![Song Gao Profile](https://pbs.twimg.com/profile_images/784459351754801152/Vja7N1yM_x96.jpg)
Song Gao
@gissong
Followers
4K
Following
4K
Statuses
3K
Associate Professor of GIScience and Director of GeoDS Lab | Data Science Institute, Univ. of Wisconsin-Madison|UCSB STKO Lab|Spatial Data Science | GeoAI
Joined April 2009
@pnasnews Our research on “Intracounty modeling of #COVID19 infection with human mobility: Assessing spatial heterogeneity with business traffic, age, and race" is available on PNAS @UWMadison @UWMadScience @UWMadisonLS @UWMadisonGeog
2
10
48
RT @SIGSPATIAL_GIS: 📢 Call for Papers: ACM SIGSPATIAL 2025, Nov 3-6 in Minneapolis, MN! Submit your research on sp…
0
7
0
#ICLR2025 Very honor to be part of the ScienceAgentBench team by contributing GeoInformatics examples!
Our ScienceAgentBench in @Nature news! DeepSeek R1 @DeepSeekR1 vs. @OpenAI o1 on data-driven scientific coding tasks: We sampled 20 tasks from ScienceAgentBench, with 5 tasks from each of the four scientific disciplines (bioinformatics, comp. chemistry, geo info science, phych. & cog. neuroscience). 1. Performance: Given three attempts, DeepSeek R1 can solve 7 out of the 20 tasks, while OpenAI o1 only solves 6 of them. In addition, o1 is able to generate 12 executable programs, and R1 can generate 10 executable programs. This suggests that DeepSeek R1 achieves the same level of performance as OpenAI o1 on ScienceAgentBench. 2. Cost: We find that OpenAI o1 is around 13 times more expensive than DeepSeek R1. On average, o1 requires 0.13 USD to solve a task in ScienceAgentBench, while R1 only costs 0.01 USD. In fact, this is also cheaper than other proprietary models evaluated in our paper, such as GPT-4o and Claude-3.5-Sonnet. 3. Latency: Currently DeepSeek R1 takes a longer time to "think" (1-3min) than OpenAI o1 (<1min) on ScienceAgentBench tasks. Some promising future work could be to further improve the CoT reasoning efficiency with length control, to make the model even better and more practical for daily use. More details will be out soon! Great efforts led by our awesome @RonZiruChen and @ShijieChen98 @osunlp!
0
1
7
RT @RonZiruChen: 🎉ScienceAgentBench is accepted at #ICLR2025! 🚀 Ready to step beyond ML R&D? Test your agents on real-world, data-driven…
0
14
0
RT @GIScience_conf: Three weeks left to prepare your proceedings paper submission for #GIScience2025! https://t.co/…
0
4
0
RT @drjingjing2026: 1/3 Today, an anecdote shared by an invited speaker at #NeurIPS2024 left many Chinese scholars, myself included, feelin…
0
632
0
Our 'region2vec' method is designed explicitly for spatial networks. A related work on (non-spatial) network community detection via neural embeddings is published on Nature Communications:
#GeoAI enhanced community detection methods 'region2vec' on spatial networks using graph neural embedding by considering node attributes, geographic adjacency, and spatial interactions all together. Open source:
0
5
29
RT @datascience_uw: DSI Director @KyleCranmer spoke with Fox47 news about @UWMadison's #6 @nsfgov HERD research ranking. The university spe…
0
2
0
RT @Yingjie_Hu: I'm recruiting a PhD student starting in Fall 2025 (application deadline is Dec. 2024). The student will focus on GeoAI for…
0
181
0
RT @GISphereGuide: 🚨 Attention GIS students! 🚨 Join us for GISphere's 6th #GISalon webinar "Exploring Career Paths in GIS: Local Governmen…
0
10
0
Many thanks to 최경아 Dr. Kyoungah Choi from KRIHS, 송인상 Dr. Insang Song from Seoul National University, and 강영옥 Professor Youngok Kang from Ewha Womans University to host me in Seoul🇰🇷
It is my great honor and pleasure to be invited as a keynote speaker for the 2024 Korean Conference in Geospatial Information Science in conjunction with the industry annual event K-Geo-Festa! #GeoAI R&D has been rapidly growing in Korea through academia-industry collaboration!
1
0
13