{"id":"708eb999-51fd-4180-8a10-16372532cf85","arxiv_id":"2608.01120","paper_version":1,"verdict":"CONDITIONAL","confidence":"HIGH","novelty_score":4.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":7,"one_line_summary":"2024 CHEOPS transit timings of AU Mic c show a reversal of the 2023 timing offset, indicating the large deviation was not sustained.","lead":"New CHEOPS observations from 2024 show that the large transit-timing deviation of the young exoplanet AU Mic c, seen in 2023, did not persist, with timings swinging back toward earlier values. The result matters because it changes the picture of what is causing the timing wobbles in one of the best-studied young planetary systems.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"2024 AU Mic c transit centers may be biased by flare masking and polynomial detrending; no injection-recovery test validates the negative O-C values that drive the reversal claim.","rationale":"The reader's identified weakest assumption is the accuracy of the 2024 mid-transit times after aggressive flare masking and polynomial/GP detrending. My independent reading converges on the same load-bearing point: the AU Mic c transits are shallow and grazing, precisely the regime where baseline model misspecification can shift the measured center by tens of minutes, and no injection-recovery validation is reported. The paper's cross-checks (DRP simultaneous fit, alternative GP settings, Bayes-factor decorrelation) are real and commendable, but they test robustness within the same data and do not establish absolute timing accuracy on this highly active star. A secondary concern is that the updated ephemeris is derived from the same 2023 and 2024 timings whose relationship is the headline result; using a pre-2023 ephemeris would make the O-C values less negative but still closer to zero than 2023, so this does not by itself overturn the claim. Because the paper is transparent about its limitations and the data are plausibly reliable, the appropriate verdict remains CONDITIONAL: accept the observational update contingent on an injection-recovery check and, ideally, a reanalysis of the 2023 AU Mic c transits with the same pipeline. My concern does not move the verdict; it sharpens the condition under which the central claim should be regarded as established.","tokens_in":20609,"tokens_out":11282,"duration_ms":99652,"concrete_test":"Run an injection-recovery test on the five 2024 AU Mic c visits: remove the best-fit transit model from the detrended light curves, inject synthetic transit signals with known mid-transit offsets spanning -60 to +60 min in 10-min steps, and rerun the identical pycheops detrending (including flare masking and the sixth-order polynomial) and Allesfitter fitting. Compare the recovered minus injected mid-transit offsets visit by visit. If the mean bias exceeds ~10 min or the scatter is comparable to the quoted 7-13 min uncertainties, the 2024 O-C values are not reliable at the level required for the reversal claim; if the mean bias is below ~5 min, the claim is supported.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim is that the 2024 CHEOPS O-C values of AU Mic c (Table 5: -38±13 to -8±11 min) show the large 2023 deviation (+48.6±4.0, +57.0±5.2 min) was not sustained. This conclusion depends on the absolute accuracy of the five 2024 mid-transit times, not just their internal consistency. AU Mic c is a shallow, grazing transit (depth 1353 ppm, impact parameter 0.85, full duration ~2.3 h), making its fitted center especially sensitive to baseline model assumptions. The analysis removes or masks ~43.4% of points, mostly flares, and applies a sixth-order polynomial detrending to the PIPE photometry before the transit fit in pycheops. The authors report consistency with a DRP-based simultaneous fit and with alternative GP setups, but these are perturbations of the same data and do not validate absolute timing accuracy. No end-to-end injection-recovery test is presented showing that the pipeline recovers known mid-transit times on this specific active-star light curve. If the flare-masking and polynomial detrending systematically shift the grazing transit centers by ~15-20 min, the 2024 negative O-C values and the apparent reversal could be artifacts. A secondary concern is that the updated linear ephemeris used to compute the Table 5 O-C values is fit to the same 2023 and 2024 points whose deviation is under study, so the zero point and period are not fully independent of the claimed reversal. The primary vulnerability remains the unvalidated absolute timing accuracy of the 2024 AU Mic c transits.