Reward-guided image editing at test time seeks to improve a specified reward while preserving source content and visual plausibility. Many existing approaches optimize candidates through pretrained generation processes, making repeated adjustment depend on costly large-model execution and, in some cases, backbone backp...
Yi-Miao Zhou, Ze-Jia Zhong, Jing-Ya Wang et al.· 0 citations
Adapting robot manipulation policies to new tasks and environments remains highly data-intensive, while the data needed for further improvement depends on the policy's current capabilities and failure modes. We introduce EmbodiRSI, an agentic system for recursive self-improvement (RSI) in a real-to-sim-to-real setting,...
Hao-Ran Lang, Hao-Tao Lu, Shi-Yu Sang et al.· 1 citation
This work proposes QAMM, a method that turns the critic-derived adjoint signal into supervision for MeanFlow's average velocity, and derives the adjoint MeanFlow target, specify its gradient boundaries, and train it with an offline actor-critic.
Yu Gong, Shu-Tong Ding, Mo-Kai Pan et al.· 0 citations
In-Context Forcing is introduced, a progressive autoregressive paradigm that utilizes contexts with decreasing noise levels that enables cross-frame parallel denoising, achieving substantial inference acceleration without sacrificing performance.
Lingxiao Yang, Liu Liu, Mo-Ran Li et al.· 2 citations
TactiDex is introduced, a real-world tactile-guided benchmark specifically designed to move dexterous manipulation beyond kinematic mimicry toward contact-level human-likeness and TactiSkill, a framework built upon a novel tri-component tactile reward that innovatively uses tactile signals as structured supervision.
Suting Ni, Hanbing Zhang, Zhenyu Wei et al.· 3 citations
We use cookies to run the site and, with your consent, for analytics and to show ads.
See our Cookie Policy.
