More Like this

1. Task Descriptors Help Transformers Learn Linear Models In-Context

   Huang, Ruomin; Ge, Rong (April 2025, ICLR 2025)

   Large language models (LLMs) exhibit strong in-context learning (ICL) ability, which allows the model to make predictions on new examples based on the given prompt. Recently, a line of research (Von Oswald et al., 2023; Aky¨urek et al., 2023; Ahn et al., 2023; Mahankali et al., 2023; Zhang et al., 2024) considered ICL for a simple linear regression setting and showed that the forward pass of Transformers is simulating some variants of gradient descent (GD) algorithms on the in-context examples. In practice, the input prompt usually contains a task descriptor in addition to in-context examples. We investigate how the task description helps ICL in the linear regression setting. Consider a simple setting where the task descriptor describes the mean of input in linear regression. Our results show that gradient flow converges to a global minimum for a linear Transformer. At the global minimum, the Transformer learns to use the task descriptor effectively to improve its performance. Empirically, we verify our results by showing that the weights converge to the predicted global minimum and Transformers indeed perform better with task descriptors. 

   more » « less
2. Learning To Learn and Remember Super Long Multi-Domain Task Sequence

   Wang, Zhenyi; Shen, Li; Duan, Tiehang; Zhan, Donglin; Fang, Le; Gao, Mingchen (January 2022, Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR))
3. Task Interdependence in Human-Robot Teaming

   Fangyun Zhao, Curt Henrichs (August 2020, IEEE International Conference on Robot and Human Interactive Communication)

   Human-robot teaming is becoming increasingly common within manufacturing processes. A key aspect practitioners need to decide on when developing effective processes is the level of task interdependence between human and robot team members. Task interdependence refers to the extent to which one’s behavior affects the performance of others in a team. In this work, we examine the effects of three levels of task interdependence—pooled, sequential, reciprocalin human-robot teaming on human worker’s mental states, task performance, and perceptions of the robot. Participants worked with the robot in an assembly task while their heart rate variability was being recorded. Results suggested human workers in the reciprocal interdependence level experienced less stress and perceived the robot more as a collaborator than other two levels. Task interdependence did not affect perceived safety. Our findings highlight the importance of considering task structure in human-robot teaming and inform future research on and industry practices for human-robot task allocation. 

   more » « less
4. Human-Aware Robot Task Planning Based on a Hierarchical Task Model

   https://doi.org/10.1109/LRA.2021.3056370  

   Cheng, Yujiao; Sun, Liting; Tomizuka, Masayoshi (April 2021, IEEE Robotics and Automation Letters)

   null (Ed.)
5. Human Comfort Index Estimation in Industrial Human–Robot Collaboration Task

   https://doi.org/10.1109/THMS.2025.3530530  

   Savur, Celal; Heard, Jamison; Sahin, Ferat (April 2025, IEEE Transactions on Human-Machine Systems)