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"This manuscript reports five new CHEOPS transit observations of AU Mic c and nine of AU Mic b obtained between June and September 2024. After masking about 43% of the data points (mostly flares), the authors detrend the PIPE photometry with a sixth-order polynomial plus a Gaussian process, fit both planets jointly with Allesfitter, update the linear ephemerides using all available transit times, and compute observed-minus-calculated (O-C) diagrams. The central claim is that the AU Mic c O-C values in 2024 (roughly -38 to -8 min) lie far below the +48.6 and +57.0 min values reported for 2023, so the large 2023 timing deviation was not sustained and the timings returned closer to the zero point, indicating a reversal of the previously reported trend. For AU Mic b the TTV signal is reported as stable (semi-amplitude 10 +/- 3 min, modulation timescale 1168 +/- 20 d). The AU Mic c TTV semi-amplitude (46 +/- 26 min) and timescale (2150 +/- 110 d) are labeled tentative, and the LiTE3 eccentricity of about 0.93 is explicitly flagged as a phenomenological fitting device. The paper is framed as a measurement paper whose primary objective is to present the 2024 timings and assess their impact on the previously reported TTV behavior.","tokens_in":20995,"tokens_out":28477,"duration_ms":228674,"significance":"If the 2024 timings are accurate, the result is significant for the interpretation of the AU Mic system: it indicates that the exceptional 2023 excursion of AU Mic c was a transient rather than the beginning of a sustained trend, with direct consequences for proposed additional perturbers (AU Mic d and e) and for the amplitude of the underlying TTV. Strengths of the paper include the cross-check with the independent DRP photometry using a simultaneously fitted baseline, robustness tests with alternative GP setups, the consistency across five separate 2024 transits, data availability, and unusually explicit caveats about the phenomenological LiTE3 model. The seasonal-offset comparison in the text (76 min between the 2023 and 2024 mean O-C levels versus the 6-15 min intra-season timing scatter) is a simple and persuasive robustness diagnostic. The principal fragilities are that the absolute accuracy of mid-transit times for a grazing, flare-contaminated transit is not validated end-to-end by an injection-recovery test, and that the reported O-C values depend on the chosen reference ephemeris at the tens-of-minutes level.","major_comments":[{"comment":"The O-C values in Table 5 and the '-38 +/- 13 to -8 +/- 11 min' span quoted in Section 3 are computed relative to a linear ephemeris fitted to all available transit times, including the same 2023 and 2024 points whose deviation is the subject of the paper. The choice of reference changes the numbers by tens of minutes: computed with the pre-anomaly Szabó et al. (2022) ephemeris for AU Mic c (Tc = 2459454.8973 BJD_TDB, P = 18.85882 d, the values used in Fig. 1 and Table 2), the 2024 residuals for visits c1-c5 become approximately +23, -1, +28, +14, and -2 min, with a mean near +12 min, rather than the -12, -36, -8, -21, and -38 min of Table 5. The qualitative conclusion that the 2023 deviation was not sustained survives either choice, but the negative sign of the 2024 offsets, the term 'reversal,' and the specific span quoted in the abstract and Section 3 are largely products of a period that has been pulled upward by the 2023 anomaly. The main text should report the 2024 residuals relative to both ephemerides, state explicitly that the updated linear ephemeris is influenced by the 2023 points, and temper the reversal wording; the cleaner statement is that the 2024 timings are consistent with the pre-2023 ephemeris.","section":"Section 2, Table 5, Section 3"},{"comment":"The central claim rests on the absolute accuracy of five mid-transit times of a grazing transit (impact parameter 0.85, depth 1353 ppm, full duration 2.3 h; Table 3), measured after masking 43.4% of the points, mostly flares, and removing a sixth-order polynomial baseline. The DRP-based cross-check, which uses aperture photometry and a simultaneously fitted baseline, and the alternative GP setups are genuine and reassuring, but they are reductions of the same 2024 data and do not by themselves certify the absolute zero point of the recovered transit centers. I request an end-to-end injection-recovery test: inject synthetic AU Mic c transits with known mid-transit times into the actual 2024 light curves, run the full PIPE/pycheops and Allesfitter chain, and report bias and scatter of the recovered centers, including the effect of asymmetric flare masking. At a minimum, quantify the sensitivity of the reported O-C values to the polynomial order (e.g., 5th and 7th) and to the exact flare-masking window. I note in fairness that the robustness margin is large: overturning the return-to-zero conclusion would require a systematic bias of roughly 40-60 min, far above the intra-season scatter, so this test bears on the exact 2024 O-C values and their interpretation rather than on the existence of the return.","section":"Section 2"},{"comment":"The reversal interpretation compares the 2024 timings with the 2023 timings of Boldog et al. (2025), which enter this paper only as adopted literature values; the 2023 light curves were not reanalyzed with the 2024 