Subway fires often cause significant casualties and property losses. There are some special bidirectional coupling scenarios during subway fires, such as firefighters moving against the evacuation flow to extinguish fires, emergency managers going to the fire scene to respond to emergencies, or other similar scenarios. How to evacuate passengers quickly and enable responders to enter the fire scene has become a big challenge for subway fire evacuation and response. The core goal is to reduce the degree of mutual interaction between these people moving in opposite directions. In this study, the impact of counterflow individuals and proactive avoidance behavior on evacuation processes was investigated through experiments and simulations. The Fire Dynamic Simulator was used to simulate the development of a fire scenario to determine the available safe egress time. Micro-evacuation experiments were conducted to obtain actual evacuation parameters, such as the speeds of different objects. With these parameters as input, a macro subway fire scenario was built to simulate the bidirectional evacuation process. Consistent conclusions were obtained from the experiments and evacuation simulations. The results indicate that the overall evacuation time increases with the number of retrograders. Proactive avoidance behavior can effectively reduce the travel time of counterflow individuals, but it causes slight delays for forward-moving evacuees. An optimization strategy was implemented through conductor guidance. All passengers can successfully evacuate under the optimization strategy, with a 25.3% improvement in overall evacuation time. This research provides some insights into the coordinated evacuation and emergency response during subway fires or similar scenarios.