Perplexity details training data pipeline: PII and opt-out sessions excluded
perplexity_ai · x · 2026-09-23
Perplexity shared more details on its Computer model training pipeline: after RL in synthetic environments, it trains on real-world sessions that expose failures beyond designed scenarios. Its sampling pipeline excludes sessions with personally identifiable information and sessions from users who opted out of training.
More from Research
- New paper proves fundamental confidence-efficiency bounds for transductive conformal prediction — _onionesque · 2026-09-23
- $1B and unlimited frontier tokens: where would you spend them to fix cybersecurity? — chrisrohlf · 2026-09-23
- Grok explains why DeepSeek picked DualPipe + ZeRO-1 over ZeRO-3 on 2048 H800s — TheZachMueller · 2026-09-23
- ObviousBench ranks LLMs on trivial-for-human questions by cost — GPT-Luna takes 2 crowns — adamallcock · 2026-09-23
- Framing automated science as a hypothesis-experiment-refute loop, with agent swarms competing — burny_tech · 2026-09-23
- Categorical deep learning is elegant but has yet to yield practical architectures — burny_tech · 2026-09-23