antonia bronars
28 posts
@BronarsToni
phd candidate in robotics @ mit csail ~ giving robots a sense of touch 👈🤖👉
What's different between these two BC policies? It's the same architecture, training budget, and data collection setup — the only difference is the controller gains! Controller gains are an understudied design parameter in robot learning. In our new work (w/ @BronarsToni*, @pulkitology), we show how they act as an inductive bias across BC, RL, and Sim2Real transfer, with real consequences on performance. Here's what we found 🧵 * Equal Contribution 📄arxiv: arxiv.org/abs/2604.02523 🔗website: younghyopark.me/tune-to-learn/
[10/n] Broader implication 2 Learning from human videos or wearables is a promising direction. These paradigms, however, often treat "future state" as its pseudo action label — implicitly assuming perfect tracking, which is effectively a stiff-controller assumption. If our findings generalize, rethinking this assumption could unlock even more of their potential.
What's different between these two BC policies? It's the same architecture, training budget, and data collection setup — the only difference is the controller gains! Controller gains are an understudied design parameter in robot learning. In our new work (w/ @BronarsToni*, @pulkitology), we show how they act as an inductive bias across BC, RL, and Sim2Real transfer, with real consequences on performance. Here's what we found 🧵 * Equal Contribution 📄arxiv: arxiv.org/abs/2604.02523 🔗website: younghyopark.me/tune-to-learn/
What's different between these two BC policies? It's the same architecture, training budget, and data collection setup — the only difference is the controller gains! Controller gains are an understudied design parameter in robot learning. In our new work (w/ @BronarsToni*, @pulkitology), we show how they act as an inductive bias across BC, RL, and Sim2Real transfer, with real consequences on performance. Here's what we found 🧵 * Equal Contribution 📄arxiv: arxiv.org/abs/2604.02523 🔗website: younghyopark.me/tune-to-learn/
What's different between these two BC policies? It's the same architecture, training budget, and data collection setup — the only difference is the controller gains! Controller gains are an understudied design parameter in robot learning. In our new work (w/ @BronarsToni*, @pulkitology), we show how they act as an inductive bias across BC, RL, and Sim2Real transfer, with real consequences on performance. Here's what we found 🧵 * Equal Contribution 📄arxiv: arxiv.org/abs/2604.02523 🔗website: younghyopark.me/tune-to-learn/
What's different between these two BC policies? It's the same architecture, training budget, and data collection setup — the only difference is the controller gains! Controller gains are an understudied design parameter in robot learning. In our new work (w/ @BronarsToni*, @pulkitology), we show how they act as an inductive bias across BC, RL, and Sim2Real transfer, with real consequences on performance. Here's what we found 🧵 * Equal Contribution 📄arxiv: arxiv.org/abs/2604.02523 🔗website: younghyopark.me/tune-to-learn/
Overturning the next-token prediction is required for achieving general reasoning! We predict that RPT (Reward Pre-Training) will overtake GPT in the future -- similar to how AlphaZero overtook AlphaGo. Learn more: arxiv.org/pdf/2502.19402 🚨Our whitepaper, “General Reasoning Requires Learning to Reason from the Get-Go” challenges the idea that simply making models bigger and feeding them more data is enough for robust, adaptable reasoning. ⚡️We argue that models should be trained for iterative reasoning from scratch while separating knowledge (i.e., facts) from reasoning!
Our demo video showcasing autonomous robotic assembly has been published! We have created a closed-loop system that exhibits robustness to failure and empowers the robot to autonomously assemble a gearbox from any initial condition. youtube.com/watch?v=cZ9M1D…
Introducing TEXterity ~ our method for simultaneous tactile estimation and control for extrinsic dexterity 👉TEXterity tackles precise in-hand manipulation with simple grippers, for tasks like assembly and tool-use 🍎website (videos + paper) : sites.google.com/view/texterity 🧵👇
