sgai editorial interpretation · Data updated 2026-05-02
🔬 Research Quality
Is original research coming out?
Target Progress
SEA-LION v4 (11 languages, 4B–33B params) + 100E (100+ projects) + ICLR 2025 hosted
Third-Party Ranking Anchors
- Wiley 2024 · #1 globally for AI papers per capita (250 per million people, 2022)
- CSRankings AI · NTU #3
- QS · NUS AI #9
Editorial Interpretation
Volume and university rankings are strong — per-capita papers #1, NTU AI #3, NUS #9, ICLR 2025 hosted, SEA-LION is one of the few non-US/UK/China foundation models at scale. But frontier-grade originality (FAIR / DeepMind tier) still trails by a step: first-author share at top venues, signature works with >1000 citations, market share of self-developed foundation models — all behind.
⚠️ Key Shortcoming
First-author share at top venues, citation counts, and self-developed foundation-model market share all trail by a step. Top PhD outflow is high. Research-to-industry transfer is strong for in-house enterprise use but weak as international export — no OpenAI / Anthropic-tier spinout. International visibility hinges on a small number of star professors.
Full Data
| Metric | Value | Source / Date |
|---|---|---|
| AI papers per capita | #1 globally (250 papers per million people, 2022) | Wiley, 2024-09 ↗ |
| NTU AI research ranking | #3 globally (after MIT and CMU) | Introl, 2025-08 ↗ |
| NUS AI academic reputation | #9 globally | Introl, 2025-08 ↗ |
| SEA-LION large language model | v4, 11+ languages, 4B-33B parameters | AISG, 2025 ↗ |
| 100 Experiments | 100+ AI projects completed (2018-2025, archived) | AISG ↗ |
| ICLR 2025 | Hosted in Singapore | ICLR, 2025 ↗ |
| DBS AI models | 800+ models, 350+ use cases, S$750M economic value generated in 2024 | Introl, 2025-08 ↗ |