We measure filter leakage in public GW170817-v3 strain. A half-second interval around the known Livingston glitch is removed before three band filters are applied. In the 100 to 300 Hz band, its response has a maximum local RMS of 3e-23 strain from 0.05 to 2 seconds after merger with a sharp Fourier cut. A 20 Hz cosine taper reduces this response by a factor of 2727. Two other bands retain strong suppression. A filter with a 0.25-second response on each side has no overlap. None of the Livingston pulse scores is unusual at a 1 percent rank against 464 identically processed controls. Gate-width checks, a Hanford follow-up and a direct convolution audit define the scope. The removed interval also contains inspiral signal and noise. It is not a pure glitch model. This is a data-specific method check, not a new source or a test of published source parameters. Code, aggregate evidence, source hashes, a figure and external retrieval steps are included.
We search public H1 and L1 strain for short pulses before GW170814. Five filters use frequencies from 60 to 300 Hz and a 20 ms Gaussian envelope. The test searches time, frequency, and detector lags within 9.77 ms. No pulse passes the 99 percent threshold from 229 control windows. Two noise-power estimates give full-search ranks of 0.035 and 0.091. Equal-amplitude pulses with strain hrss of 5e-23 sqrt(s) per detector pass the main threshold in 45 to 57 of 57 reserved trials, depending on frequency. Direct waveform addition checks the filter response. This is a restricted null result, not a source detection or a limit on eccentric binaries. Only one pulse width and five frequencies are tested. Noise changes, hardware injection lines, detector response, and untested waveforms limit the result. The package contains code, aggregate results, figures, input hashes, and retrieval steps.
We test public H1, L1, and V1 strain for a delayed copy of GW170814. One approximate merger waveform and detector response ratios fitted to the main event define the test. Two network choices and three filter settings give control ranks from 0.193 to 0.346. No target passes the 99 percent threshold from 242 control windows. In the main filter setting, both networks recover 62 of 62 copies at half the fitted main-event amplitude in reserved noise windows. Direct waveform addition checks the injection calculation. The result is a restricted null result. No new source, planet, lens, or unexplained signal is found. Noisy fitted responses, one waveform, hardware injections, small control samples, and shared filter support limit the result. Code, aggregate scores, exact data hashes, figures, and external retrieval steps are included.
We test for a delayed copy of GW170814 with one fixed non-spinning merger template in two LIGO detectors. The search covers 4 to 12 seconds after the event. All three filter settings give scores below the empirical 99 percent noise threshold. Ranks against 246 control intervals give p values from 0.202 to 0.425. The main test detects the known merger and recovers 31 of 40 copies at half the fitted model amplitude, and 39 of 40 at three quarters. Nearby controls give similar null ranks. Finite filter support prevents direct leakage from the main merger into the search. Both files carry a continuous-wave hardware-injection flag. One approximate template, noise drift, a small background sample and restricted delays limit the result. No new source, planet or lens is found. Code, aggregate evidence, input hashes, external-data notices and retrieval steps are included.
We search public Hanford and Livingston strain for short pulses from 0.3 to 1.2 seconds after GW150914. Three tests use finite-support filters, 31 to 125 ms windows, both signs, and detector lags within 9.77 ms. Each score includes the full window and lag search. Ranks against 505 control intervals give p-values of 0.387, 0.346, and 0.494. One-second detector shifts and 60 nearby intervals give similar ranks. No delayed pulse candidate passes the 99 percent background threshold. The main filter leaves strong noise lines and does not pass the known merger at that threshold. The narrower-band filter does pass the merger. For equal-amplitude 150 Hz Gaussian pulses with a 20 ms envelope, the main test recovers 54 of 100 injections at a strain hrss of 1.6e-22 sqrt(s) per detector, and 100 of 100 at 3.2e-22. These results measure limited sensitivity. They do not exclude weak echoes or test a full echo model. The bundle includes code, aggregate results, source hashes, figures, and retrieval steps. No new source or planet is found.
We compared all 3,558 public LIGO-Virgo-KAGRA low-latency alerts in the O4a and O4b catalog windows with the final GWTC-4.1 and GWTC-5.0 catalogs. We used only public GraceDB notices and catalog data releases. For 2,500 compact-binary alerts, the sum of first-notice p_astro was 421, the sum of catalog p_astro for the same superevents was 309, and 266 went into the catalogs. For alerts with p_astro from 0.1 to 0.9 (n = 650), the alert p_astro sum was 184, the catalog sum was 98, and 74 were catalogued. Most of the difference came from SPIIR-first alerts in O4a (alert sum 75.5, catalog sum 13.3). MBTA-first alerts also had high values, and GstLAL and PyCBC had small differences. A logistic recalibration with pipeline terms, fitted on O4a only, decreased the O4b Brier score from 0.0344 to 0.0289 (95% CI of the difference -0.0091 to -0.0019). None of the 20 retracted alerts is in a catalog, and 9 of them had p_astro above 0.9. Relative to the catalog parameter-estimation sky maps, first-notice BAYESTAR 90% regions contained 0.858 (0.840-0.875) of the catalog probability and the last-notice online-PE regions contained 0.828 (0.818-0.838), although the online-PE regions were 26% smaller. Fourteen of 189 first-notice distances were too small by more than 3 standard deviations (all GstLAL-first heavy binary black holes; post-hoc). Thirty of 301 catalog candidates had no public alert. Limitations: the catalog p_astro and posteriors are not ground truth, alerts and catalog use the same data, and sky tests use only the 190 catalogued events with released PE sky maps.