Kuaishou’s TWICE splits ad conversion learning into two clocks
_reachsumit · x · 2026-07-29
Kuaishou proposes TWICE, a two-clock framework for long-horizon conversion prediction in online ads.
- Clicks and conversions supervise separate branches: a click clock for timely but partial supervision, and a conversion clock for delayed outcomes.
- The model factorizes long-horizon CVR into target-window conversion probability and a grouped elapsed-delay CDF.
- It also uses fixed predicted CVR mass as cohort exposure to handle mixed arrival-time slices with different traffic and propensity.
- The paper claims the resulting framework improves delayed-feedback conversion modeling at scale.
More from Research
- Paper finds semantic IDs lose fine-grained item detail in generative recommendation — _reachsumit · 2026-07-29
- LLM shopping agents let exaggerated claims win 73%–78% of top picks, paper finds — _reachsumit · 2026-07-29
- Protein-complex benchmark finds MSA depth matters more than correct pairing in AlphaFold3 — ferruz_noelia · 2026-07-29
- Kuaishou open-sources KuaiLive-M3, a 35M-interaction dataset for live-stream recommendation — _reachsumit · 2026-07-29
- Huawei’s SearchArt trains a 27B long-horizon search agent with verified synthetic tasks — _reachsumit · 2026-07-29
- Agent Retrieval Bench tests whether coding agents can find the right repo files before patching — _reachsumit · 2026-07-29