Dear All,
We are excited to announce the launch of the Third Scientific Figure Captioning Challenge (SciCap Challenge 2025)!
This year’s challenge focuses on personalized caption generation using our newly released LaMP-Cap dataset. The dataset is specifically curated for generating personalized figure captions with rich multimodal profiles.
The challenge will be hosted as part of the LM4Sci Workshop at COLM 2025, taking place October 7–10 in Montreal, Canada. We welcome participation from researchers, students, industry practitioners, and anyone interested in advancing personalized scientific communication.
IMPORTANT DATES (All times in Central Time Zone, CST = UTC−6):
- Challenge Submission Deadline: Monday, August 11, 2025
- Technical Report (Required) Submission Deadline: Friday, August 15, 2025
- Result Notification: Monday, September 1, 2025
- LM4Sci Workshop: Friday, October 10, 2025
CHALLENGE DETAILS:
- Two winning teams will be selected:
- Leaderboard Winner (based on automatic metrics): $300 USD
- Quality Winner (based on human evaluation): $300 USD
- The top 4 teams on the leaderboard and the Quality Winner will be invited to present posters in person at the workshop.
- All teams must submit a 2–4 page technical report describing their system design, methods, data processing, and results, using the COLM 2025 paper template.
- Important: Do NOT use the original author-written captions of the target figures in the test set as input to your system. These are for evaluation only. Use of gold captions as input will result in disqualification.
More information and dataset access: https://scicap.ai/
We hope to see many of you join this year’s challenge!
Best regards,
The SciCap Challenge Team
Ho Yin 'Sam' Ng (Penn State University)
Ting-Yao (Edward) Hsu (Penn State University)
Aashish Anantha Ramakrishnan (Penn State University)
Ryan Rossi (Adobe)
Sungchul Kim (Adobe)
Tong Yu (Adobe)
Ting-Hao 'Kenneth' Huang (Penn State University)