pipeline, masking, or noise model. Since the headline claim is that a previously reported deviation was 'not sustained,' the most plausible non-astrophysical alternative, namely that the 2023 or 2024 mid-transit centers are affected by epoch-dependent analysis systematics, cannot currently be excluded. The authors state that a dedicated reanalysis is beyond the scope of the present work; that is acceptable, but the abstract's 'not sustained' should then be formulated as a statement about the new observations (the 2024 points are inconsistent with an extrapolation of the 2022-2023 trend) rather than as a verdict that the 2023 measurement was an excursion. A homogeneous reanalysis of the 2023 CHEOPS visits with the present pipeline would turn the reversal claim from probable into demonstrated, and I recommend adding it if at all feasible.","section":"Section 3"}],"minor_comments":[{"comment":"The ~80 min 2023 deviation attributed to Boldog et al. (2025) and the 'up to ~60 min' quoted for the same points with the updated ephemeris are two different numbers; please state explicitly that the difference arises from the change of reference ephemeris.","section":"Introduction, Section 3"},{"comment":"The prior for Pmod,b, U(1130,1170) d, brackets the previously reported ~1150-d timescale, so the recovered 1167.6 +/- 20.3 d, and the abstract's '1168 +/- 20 d,' are strongly prior-dominated, as is the narrow Tc,3 prior U(2459470,2459480) used for planet b; please justify these priors or widen them and soften the 'consistent with previous determinations' wording accordingly.","section":"Table B.1"},{"comment":"Please give the fraction of masked points per visit; a single global 43.4% figure is hard to interpret when the visit efficiencies range from 57% to 77% (Table 1).","section":"Section 2, Table 1"},{"comment":"With e3 about 0.93 and omega3 = 143 +/- 78 deg, the LiTE3 curve for AU Mic c is effectively unconstrained; the red 'best-fitting' curve and cyan uncertainty bands in the right panel overstate the model's information content and should be labeled as an illustrative phenomenological fit.","section":"Fig. 2"},{"comment":"The comparison of the 2024 residuals under the old and new ephemerides is confined to an appendix; moving this comparison into the main text, e.g., as part of Fig. 2, would directly address the ephemeris-dependence that affects the central numbers.","section":"Fig. A.3"}],"recommendation":"major_revision","confidential_remarks":"To the editor: this manuscript is, in part, a consortium self-check: the 2023 measurements being revisited come from the same group's previous paper (Boldog et al. 2025) and the new data are from the same CHEOPS GTO program. This is not improper and the authors are appropriately candid about it, but it means the 'not sustained' claim rests partly on the same team revising its own earlier numbers; an independent or homogeneous reanalysis of the 2023 data would materially increase confidence in the result. I also want to flag that the quantitative content of Table 5 (negative 2024 O-C values) is more ephemeris-dependent than the text conveys; the authors should be asked to give the pre-2023-ephemeris residuals a prominent place in the revised version."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague,\n\nHere is my take on arXiv:2608.01120. The paper is a straightforward observational update from the CHEOPS monitoring of AU Mic, and the headline claim—that the large 2023 transit-timing deviation of AU Mic c did not persist into 2024—looks supported by the new data. The 2024 O-C values for c are all negative, spanning -38 to -8 minutes, while the 2023 values were +49 and +57 minutes. The result matters because it resets the narrative on AU Mic c's TTV amplitude, from a growing ~60-minute deviation to a more tentative, possibly few-minute wobble.\n\nWhat the paper does well: it presents new CHEOPS transit times for both b and c, not published before; it cross-checks the PIPE light curves with the DRP pipeline and with alternative GP setups; and it is openly cautious about its own model. The LiTE3 fit for c has e=0.93 and is explicitly labeled phenomenological, not a physical orbit. That is honest.\n\nSoft spots: the main vulnerability is the absolute timing accuracy of the shallow, grazing c transits. The analysis masks 43.4% of the points (mostly flares) and applies a sixth-order polynomial plus GP detrending. No end-to-end injection-recovery test is presented that shows the pipeline recovers known mid-transit times on this particular active-star light curve. The two-pipeline consistency is reassuring, but both pipelines could in principle share a systematic bias. I would not call this fatal, but a referee should ask for an injection-recovery test or a careful discussion of plausible systematic timing shifts. A second, minor point: the updated linear ephemeris used for the O-C values is fitted to the same 2023 and 2024 points whose deviations are under study. The authors show a comparison with the previous ephemeris in Figure A.3, so the reversal is not a pure artifact, but the zero point is not fully independent.