GitHub: https://github.com/TheEighthDay/SeekWorld

SeekWorld_RL_PLUS

This is a direct RL model for geolocation reasoning on the SeekWorld dataset(Train-No-Process), based on Qwen2.5-VL-7B.

It can outperform the top-tier open-source and closed-source models such as Qwen-VL-32B-Instruct, GPT4o, QvQ-max, and Gemini-Flash-thinking!!!

Model Description

This model is used to identify geographical location information in pictures, including the country and the first-level administrative division (administrative_area_level_1).

Evaluation Results

Model	Global-Test	China-Test	Overall Accuracy
Bigger model
🔒GPT4o-240806	56.50	31.90	43.26
🔒Doubao-1.5-vision-pro-32k-250115	43.75	40.48	41.99
🔒🧠Gemini-2.0-flash-thinking-exp-01-21	56.25	29.49	41.85
🧠QvQ-72B-max-2025-03-25	48.13	31.63	39.25
Qwen-2.5-32B-VL	38.12	24.13	30.59
Small model (7B)
SeekWorld-7B [Cold-Start SFT + RL] (ours)	-	-	-
SeekWorld-7B [Direct RL] (ours)	59.69	34.65	46.21
Qwen-2.5-7B-VL [Direct RL]	51.25	31.90	40.84
Qwen-2.5-7B-VL [Direct SFT]	37.19	25.47	30.88
Qwen-2.5-7B-VL	33.44	24.40	28.57
Qwen-2.5-7B-VL + CoT	25.31	21.45	23.23

Models with 🔒 are **proprietary**, while those with 🧠 are enhanced with **thinking capabilities**.

• Eval Time: 2025-04-15 17:13:28