Hugging Face's logo

AIMS2025
/
DeepSignal

Text Generation
GGUF
English
qwen3
llama.cpp
lmstudio
traffic-signal-control
simulation
conversational
Model card Files
xet
Community
AIMS2025 commited on
Commit
eb0e3d0
·
verified ·
1 Parent(s): e2d7816

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +7 -7
README.md CHANGED
@@ -37,14 +37,14 @@ where the number is the phase index (starting from 0) and the seconds is the dur
37
 
38
  ### Performance Metrics Comparison by Model *
39
 
40
- | Model | Avg Saturation | Avg Queue Length (veh/s) | Avg Throughput (veh/5min) | Avg Response Time (s) |
41
  |:---:|:---:|:---:|:---:|:---:|
42
- | [`GPT-OSS-20B (thinking)`](https://huggingface.co/openai/gpt-oss-20b) | 0.380 | 0.476 | 77.910 | 6.768 |
43
- | **DeepSignal-4B (Ours)** | 0.422 | 0.498 | **79.883** | 2.131 |
44
- | [`Qwen3-30B-A3B`](https://huggingface.co/Qwen/Qwen3-VL-30B-A3B-Instruct) | 0.431 | 0.580 | 79.059 | 2.727 |
45
- | [`Qwen3-4B`](https://huggingface.co/Qwen/Qwen3-4B-Instruct-2507) | 0.466 | 2.454 | 75.712 | 1.994 |
46
- | Max Pressure | 0.465 | 0.640 | 77.236 | ** |
47
- | [`LightGPT-8B-Llama3`](https://huggingface.co/lightgpt/LightGPT-8B-Llama3) | 0.523 | 1.259 | 75.512 | 3.025*** |
48
 
49
  `*`: Each simulation scenario runs for 60 minutes. We discard the first **5 minutes** as warm-up, then compute metrics over the next **20 minutes** (minute 5 to 25). We cap the evaluation window because, when an LLM controls signal timing for only a single intersection, spillback from neighboring intersections may occur after ~20+ minutes and destabilize the scenario. All evaluations are conducted on a **Mac Studio M3 Ultra**.
50
  `**`: Max Pressure is a fixed signal-timing optimization algorithm (not an LLM), so we omit its Avg Response Time; this metric is only defined for LLM-based signal-timing optimization.
 
37
 
38
  ### Performance Metrics Comparison by Model *
39
 
40
+ | Model | Avg Saturation | Avg Queue Length (veh/min) | Avg Throughput (veh/5min) | Avg Response Time (s) |
41
  |:---:|:---:|:---:|:---:|:---:|
42
+ | [`GPT-OSS-20B (thinking)`](https://huggingface.co/openai/gpt-oss-20b) | 0.380 | 14.088 | 77.910 | 6.768 |
43
+ | **DeepSignal-4B (Ours)** | 0.422 | 15.703 | **79.883** | 2.131 |
44
+ | [`Qwen3-30B-A3B`](https://huggingface.co/Qwen/Qwen3-VL-30B-A3B-Instruct) | 0.431 | 17.046 | 79.059 | 2.727 |
45
+ | [`Qwen3-4B`](https://huggingface.co/Qwen/Qwen3-4B-Instruct-2507) | 0.466 | 57.699 | 75.712 | 1.994 |
46
+ | Max Pressure | 0.465 | 23.022 | 77.236 | ** |
47
+ | [`LightGPT-8B-Llama3`](https://huggingface.co/lightgpt/LightGPT-8B-Llama3) | 0.523 | 54.384 | 75.512 | 3.025*** |
48
 
49
  `*`: Each simulation scenario runs for 60 minutes. We discard the first **5 minutes** as warm-up, then compute metrics over the next **20 minutes** (minute 5 to 25). We cap the evaluation window because, when an LLM controls signal timing for only a single intersection, spillback from neighboring intersections may occur after ~20+ minutes and destabilize the scenario. All evaluations are conducted on a **Mac Studio M3 Ultra**.
50
  `**`: Max Pressure is a fixed signal-timing optimization algorithm (not an LLM), so we omit its Avg Response Time; this metric is only defined for LLM-based signal-timing optimization.