\n\nThe 2023 data are not reanalyzed; that is a stated limitation, not a flaw, since the \"not sustained\" claim only requires the 2024 timings to be much closer to zero.\n\nWho is this for: people following TTVs in young systems and anyone planning follow-up of AU Mic. It deserves a serious peer review—desk rejection would be wrong. I would recommend accepting it with referee requests for an injection-recovery test and a brief discussion of the ephemeris dependence.","headline":"A solid, honest observational update: the 2024 CHEOPS timings do show AU Mic c's large 2023 TTV deviation did not persist, though the shallow c transits deserve one more validation step.","tokens_in":22104,"tokens_out":4069,"would_cite":true,"duration_ms":36264,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"2024 CHEOPS transits show AU Mic c's large 2023 timing deviation was not sustained.","keywords":["transit-timing variations","AU Mic","CHEOPS","exoplanet photometry","O-C diagram","young stellar systems","stellar activity","TTV modeling"],"falsifier":"Re-analyze the 2023 CHEOPS light curves with the same photometry and detrending used for 2024: if the 2023 observed-minus-calculated values move from about +49 and +57 minutes to near zero, the claimed reversal becomes an artifact of differing treatment. Alternatively, observe the 2025–2026 transits: if the offsets return to strongly positive values near the 2023 points, the 2024 reversal is part of a longer oscillation rather than a lasting change in behavior.","tokens_in":20407,"feed_emoji":"⏱️","tokens_out":8318,"duration_ms":66706,"temperature":0.7,"pith_summary":"The paper reports new CHEOPS space-telescope transit observations of the young planetary system AU Mic taken in June–September 2024 and asks what they do to the previously reported transit-timing signal. Its central claim is that the large timing deviation of AU Mic c seen in 2023, with observed-minus-calculated offsets near +49 and +57 minutes, was not sustained: the 2024 offsets cluster from about −38 to −8 minutes, indicating a return toward zero and a reversal of the 2022–2023 trend. This matters because transit-timing variations are the main observational route to hidden planets in this system; if the 2023 excursion was an outlier rather than a growing signal, the true TTV amplitude of AU Mic c may be much smaller than previously suggested. The companion planet AU Mic b shows a stable, well-established timing modulation with semi-amplitude 10±3 minutes and a characteristic timescale of 1168±20 days.","feed_headline":"AU Mic c's big 2023 timing shift did not last","feed_subtitle":"Fresh 2024 CHEOPS transits put the young planet back near schedule, reopening the case for hidden worlds.","key_machinery":"The central object is the observed-minus-calculated (O−C) diagram: the difference between each observed mid-transit time and the prediction of a linear ephemeris whose period and reference time are fitted to all available transits. The 2024 mid-transit times are measured from CHEOPS faint-star imagette photometry, after masking flares and outliers (43.4% of points removed), modeling the baseline with a sixth-order polynomial plus a Gaussian-process regression with a simple-harmonic-oscillator kernel, and jointly fitting both planets' transits; an independent reduction of the mission's standard pipeline data reproduces the same timing behavior. The resulting O−C series is then described phenomenologically with a light-time-effect model, whose fitted period is used only as a characteristic modulation timescale and is explicitly not interpreted as the orbit of a real third body.","core_discovery":"On the authors' own terms, the discovery is that the large 2023 offset in AU Mic c's transit-timing curve was a transient episode rather than the continuation of a steady trend. Using updated linear ephemerides fitted to all available TESS, Spitzer, and CHEOPS transit times, the 2024 CHEOPS mid-transit times yield observed-minus-calculated values for AU Mic c between −38±13 and −8±11 minutes, while the two 2023 CHEOPS visits gave +48.6±4.0 and +57.0±5.2 minutes. The authors stress that both the fitted TTV semi-amplitude (46±26 minutes) and the modulation timescale (2150±110 days) for planet c remain tentative, and they explicitly leave open whether the 2023 deviation was dynamical, activity-related, or partly a data-treatment effect; their central point is simply that the large deviation was not reproduced in 2024. For AU Mic b the signal is coherent, with a 10±3 minute semi-amplitude and 1168±20 day modulation consistent with previous work, reinforcing confidence in that planet's timing behavior.","pith_inferences":["If the 2023 excursion is attributed to stellar activity or data treatment rather than dynamics, then single-season TTV anomalies in young, active stars should generally be treated as provisional