SUMO Traffic Simulation Agent Admin

Correction Rate
-
Share of reviewed runs that needed a trainable correction.
Mixed Fail Rate
-
Runs that failed across both parameter and geometry dimensions.
Simulations
-
Total scenario runs captured in the current dataset.
Parameter Corr.
-
Trainable parameter corrections logged so far.
Geometry Corr.
-
Geometry-level fixes captured from review sessions.

LLM Correction Evaluation

Start here when you want to understand where the model is still weak. The charts below show which fields fail most often and which geometry changes cluster together.

Field Error Counts

Geometry Correction Types

View average parameter delta (raw JSON) 
loading...

System Summary

Secondary operational context such as grade distribution and overall simulation fidelity.

Open supplementary metrics 
loading...

Recent Simulations

Use this table to inspect the latest prompts and compare their resulting grades.
IDCreatedGradePrompt

Recent Modifications

Review how users corrected or tuned outputs and whether those edits are reusable for retraining.
IDIntentTypeTrainablePrompt