Learning accurate off-policy state-action value functions (Q-functions) from offline datasets becomes harder as the task horizon grows. One-step temporal-difference (TD) learning bootstraps after every transition, so small per-step errors accumulate along the horizon, and early states can be badly misvalued even when t...
Hao-Ran Xu, Pragya Srivastava, Xin-Jie Liu et al.· 0 citations
This work introduces AnchorQ, which enables coarse-to-fine policy improvement through a structured, refinable latent action representation, and shows that anchors reduce identification complexity at the cost of support bias; particle refinement trades this bias against action-class complexity; and critic accuracy deter...
Carl Qi, Wei-Jun Dong, Xiao-Jie Wang et al.· 0 citations
While Vision-Language-Action models have enabled impressive cross-task transfer in zero-shot settings, \emph{adaptation} methods are often necessary to achieve consistent performance on novel tasks. However, even though a wealth of methods for adaptation have been proposed, it is unclear how to choose the best method b...
Maria Attarian, Yi-Fan Ruan, Claas A Voelcker et al.· 0 citations
ARMOR achieves state-of-the-art performance by improving over the previous approaches by improving over the previous approaches by up to 30% on failure detection rate and up to 100% in reasoning measured through LLM fuzzy match score, demonstrating robustness to heterogeneous supervision and open-ended reasoning beyond...
Carl Qi, Xiaojie Wang, Si-Long Yong et al.· arXiv.org· 10 citations
This work proposes scalable coordination via optimal unified transport (SCOUT), the first offline MARL framework to combine a generative foundation model with a learned value function through test-time action refinement, and proves a single-term KL bound on the joint soft-value gap that vanishes as transport converges.
MT2IE is presented, an evaluation framework in which a single multimodal large language model (MLLM) acts as an evaluator agent, iteratively generating the evaluation prompts and scoring the resulting images, showing that MT2IE's image-text consistency scores have higher correlation with human judgment than metrics pre...