6 accepted papers
Learning human objectives from preference feedback has significantly advanced reinforcement learning (RL) in domains where objectives are hard to formalize. However, traditional methods based on pairwise trajectory comparisons face notable challenges, including the difficulty in comparing trajector…
Defining a reward function is usually a challenging but critical task for the system designer in reinforcement learning, especially when specifying complex behaviors. Reinforcement learning from human feedback (RLHF) emerges as a promising approach to circumvent this. In RLHF, the agent typically le…
In practice, reinforcement learning (RL) agents are often trained with a possibly imperfect proxy reward function, which may lead to a human-agent alignment issue (i.e., the learned policy either converges to non-optimal performance with low cumulative rewards, or achieves high cumulative rewards bu…
Large language models (LLMs) have demonstrated exceptional capabilities across a wide range of tasks, from text generation to complex problem-solving. LLM APIs provide easy access to these models by streamlining deployment and usage. Combining LLMs with complementary strengths has been shown to yiel…
Robots that use gestures in conjunction with speech can achieve more effective and natural communication with human teammates, however, not all robots have capable and dexterous arms. Augmented Reality technology has effectively enabled deictic gestures for morphologically limited robots in prior wo
"The exponentially large discrete search space in mixed-precision quantization (MPQ) makes it hard to determine the optimal bit-width for each layer. Previous works usually resort to iterative search methods on the training set, which consume hundreds or even thousands of GPU-hours. In this study, w…