Original research titles are collected from open-access scholarly repositories and processed through a structured LLM-based reformulation pipeline to produce stylistically diverse variants.
Data Generation
Each original title is transformed into 30 reformulated variants:
Technical Titles
10 variants — formal, methodology-focused, domain-specific language targeting expert audiences.
Accessible Titles
10 variants — plain-language rewrites designed for broad comprehension without prior domain knowledge.
Catchy Titles
10 variants — creative, hook-driven titles using analogies and engaging language to spark curiosity.
Data Availability
- Training, validation, and test splits will be released according to the shared-task schedule
- Access is provided to registered participants through the official task platform
- All source titles are drawn from open-access repositories
See Important Dates for the full release schedule.