Paras Stefanopoulos
@stefanopopoulos
Followers
58
Following
340
Media
4
Statuses
44
CTO @parsedlabs | disciple of history
San Francisco
Joined October 2022
We’re releasing a product that trains fast, domain-aware search models on your knowledge base. Drop in your KB and we synthesise data, then use RL with verifiable rewards to train <4B models. It trains in a couple of hours, is about an order of magnitude faster than your
1
5
13
Introducing Lumina. We've built an adaptive evaluation engine that discovers failures and evolves its own outputs, all by iterating with the customer in the loop. Proper evals can only be constructed by “touching grass”, and we think this holds incredible promise for steering
0
6
12
RGT is available in our platform right now for our customers. Havin' fun, building frontier tech, seeing downstream customers getting real value from OS models and eating Kababs 🔥 We plan on exposing more of our web-app so the public can interact with these methods as well as
🧵 We just published our work on Rationale-Guided Training (RGT), a stupidly simple method that allows you to circumvent the difficult with RL and get much better performance than plain SFT Everyone's trying to build "RL-as-a-service" for LLMs, and instead we found something way
0
2
3
The trend for $GOOGL continues. Non-Search revenue is closing in on being 50% of revenue generated.
27
278
5K
We discovered that teaching models why answers are correct, not just what to output, dramatically improves training efficiency. By making latent strategies explicit during training (e.g., "don't infer diagnoses from medications"), we achieve the same performance with 10x fewer
2
2
10
Introducing attention-based attribution: why cosine similarity is cosplay. Averaging the right transformer layers yields true attribution from attention, delivering reliable chunk-level auditability with sub-100 ms overhead and lower memory. It even works on a closed model!
2
3
9
I would like to sell my shares
I want to buy shares in LoRA-XS
0
1
2
Introducing some recent research from the team. @max_kirkby and @charles0neill show that low-rank LoRA matches full fine-tuning performance. A post on what happens when theoretical findings meet real-world production tasks.
Exciting findings from our work at @parsedlabs with @charles0neill. We demonstrate that low-rank LoRA delivers full fine-tuning quality in production. Our experiments reveal several promising relationships between training loss, evaluation metrics and dataset size (more on this
1
2
6
This experiment is kind of useless How much edge do you think an LLM has on a market? You may say 1% (it’s definitely negative) Even at 1%, you’ll need 10k+ actions and observations to draw any conclusions
Gemini just completely reversed its positions Was short everything, now long GPT5 starting to get long now too
0
0
2
one day you’ll realise that nobody was actually watching and you could have done what you wanted.
189
6K
43K
I’ve missed 1 flight in my life. Cost me ~4 hours. I save an hour each time. E[life gain] = 57 minutes Take a few minutes from E[life gain] for heightened cortisol.
0
0
0
Going to the airport late is +ev I’ve found optimal timing is 40 minutes before departure for domestics. 1h 20 internationals. - Fast tracked through every gate - Boarding is done by the time you get to the gate
1
0
2
I miss spending afternoons optimising cellular automata ideas and not caring if they’re useful to the world… This was a predator vs prey simulation White: food Black: prey Red: predator 2k16
1
0
4
nobody cares about your potential. potential is worthless. everyone has potential. what separates the greats from the wannabes is reality. finished work. shipped products. written pages. solved problems. stop talking about what you could do and go create proof of what you did.
87
2K
12K
@EdwardOThorp @FoundersPodcast TBF Zemurray is an odd one out from that list of names. I just think his story is a crazy example of the way the world bends to relentless effort over a lifetime.
0
0
1
@EdwardOThorp @FoundersPodcast Ed Thorp, Dyson, Munger are my personal blueprints. Zemurray, Chung Ju-Yung, Musk, Edison, B. Franklin, so many others... so many incredible humans have built our world. I am obsessed with the stories of who have done significant things. Life is beautiful!
1
0
2
"None of this would have happened if... I had asked myself beforehand... What do you want to happen, and... What do you think will happen?" Have been guided by this quote for a few years. Saved my ass again this morning. <3 @EdwardOThorp @FoundersPodcast
1
1
4