	5.8301	0.1657	5.8301	9302.7960	0.1752
LightGPT-8B-Llama3	0.1538	5.7688	0.1538	5.7688	5510.2010	0.2086
SFT	0.1604	6.0021	0.1604	6.0021	5390.4867	0.2093
Qwen3-4B	0.2152	8.2083	0.2152	8.2083	7112.8433	0.2522
Max Pressure	0.2059	8.1034	0.2059	8.1034	7395.1463	0.2556
GPT-OSS-20B	0.2591	10.4292	0.2591	10.4292	8742.7267	0.3175

Congestion Level Distribution by Model (%)

Model	Light congestion	Smooth	Very smooth
DeepSignal-4B	0.00	12.00	88.00
GPT-OSS-20B	2.00	53.33	44.67
LightGPT-8B-Llama3	0.00	21.00	79.00
Max Pressure	0.00	36.44	63.56
Qwen3-30B-A3B	0.00	10.00	90.00
Qwen3-4B	2.33	32.00	65.67
SFT	0.00	23.33	76.67

Visualization

Notes

• The results above are reported from a SUMO-based traffic simulation evaluation.
• If you need to reproduce the evaluation, include the exact scenario configuration, random seeds, and controller settings in a separate README or paper appendix.