[
  {
    "text": "Achieves δ-error in polylog(1/δ) steps, exponential improvement over previous results",
    "status": "unverified"
  },
  {
    "text": "Under minimal assumptions, complexity is Õ(d·polylog(1/δ)) where d is data dimension",
    "status": "unverified"
  },
  {
    "text": "First polylog(1/δ) sampler for general log-concave distributions using only gradient evaluations",
    "status": "unverified"
  }
]