HalLoc: Token-level Localization of Hallucinations for Vision Language Models | IEEE Conference Publication | IEEE Xplore