until confirmed in a second season.","The reversal is also compatible with a sinusoidal TTV whose period is longer than the observed baseline; with only a few cycles sampled, a stochastic or activity-driven process cannot yet be excluded.","An immediate, decisive test is to re-run the present pipeline on the 2023 CHEOPS visits, which the authors state is beyond their scope: if those offsets then shrink toward zero, the apparent reversal would be a methodological artifact rather than an astrophysical change.","Because AU Mic b and c have different impact parameters, their simultaneous 2024 transits probe different stellar chords; comparing their depth and timing changes can disentangle spot-crossing effects from true dynamical TTVs."],"forward_implications":["If the reversal is real, the 2023 AU Mic c offsets should not be read as evidence of a steadily growing TTV; the case for the proposed non-transiting planet AU Mic d must be revisited, and planet c's genuine TTV amplitude may be close to the few-minute level predicted by earlier work.","Continued CHEOPS and TESS monitoring across 2025–2026 should separate a long-period sinusoidal modulation (characteristic timescale about 2150 days) from a one-off 2023 anomaly.","The stable AU Mic b signal, with 10±3 minute semi-amplitude on a roughly 1168-day modulation, supports a coherent dynamical origin for that planet's TTVs and makes its future transit times predictable.","The 2024 transit depth of AU Mic b is more than 3σ smaller than in 2023, suggesting occulted active regions affect measured depths; tracking both planets' depths over time can map the stellar spot distribution.","The updated linear ephemerides for both planets sharpen predictions for future follow-up observations and for dynamical modeling of the system."],"supporting_citations":[{"why":"Provides the 2023 CHEOPS transits showing the large AU Mic c deviation and the previous TTV analysis that this paper's 2024 data test.","marker":"Boldog et al. (2025)"},{"why":"Supplies the reference mid-transit times and periods used as fixed priors, plus the earlier prediction of a small TTV amplitude for AU Mic c.","marker":"Szabó et al. (2022)"},{"why":"Proposes the non-transiting planet AU Mic d to explain the TTVs; the reversal makes this explanation uncertain.","marker":"Wittrock et al. (2023)"},{"why":"Defines the point-spread-function photometric extraction used to produce the CHEOPS imagette light curves.","marker":"Brandeker et al. (2024)"},{"why":"Provides the detrending tool and decorrelation methodology used for the 2024 light curves.","marker":"Maxted et al. (2022)"},{"why":"Describes the independent CHEOPS data-reduction pipeline whose transit timings agree with the main analysis.","marker":"Hoyer et al. (2020)"},{"why":"Discovery paper for AU Mic b and c, contributing the TESS and Spitzer transit times included in the O-C dataset.","marker":"Plavchan et al. (2020)"},{"why":"Adds TESS sector 1 and 27 transit times for both planets used in the O-C diagram.","marker":"Gilbert et al. (2022)"}],"fun_headline_variants":["AU Mic c's 2023 timing spike fades in new CHEOPS data","AU Mic c's big 2023 TTV offset was transient","Reversal: AU Mic c's 2024 transits undo the 2023 timing jump","AU Mic c's 2023 transit anomaly not repeated in 2024","New CHEOPS data show AU Mic c's timing reversal"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that the measured 2024 mid-transit times are accurate to the quoted few-minute-to-tens-of-minutes precision, meaning the masking of 43.4% of data (mostly flares) and the sixth-order polynomial plus Gaussian-process detrending do not systematically shift the derived transit centers by tens of minutes.","fun_headline_variants_meta":{"raw":{"variants":["AU Mic c's 2023 timing spike fades in new CHEOPS data","AU Mic c's big 2023 TTV offset was transient","Reversal: AU Mic c's 2024 transits undo the 2023 timing jump","AU Mic c's 2023 transit anomaly not repeated in 2024","New CHEOPS data show AU Mic c's timing reversal"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000527,"raw_usage":{"total_tokens":2563,"prompt_tokens":986,"completion_tokens":1577,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":602,"completion_tokens_details":{"reasoning_tokens":1475}},"tokens_in":602,"tokens_out":1577,"duration_ms":9089,"temperature":1.0,"reasoning_tokens":1475,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-15T15:12:21.973635+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Re-analyze the 2023 CHEOPS light curves with the same photometry and detrending used for 2024: if the 2023 observed-minus-calculated values move from about +49 and +57 minutes to near zero, the claimed reversal becomes an artifact of differing treatment. Alternatively, observe the 2025–2026 transits: if the offsets return to strongly positive values near the 2023 points, the 2024 reversal is part of a longer oscillation rather than a lasting change in behavior.","supporting_citations":[],"review_version":1}