The competition will task participants to provide critical information necessary for incident report generation. The prerequisite tasks are: 1. **Time-to-Incident Start Estimation (time2incident):** Predict the time a situation becomes hazardous by providing a frame number. 2. **Incident Detection:** Determine whether the video contains a **Hazard**, **Accident**, or **No incident**. 3. **Incident Recognition:** Classify the type of **Hazard** or **Accident** (Label). 4. **Ego-Car and Other Parties Involvement:** Dentify whether the **Ego vehicle** is involved, and specify if other **cars**, **objects**, or **animals** are part of the event. 5. **Crash Severity:** Assess the **level of danger**. 6. **Caption Before the incident** Provide a caption describing the video segment before the incident begins. 7. **Caption After the incident** Provide a caption describing the cause of the accident. More details on the exact format (and classification interpretations) can be found in the "Data" tab. Participant should use the outputs from these tasks along with language models to generate detailed, context-rich captions describing the incident. For this competition, we have compiled and annotated approximately 700 dashcam videos. Captions were produced by four human annotators and two large Vision-Language Models (with human review), ensuring a fair and comprehensive evaluation. We include this example report to provide a sense of how the inputs contestants submit will be formatted into a final report: Before the incident, **Ego-car is driving through a city road with heavy traffic**. At approximately frame 110, the incident started developing into an **Accident** under the classification **vehicle drives into another vehicle**. The incident severity was **4. Other cars collided with person/car/object but ego-car is ok**, and involved **two vehicles**. The cause of the incident was **a vehicle in front of ego-car tries turns left in front of oncoming traffic and collides with another vehicle**