SabaPivot's picture
|
download
raw
838 Bytes
# Source audit
- Paper: https://arxiv.org/abs/2606.05919 (v1, 4 June 2026)
- OpenReview: https://openreview.net/forum?id=ghd0zmtpB9
- Released implementation: https://github.com/nk027/findingMIS at commit `12fdd4775a3df694926bfabe45421aebaa281eb1`
- Independent rerun: https://huggingface.co/jobs/SabaPivot/6a5ce3ecbee6ee1cf4ed12f3
Claim 1 is stated in Section 3.1 and Algorithm 1. Claim 2 is Theorem 1; the O(n) per-step complexity presumes expected-linear top-k selection, while the released size-k heap is O(n log k). Claim 3 is Theorem 2 and requires fixed k, positive oracle residual variance, uniform score/denominator stability, and a gap condition for exact selection. Claim 4's sub-200 ms number is specific to n=10^6,k=10^5; the median-three-updates number is aggregated across the whole grid. Claim 5 is exactly Algorithm 2.

Xet Storage Details

Size:
838 Bytes
·
Xet hash:
2aaab14f8ad1885717357f3c4dd0902997ac34ec6f5ce1151397bb5e4284a452

Xet efficiently stores files, intelligently splitting them into unique chunks and accelerating uploads and downloads. More info.