HM3D-M³ is a benchmark dealing with goal navigation scenarios in which multiple agents use multiple modalities while navigating toward multiple goals. At 04:13 UTC on October 15, 2026, the latest commit 8963b1d of nju-r-l-group-embodied-lab/hm3d-m3-code is a change that reflects in the citation information the fact that the paper of this project was uploaded as an arXiv preprint (arXiv:2610.20121).
In the link line at the top of the README, a Paper (arXiv) entry was added immediately after Project page, and the previously existing Episodes link now points as-is to the v1.0.0 tag on GitHub Releases. The BibTeX example that was in the README has had the 'Under review' notation in the note field removed and the eprint, archivePrefix, primaryClass(cs.RO), and url fields filled in. In CITATION.cff, 'arXiv preprint arXiv:2610.20121' and the corresponding abs URL were added as a journal entry. The author affiliation is listed as Nanjing University's School of Intelligence Science and Technology.
The scope of the change is limited to citation metadata. Neither the code nor the data nor the episode composition was touched, and the commit message is a single line about adding the preprint citation. Conversely, this commit alone does not tell which metrics HM3D-M³ presented, nor how it compares to other navigation benchmarks. The preprint number can be confirmed, but the repository record contains no summary of the paper body. It is better to read arXiv:2610.20121, directly check the design and the evaluation results, and then decide whether to use it.