Kyle Wong

468 posts

Kyle Wong banner
Kyle Wong

Kyle Wong

@ewveggies

Training computer use agents @ Amazon AGI Labs. Prev ML @Apple, @SimularAI, @ucsbNLP

San Francisco, CA Beigetreten Haziran 2024
76 Folgt558 Follower
Angehefteter Tweet
Kyle Wong
Kyle Wong@ewveggies·
Excited to share what our lab has been baking: Amazon Nova Act! Trained with large scale RL on diverse web gyms, Nova Act achieves SOTA on multiple public web agent benchmarks. Check it out!🚀 labs.amazon.science/blog/amazon-no…
English
0
1
12
10.6K
Kyle Wong
Kyle Wong@ewveggies·
@_Suresh2 Their collapse doesn’t seem reward model driven. Fig 15 shows instability on aime and lcb, and the report mentions for stem/code they only do RLVR
English
0
0
0
53
Suresh
Suresh@_Suresh2·
@ewveggies the on-policy bit usually falls apart once you scale the reward model updates past a few thousand steps
English
1
0
1
110
Kyle Wong
Kyle Wong@ewveggies·
Beautiful tech report, perhaps the best western model report I’ve ever read. Lots of great insights: no synthetic data in midtrain, teacher models are RLed directly on top of midtrain, and adaptive clip higher. But still seems like they didn’t fully nail true on-policy as they admit their RL stage is unstable, leading to a hacky self-distillation stage (imo)
elie@eliebakouch

WOW microsoft new "MAI Thinking 1" model comes with a 109 page tech report that looks REALLY detailed, this is amazing

English
5
8
112
12.3K
TechPat
TechPat@TechnologyPat·
@ewveggies Best Western is making AI now? Pump the stonk
English
1
0
2
63
Kyle Wong
Kyle Wong@ewveggies·
Idk if Claudes/GPTs use synthetic data since they don't share anything. But if you want to train the BEST model, and you are confident that your model has the 'best pre-training' (most world knowledge) and 'best mid-training' (best domain specific capabilities and behaviors), then it doesn't make sense to distill off-policy data from other models, since: 1. Those models have less world and domain knowledge 2. Lots of SFT on off-policy synthetic data pulls your model into a more narrow distribution
English
1
1
2
249
Manoj 🪐
Manoj 🪐@SaturnKit·
@ewveggies Can someone explain "zero synthetic data or distillation from previous models"... What are the advantages? Are other models like Claude /Sonnet/Gpts use synthetic? 🤔
English
1
0
2
262
Kyle Wong
Kyle Wong@ewveggies·
Gotta love SF poker: Some absolute degen pre-flop 5 bet jams with 34 offsuit, gets called by pocket queens and pocket aces. Flop comes 2 5 6, flopped the absolute nuts and won $500 pot. Sickest hand I’ve ever seen.
Kyle Wong tweet media
Corgi@UseCorgi

So much fun hosting a poker night with our friends @SignalFire, packed with founders and operators from the ecosystem. Reminder that we have a space in the heart of SF for community events like this. The whole point is creating room for people to meet and for serendipity to do its thing.

English
0
0
4
756
Kyle Wong
Kyle Wong@ewveggies·
To my grand total of 3 followers who are gonna see this, the 1000 Leetcode streak has been achieved
Kyle Wong tweet mediaKyle Wong tweet media
English
2
0
14
471
Kyle Wong
Kyle Wong@ewveggies·
hey everyone! i’m kyle - new grad 2025 @UCSB - no prev internships - no prev research - no employment ever - bottom 5th percentile in math/coding contests - 2 stars on github class projects - turned down competitve offers @McDonalds and @BurgerKing to hustle on my own - got kicked out of parents basement yesterday; disowned - staying in sf for a few days! looking to raise for my neolab hmu if interested!
Kyle Wong tweet mediaKyle Wong tweet mediaKyle Wong tweet media
Samuel Zhang@samuelxzhang

hey everyone! i'm samuel - 2nd year cs @uwaterloo - prev eng @memories_ai, ai research @uwaterloo - 99th percentile in multiple national math/coding contests - prev national level fencer; 275lbs max bench - 1.2k+ stars on github projects - turned down swe offers @Gemini @openart_ai and @ yc startups this summer to build smth of my own - got flown out for yc s26 interview yesterday; rejected - staying in sf for a few more days; looking to raise from other investors hmu if interested!

English
44
35
1.3K
98.7K
ICE
ICE@ICE257_·
Hey everyone! I’m King but you can call me ICE (fuck how do we do this) - AI/ML dev. - Heavy Codexoor - I build cool af stuff - idk check my portfolio it’s a bit more cohesive kingsleyaremu.vercel.app
Kyle Wong@ewveggies

hey everyone! i’m kyle - new grad 2025 @UCSB - no prev internships - no prev research - no employment ever - bottom 5th percentile in math/coding contests - 2 stars on github class projects - turned down competitve offers @McDonalds and @BurgerKing to hustle on my own - got kicked out of parents basement yesterday; disowned - staying in sf for a few days! looking to raise for my neolab hmu if interested!

English
1
1
5
251

Entdecken

@27upon2 @_Suresh2 @TechnologyPat @truthixifi @sanjayramesh64 @lukas_hellesch @UCSB @McDonalds