Tencent's KDD Cup Track Ends With $600K Prizes; Industrial Winners Built Everything on DeepSeek
新智元 · wechat · 2026-08-20
Tencent's advertising algorithm competition (TAAC), running as an official KDD Cup track, concluded in Jeju with 14,000 participants from 52 countries and 5,700+ teams, plus a total prize pool of 6M RMB — the academic champion alone took $300K, a record for Chinese corporate tech contests.
Tencent for the first time opened real anonymized advertising data serving billions of daily users (100+ feature fields). The task targeted unified sequential modeling and feature interaction for large-scale recommendation, plus exploring recommendation Scaling Laws. Academic champion lozyeah (a 3-person lab team) proposed CRAFT, borrowing flow matching to generate per-layer "intent vectors" that prevent information collapse across stacked blocks. Industrial champion sunshot's QueryFormer took a counterintuitive approach — craft a high-quality query before reading long sequences — winning with AUC 0.83254; notably, their entire modeling and development was done with only the DeepSeek web app, which also drafted most of their paper.
Both innovation awards went to undergrads and complete newcomers. Almost every winning team relied on an AI teammate, showing the competition itself being redefined by AI.
More from Companies & People
- 'Claude is an incredible model' — users say Anthropic is pulling ahead of OpenAI — iruletheworldmo · 2026-10-16
- Accenture caps employee Claude spending at $29/month — LinkedInNews · 2026-10-16
- Ex-Apple engineer recounts 6 years on Siri, from QA tooling to offline Siri — jdluk87 · 2026-10-16
- Chile's C³ index: 65% of firms train staff on AI but only 8% use it to innovate — robertonappe · 2026-10-16
- Musk teases "crazy stuff" at Tesla event, needs audience to prove it's not AI — XFreeze · 2026-10-16
- Synthwave is tech's current aesthetic, from Tesla Roadster to TBPN intros — granawkins · 2026-10-16