@pedma7

Thread

ok so for the past week ive been working making feature research as streamlined as possible. the end goal would be just having to design the feature + transformations and ship it through the testing pipeline until it hits the portfolio. i already had a defined method for feature construction, but...

@Rashikathalu

Thread

Pelli chupulu Part 3 Scene 3 Pooja-keerty(akka chellilla puvvu guvvata) Pragati vallu vachaka pooja and keerty okasare lecharu memu matladali ani alage valla mogullu kuda Pooja-mundu memu Keerty mogudu-haa Avunu Keery-ledu memu Pooja mogudu-haa Avunu ...

@moasayev

Thread

Nöbetçi Finans #25 22 Haziran- 26 Haziran 2026 Merkez Bankaları Özet ; Kısa Vadede Altın , El Nino , ABD Tahvil 22 Haziran Pazartesi Hafta Asya’dan gelen faiz ve kredi verileriyle başlayacak. Çin’de: 1 yıllık ve 5 yıllık Loan Prime Rate (LPR) oranları açıklanacak. Bu oranlar, Çin’de kredi ko...

@Rashikathalu

Thread

Part 3 Pelli chupulu Scene 2 Pragati and ramya Mogudu kacheri. Ramya and pragati Mogudu kindaki vachi ala anaka. Pragati and ramya Mogudu lecharu. PRAGATI-nenu kuda ma anayagarito knchm matladali. Ramya mogudu-avunu nenu kuda ma chelli to matladali chala important ani kanu kotadu ...

@Rashikathalu

Thread

Andariki hii Welcome to Rankula malem Part 3 Pelli chuupulu Andaruu Ramya Krishna intlo kalusukunnaru Ramya-Randi vadina garu randi annayyagaru Pragati-ha ba chesaru erpatlu pelli chupulake ...

@Rashikathalu

Thread

Part1 support chesina vala andariki chala tnq. Today part 2 story is Family2 Intro:Same ade oka nrml roju vizag lo oka nrml intlo oka ratri srelela padukunte tanaki gatiga sounds vinapadday bedroom lonundi kindaki vachi chusindi ...

@Rashikathalu

Thread

Tnx for waiting inka late cheyakunda part 1 loki velipodam part 1:- family 1 intro. Oka nrml roju hyd lo nrml intlo oka ratri ramya krishna eppatlane panulanni muginchukuni mancham meeda padukundi nagnam ga Kinda tana pukuni okadu tega nakestunte tanu pichi pichi ga mulugutondi <a target="_blank" ...

@MatthieuWyart

Thread

LLMs learn by predicting tokens. World models (JEPA, data2vec) learn by predicting their own abstractions. Which needs more data? For data with hidden hierarchy, we prove the gap is exponential. <a target="_blank" href="https://arxiv.org/pdf/2605.27734" color="blue">arxiv.org/pdf/2605.27734</a> ...

@alesfav

Thread

AI needs vastly more data than we do. One idea might close the gap: don't predict raw signals (tokens), predict your own abstract latent representation (JEPA, data2vec). With @DanKorchinski @MatthieuWyart, on a toy model, we prove how much that helps: the gap is exponential. 🧵 ...