Daniel Brown
@daniel-brown.bsky.social
CS assistant prof @Utah. Researches human-robot interaction, human-in-the-loop ML, AI safety and alignment. https://users.cs.utah.edu/~dsbrown/
The main idea was that for linear rewards, we can determine, via an intersection of half-spaces, the set of reward functions that make a policy optimal and that this set of rewards is defined by a small number of "non-redundant" demonstrations or comparisons.
4/8
4/8
October 10, 2025 at 4:03 PM
The main idea was that for linear rewards, we can determine, via an intersection of half-spaces, the set of reward functions that make a policy optimal and that this set of rewards is defined by a small number of "non-redundant" demonstrations or comparisons.
4/8
4/8
It was a fun paper and has some interesting nuggets, like the fact that there exist sufficient conditions under which we can verify exact and approximate AI alignment across an infinite set of deployment environments via a constant-query-complexity test.
3/8
3/8
October 10, 2025 at 4:03 PM
It was a fun paper and has some interesting nuggets, like the fact that there exist sufficient conditions under which we can verify exact and approximate AI alignment across an infinite set of deployment environments via a constant-query-complexity test.
3/8
3/8
Can you trust your reward model alignment scores?
New work presented today at the COLM Workshop on Socially Responsible Language Modelling Research led by Purbid Bambroo and in collaboration with @anamarasovic.bsky.social that probes LLM preference test sets for redundancy and inflated scores.
1/8
New work presented today at the COLM Workshop on Socially Responsible Language Modelling Research led by Purbid Bambroo and in collaboration with @anamarasovic.bsky.social that probes LLM preference test sets for redundancy and inflated scores.
1/8
October 10, 2025 at 4:03 PM
Can you trust your reward model alignment scores?
New work presented today at the COLM Workshop on Socially Responsible Language Modelling Research led by Purbid Bambroo and in collaboration with @anamarasovic.bsky.social that probes LLM preference test sets for redundancy and inflated scores.
1/8
New work presented today at the COLM Workshop on Socially Responsible Language Modelling Research led by Purbid Bambroo and in collaboration with @anamarasovic.bsky.social that probes LLM preference test sets for redundancy and inflated scores.
1/8
We apply our approach to surgically-inspired deformable tissue manipulation and find it achieves a 10% lower reliance on human interventions compared to prior work that leverages variance-based uncertainty estimates.
3/5
3/5
September 29, 2025 at 6:27 PM
We apply our approach to surgically-inspired deformable tissue manipulation and find it achieves a 10% lower reliance on human interventions compared to prior work that leverages variance-based uncertainty estimates.
3/5
3/5
Check out our new paper being presented today at #CoRL2025 on uncertainty quantification: openreview.net/forum?id=K7K....
We propose a new second-order metric for uncertainty quantification in robot learning that we call "Agreement Volatility."
1/5
We propose a new second-order metric for uncertainty quantification in robot learning that we call "Agreement Volatility."
1/5
September 29, 2025 at 6:27 PM
Check out our new paper being presented today at #CoRL2025 on uncertainty quantification: openreview.net/forum?id=K7K....
We propose a new second-order metric for uncertainty quantification in robot learning that we call "Agreement Volatility."
1/5
We propose a new second-order metric for uncertainty quantification in robot learning that we call "Agreement Volatility."
1/5
If you're in Melbourne, come check out Connor's talk in the Teleoperation and Shared Control session today!
Paper: arxiv.org/abs/2501.08389
Website: sites.google.com/view/zerosho...
This is joint work with two of my other amazing PhD students Zohre Karimi and Atharv Belsare!
3/3
Paper: arxiv.org/abs/2501.08389
Website: sites.google.com/view/zerosho...
This is joint work with two of my other amazing PhD students Zohre Karimi and Atharv Belsare!
3/3
March 3, 2025 at 8:34 PM
If you're in Melbourne, come check out Connor's talk in the Teleoperation and Shared Control session today!
Paper: arxiv.org/abs/2501.08389
Website: sites.google.com/view/zerosho...
This is joint work with two of my other amazing PhD students Zohre Karimi and Atharv Belsare!
3/3
Paper: arxiv.org/abs/2501.08389
Website: sites.google.com/view/zerosho...
This is joint work with two of my other amazing PhD students Zohre Karimi and Atharv Belsare!
3/3
We study how to enable robots to use end-effector vision to estimate zero-shot human intents in conjunction with blended control to help humans accomplish manipulation tasks like grocery shelving with unknown and dynamically changing object locations.
2/3
2/3
March 3, 2025 at 8:34 PM
We study how to enable robots to use end-effector vision to estimate zero-shot human intents in conjunction with blended control to help humans accomplish manipulation tasks like grocery shelving with unknown and dynamically changing object locations.
2/3
2/3
Shared autonomy systems have been around for a long time but most approaches require a learned or specified set of possible human goals or intents. I'm excited for my student Connor Mattson to present our work at #HRI2025 on a zero-shot, vision-only shared autonomy (VOSA) framework.
1/3
1/3
March 3, 2025 at 8:34 PM
Shared autonomy systems have been around for a long time but most approaches require a learned or specified set of possible human goals or intents. I'm excited for my student Connor Mattson to present our work at #HRI2025 on a zero-shot, vision-only shared autonomy (VOSA) framework.
1/3
1/3