ReActor: Reinforcement Learning for Physics-Aware Motion Retargeting

Posted by klaurent Cornell

Kavli Affiliate: David Muller | Summary:Retargeting human kinematic reference motion onto a robot’s morphology remains a formidable challenge. Existing methods often produce physical inconsistencies, such as foot sliding, self-collisions, or dynamically infeasible motions, which hinder downstream imitation learning. We propose a bilevel optimization framework that jointly adapts reference motions to a robot’s morphology while training […]


Continue.. ReActor: Reinforcement Learning for Physics-Aware Motion Retargeting

Long-term maintenance of H3K27me3 in postmitotic neurons is dispensable for gene expression regulation

Posted by dbos Rockefeller

Kavli Affiliate: C. David Allis and Mary E. Hatten | Authors: Irma Laas, Matthew R Paul, Natarajan Bhanu, Lijuan Feng, Eve-Ellen Govek, Benjamin A Garcia, Thomas S Carroll, C David Allis, Mary E. Hatten and Kärt Mätlik | Summary: Neuronal maturation is associated with extensive changes in gene expression and chromatin organization. However, the molecular […]


Continue.. Long-term maintenance of H3K27me3 in postmitotic neurons is dispensable for gene expression regulation

Real-time AI integration for MR to detect artifacts and guide pulse sequence adaptations

Posted by dbos Hopkins

Kavli Affiliate: Jeremias Sulam/p> | Authors: Aaron T. Gudmundson, Zahra Shams, Abdelrahman Gad, Shuyuan Wang, Dunja Simicic, Saipavitra Murali-Manohar, Gizeaddis Lamesgin Simegn, Ipek Özdemir, Christopher W. Davies-Jenkins, Vivek Yedavalli, Georg Oeltzschner, Omer Burak Demirel, Jeremias Sulam, Michael schär, Sandeep Ganji and Richard A. E. Edden | Summary: Purpose To present a first-of-its-kind artificial intelligence (AI-)integrated […]


Continue.. Real-time AI integration for MR to detect artifacts and guide pulse sequence adaptations

ROSE: Rollout On Serving GPUs via Cooperative Elasticity for Agentic RL

Posted by klaurent Caltech

Kavli Affiliate: Wei Gao| Summary:Agentic reinforcement learning (RL) is reshaping LLM post-training, but end-to-end training time is dominated by compute-intensive, multi-turn rollouts whose resource demand varies significantly across training steps. Resource-fixed systems cannot adapt to this variation, while resource-elastic approaches that provision external GPUs on demand suffer from high allocation overhead and limited availability. We […]


Continue.. ROSE: Rollout On Serving GPUs via Cooperative Elasticity for Agentic RL

ROSE: Rollout On Serving GPUs via Cooperative Elasticity for Agentic RL

Posted by klaurent Caltech

Kavli Affiliate: Wei Gao| Summary:Agentic reinforcement learning (RL) has emerged as a key driver for improving the multi-step reasoning and tool-use capabilities of LLMs. However, its efficiency is bottlenecked by long-tail rollouts with multi-turn environment interactions, making static GPU provisioning a poor fit: overprovisioning wastes GPUs on stragglers, while underprovisioning increases contention and slows training. […]


Continue.. ROSE: Rollout On Serving GPUs via Cooperative Elasticity for Agentic RL

Pair-Breaking and Dimensionality in Spin-Orbit Coupled Superconductors

Posted by klaurent Caltech

Kavli Affiliate: Joseph Falson | Summary:The response of ultra-thin superconducting materials under parallel magnetic fields is often leveraged to obtain insight into the nature of the condensate, including features attributable to unconventional forms of pairing. Despite there being multiple competing mechanisms responsible for suppressing superconductivity, it is common for these analyses to overlook certain depairing […]


Continue.. Pair-Breaking and Dimensionality in Spin-Orbit Coupled Superconductors

FedFrozen: Two-Stage Federated Optimization via Attention Kernel Freezing

Posted by klaurent Peking

Kavli Affiliate: Feng Long| Summary: Federated learning with heterogeneous clients remains a significant challenge for deep learning, primarily due to client drift arising from inconsistent local updates. Existing federated optimization methods typically address this issue through objective-level regularization or update-correction mechanisms. Recent studies, however, suggest that Transformer-based architectures may be inherently more robust than conventional […]


Continue.. FedFrozen: Two-Stage Federated Optimization via Attention Kernel Freezing

ResiHP: Taming LLM Training Failures with Dynamic Hybrid Parallelism

Posted by klaurent Caltech

Kavli Affiliate: Wei Gao| Summary:Hybrid parallelism underpins large-scale LLM training across tens of thousands of GPUs. At such scale, hardware failures on individual devices lead to performance skew across devices, diminishing overall training efficiency. Existing resilient systems overlook sequence length variability in datasets and device performance skew under hybrid parallelism. As a result, (1) iteration […]


Continue.. ResiHP: Taming LLM Training Failures with Dynamic Hybrid Parallelism

ResiHP: Taming LLM Training Failures with Dynamic Hybrid

Posted by klaurent Caltech

Kavli Affiliate: Wei Gao| Summary:Hybrid parallelism underpins large-scale LLM training across tens of thousands of GPUs. At such scale, hardware failures on individual devices lead to performance skew across devices, diminishing overall training efficiency. Existing resilient systems overlook sequence length variability in datasets and device performance skew under hybrid parallelism. As a result, (1) iteration […]


Continue.. ResiHP: Taming LLM Training Failures with Dynamic Hybrid

Coherence limitations of a Fourier-engineered $cos(2varphi)$ transmon qubit

Posted by klaurent Delft

Kavli Affiliate: Christian Andersen | Summary: Intrinsically protected superconducting qubits are a promising route toward enhancing coherence times and advancing hardware towards applications in quantum computing. The $cos(2varphi)$ qubit achieves protection against qubit relaxation by allowing only the coherent tunneling of pairs of Cooper pairs, resulting in Cooper-pair parity symmetry and thereby suppressing charge-induced errors. […]


Continue.. Coherence limitations of a Fourier-engineered $cos(2varphi)$ transmon qubit