🛠️ Steven Gong

Search

SearchSearch
Jul 16, 2025 · 1 min read · 10 views · 1 in the last 30 days

Where readers are reading from

FijiTanzaniaWestern SaharaCanadaUnited States of AmericaKazakhstanUzbekistanPapua New GuineaIndonesiaArgentinaChileDemocratic Republic of the CongoSomaliaKenyaSudanChadHaitiDominican RepublicRussiaThe BahamasFalkland IslandsNorwayGreenlandFrench Southern and Antarctic LandsEast TimorSouth AfricaLesothoMexicoUruguayBrazilBoliviaPeruColombiaPanamaCosta RicaNicaraguaHondurasEl SalvadorGuatemalaBelizeVenezuelaGuyanaSurinameFranceEcuadorPuerto RicoJamaicaCubaZimbabweBotswanaNamibiaSenegalMaliMauritaniaBeninNigerNigeriaCameroonTogoGhanaIvory CoastGuineaGuinea-BissauLiberiaSierra LeoneBurkina FasoCentral African RepublicRepublic of the CongoGabonEquatorial GuineaZambiaMalawiMozambiqueEswatiniAngolaBurundiIsraelLebanonMadagascarPalestineThe GambiaTunisiaAlgeriaJordanUnited Arab EmiratesQatarKuwaitIraqOmanVanuatuCambodiaThailandLaosMyanmarVietnamNorth KoreaSouth KoreaMongoliaIndiaBangladeshBhutanNepalPakistan · 0 readersAfghanistanTajikistanKyrgyzstanTurkmenistanIranSyriaArmeniaSwedenBelarusUkrainePolandAustriaHungaryMoldovaRomaniaLithuaniaLatviaEstoniaGermanyBulgariaGreeceTurkeyAlbaniaCroatiaSwitzerlandLuxembourgBelgiumNetherlandsPortugalSpainIrelandNew CaledoniaSolomon IslandsNew ZealandAustraliaSri LankaPeople's Republic of China · 0 readersTaiwanItalyDenmarkUnited KingdomIcelandAzerbaijanGeorgiaPhilippinesMalaysiaBruneiSloveniaFinlandSlovakiaCzech RepublicEritreaJapanParaguayYemenSaudi ArabiaTurkish Republic of Northern CyprusCyprusMoroccoEgyptLibyaEthiopiaDjiboutiSomalilandUgandaRwandaBosnia and HerzegovinaNorth MacedoniaSerbiaMontenegroKosovoTrinidad and TobagoSouth SudanSelect a country to explore · Last 30 days
Singapore0
China0
Pakistan0
Traffic · last 30 days000Oct 6Oct 16Oct 26Nov 4
views
Traffic · all time000Jul 16Jul 19Jul 22Jul 25Jul 27
views

Policy

Categorical Policy

Learned this from spinning up https://spinningup.openai.com/en/latest/spinningup/rl_intro.html.

A categorical policy is like a classifier over discrete actions.

  • It’s essentially the same ideas that used to train a next-token autoregressive model in the LLM world

Graph View

Backlinks

  • Policy

Created with Quartz, © 2026

  • Blog
  • LinkedIn
  • Twitter
  • GitHub