NathanThinks Profile Banner
nathan lile Profile
nathan lile

@NathanThinks

Followers
2K
Following
28K
Media
248
Statuses
2K

ceo/cofounder @ https://t.co/bDd3J4Lmzf hiring in SF 🌁 scaling synthetic reasoning. recurrent rabbit hole victim. nothing great is easy.

San Francisco
Joined August 2013
Don't wanna be here? Send us removal request.
@NathanThinks
nathan lile
6 months
Superintelligence isn't about discovering new things; it's about discovering new ways to discover. I think our latest work formalizes Meta Chain-of-Thought which we believe lies on the path to ASI. When we train models on the problem-solving process itself—rather than the final.
@rm_rafailov
Rafael Rafailov @ NeurIPS
6 months
We have a new position paper on "inference time compute" and what we have been working on in the last few months! We present some theory on why it is necessary, how does it work, why we need it and what does it mean for "super" intelligence.
Tweet media one
4
28
133
@NathanThinks
nathan lile
3 days
up _and_ left 😲
Tweet media one
0
4
11
@NathanThinks
nathan lile
4 days
RT @tszzl: you have no idea how hard it is to get an rlhf model to be even ā€œcentristā€ much less right reactionary. they must have beat this….
0
181
0
@NathanThinks
nathan lile
8 days
RT @sama: I’m not big on identities, but I am extremely proud to be American. This is true every day, but especially today—I firmly believe….
0
3K
0
@NathanThinks
nathan lile
9 days
RT @reach_vb: Apple dropping diffusion based Coding LLMs on Hugging Face was not on my bingo
Tweet media one
0
85
0
@NathanThinks
nathan lile
16 days
RT @FredericLambert: Xiaomi got 200,000 orders in 3 minutes for the YU7 and I’m not even surprised. The value proposition is just nuts. I….
0
52
0
@NathanThinks
nathan lile
17 days
RT @tractoai: the future is about smart tokens.
0
2
0
@NathanThinks
nathan lile
18 days
What if models could learn which problems _deserve_ deep thinking?. No labels. Just let the model discover difficulty through its own performance during training. Instead of burning compute šŸ”„šŸ’ø on trivial problems, it allocates 5x more on problems that actually need it ↓.
@synth_labs
SynthLabs
18 days
Our new method (ALP) monitors solve rates across RL rollouts and applies inverse difficulty penalties during RL training. Result? Models learn an implicit difficulty estimator—allocating 5x more tokens to hard vs easy problems, cutting overall usage by 50%. šŸ§µšŸ‘‡1/10
Tweet media one
1
6
37
@NathanThinks
nathan lile
21 days
RT @JessePeltan: China is winning the race to Type 1 Civilization and we're not even aware it's happening. By 2030, China will have the ma….
0
856
0
@NathanThinks
nathan lile
25 days
RT @ashVaswani: Check out our latest research on data. We're releasing 24T tokens of richly labelled web data. We found it very useful for….
0
81
0
@NathanThinks
nathan lile
1 month
RT @JamesAlcorn94: congrats @rm_rafailov on your hard-earned acceptance to the USofA as alien of officially extraordinary ability. The alie….
0
2
0
@NathanThinks
nathan lile
1 month
RT @rm_rafailov: When we first published our work on this 9 months ago it was rejected for being impractical in realistic cases. Six month….
0
14
0
@NathanThinks
nathan lile
1 month
@NathanThinks
nathan lile
2 months
btw we have ongoing research on this front! we're open-science, pro-publication, and love collaboration. want to push this frontier forward? we're growing our SF team & always open to research partners—reach out, my DMs are open šŸ“©.
0
0
1
@NathanThinks
nathan lile
1 month
Generative Reward Models impact compounds daily. way stronger interest now than when we published last fall šŸ‘‡. many excellent recent extensions—cool seeing where .researchers take GenRM
Tweet media one
Tweet media two
Tweet media three
@NathanThinks
nathan lile
9 months
we bootstrapped our way to generalized meta-reasoning capabilities with generative reward models. classical reward models can be worse than random on new reasoning tasks šŸŽ². we see improvements in robustness, generalization, interpretability and an opportunity to unify RLHF/RLAIF.
1
2
19
@NathanThinks
nathan lile
1 month
RT @nathanfielder: I was going to call this dumb, but former NTSB board member John Goglia just texted me and told me to reply with this in….
0
2K
0
@NathanThinks
nathan lile
1 month
RT @ChaseBlagden: Thank you to @synth_labs and friends for making this possible!🄳
Tweet media one
0
1
0
@NathanThinks
nathan lile
2 months
RT @MetaPuppet: This is Plastic. Made with Veo3. Spoilers in the next post. Watch before reading
0
536
0
@NathanThinks
nathan lile
2 months
RT @NathanThinks: btw we have ongoing research on this front! we're open-science, pro-publication, and love collaboration. want to push th….
0
8
0
@NathanThinks
nathan lile
2 months
RT @HashemGhaili: Prompt Theory (Made with Veo 3). What if AI-generated characters refused to believe they were AI-generated? https://t.co/….
0
4K
0
@NathanThinks
nathan lile
2 months
Platonic GANs. >Repeat after me—your embeddings were never yours.
@jxmnop
jxmo
2 months
excited to finally share on arxiv what we've known for a while now:. All Embedding Models Learn The Same Thing. embeddings from different models are SO similar that we can map between them based on structure alone. without *any* paired data. feels like magic, but it's real:🧵.
0
0
4
@NathanThinks
nathan lile
2 months
btw we have ongoing research on this front! we're open-science, pro-publication, and love collaboration. want to push this frontier forward? we're growing our SF team & always open to research partners—reach out, my DMs are open šŸ“©.
@NathanThinks
nathan lile
2 months
excellent work by @jaseweston & team—extending our "Generative Reward Models" work with RL (GRPO) to optimize LLM reasoning during judgment. scalable (synthetic) evaluation continues to be AI's key bottleneck!
Tweet media one
17
8
56