Post training research
san francisco, ca
Joined January 2025
- Tweets13
- Following29
- Followers569
- Likes101
TrainLoop retweeted
2,500+ sign-ups, 1,200+ attendees, and so much love! 💙 Our Dermalogica x Nolla pop-up took over the West Village with skin scans, skincare, and more!
Huge thanks to everyone who came out and @dermalogica for making it happen. More to come soon.
TrainLoop is proud to sign Open Weights and American AI Leadership.
As our economy transitions to the Autonomy Age, the best organizations will chart their own path, building their intelligence foundation atop open models.
aka.ms/OpenLetter
TrainLoop retweeted
We partnered with @mercor to test a simple idea:
What if knowledge-work agents were just… coding agents?
Result: +25% performance, 2x faster, cheaper, and new SOTA on APEX-Agents. @josancamon19
+48% from a single RL step, and 100k rollouts from a single policy.
@LoganGrasby wrote up his findings that off policy training with OAPL is more robust than we thought!
can we train a model in single RL step?
During recent experiments, @Logangrasby found that a single step of OAPL increased model performance from ~0 to 48% on a clinical reasoning and prediction task.
Turns out, data staleness might matter less than we think. with @pathos :
Task-specific training can be far more efficient than we realize. This work by @hasith_v explores the LoRA-trained solution space for GSM8k, finding a massive "plane" of solutions, representable in a single rank.
We trained LoRA adapters of different ranks to understand training dynamics, finding that adapters for GSM8k live in a surprisingly vast, low-rank solution space.
This hints that some model skills are easy to learn, and training is more forgiving than we think. @hasith_v 1/6 🧵
Specialized models for the most important tasks!
We post-trained MedGemma to be SoTA in visual medicine ddx, outperforming Opus 4.6, Gemini 3.1 and GPT-5.4 while running at ~1/30th the cost. @nollahealth Part 1 - improving visual reasoning 🧵1/6
TrainLoop retweeted
🧵 Labs and VC's are throwing cash at RL environments, especially for computer and browser use. Yet, with just 4 customers and over 30+ vendors, is cloning every website in the world really the path to scale? of course not.
Introducing TRACE: Trajectory Recording and Capture of Environments.