{"id":"8fa3af4f-b60d-46a1-97b4-39d3d89e26e3","arxiv_id":"2608.06368","paper_version":1,"verdict":"ACCEPT","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"low","formal_verification":"none","parameter_count":0,"one_line_summary":"Mean first-passage time sensitivities of any finite Markov chain lie in [-1,1] and sum to -1, a conserved control budget that caps kinetic-proofreading discrimination at the number of checkpoints.","lead":"This paper proves a universal rule: for any finite Markov chain, the logarithmic sensitivity of the mean first-passage time to any single rate is bounded by one in magnitude, and all these sensitivities add up to minus one. This conserved control budget tells researchers how much speeding up one reaction step can change overall completion time, and why high kinetic-proofreading discrimination comes with fragile concentration dependence.","discovery_kind":"first_principles","skeptic_critique":{"model":"deepseek-v4-flash","headline":"No significant objection identified: the proof hinges on the imported channel-flux response identity (Eq. 5), but its extension to the redirected absorbing chain is sound.","rationale":"The reader's ACCEPT is justified. The summation rule is an exact Euler consequence of tau(lambda k) = lambda^{-1} tau(k), and the local unit bound reduces to Eq. (5). The triangle and Kac inequalities used after Eq. (5) are elementary and correctly applied, including the self-loop case that is easy to get wrong. The relevant-state irreducibility of the redirected chain holds once 'retained class' is read as the absorbing class containing A, which is the only class relevant to tau_{A to B}. The proofreading application follows from the closed forms in Appendix C and the budget accounting; Eq. (16) is a direct consequence of B_- = 1 + B_+ and the per-edge cap. No fitting, post hoc exclusion, or circular reasoning is present. The absence of a proof of Eq. (5) in the text lowers confidence slightly but does not change the verdict; the stated test would remove even that residual concern.","tokens_in":12626,"tokens_out":34186,"duration_ms":404677,"concrete_test":"Independently re-derive Eq. (5) from the generator by differentiating the stationary distribution: d log pi_x / d log k_e = j_e (tau_{n_e to x} - tau_{tilde m_e to x}), add the Kronecker-delta term, and verify against exact linear algebra on random 5- to 8-state networks that include a direct A-to-B self-loop and multiple absorption channels; confirm that the resulting s_e matches Eq. (9) and the direct MFPT derivative to machine precision.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The single step that could sink Eq. (3) is Eq. (5), the channel-flux response identity imported from the authors' prior work. The rest of the unit-bound proof, triangle squeezing, the Kac/traffic inequality, and convex averaging, is written out and internally consistent. I checked the two places where the imported identity could fail in this setting. First, for a perturbed absorption channel e = alpha, the Kronecker-delta term plus - j_alpha tau_{tilde m_alpha -> n_alpha} lies in [0,1] because the Kac inequality bounds j_alpha tau_{A -> n_alpha} by one. Second, a direct A to B edge becomes a self-loop in the redirected network; the Kac decomposition of its firing process still holds, and the apparent counterexample J_red = 1/tau resolves once the time spent away from A between completion events is included. I found no sign error, no missing term, and no circularity. The residual gap is purely that Eq. (5) is asserted rather than proved in this manuscript; that is a completeness issue for a standalone proof, not a demonstrated error in the central claim.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper establishes two universal laws for the logarithmic sensitivity s_e = ∂ log τ_{A→B}/∂ log k_e of the mean first-passage time in a finite Markov chain with absorbing target B: the single-edge bound |s_e| ≤ 1 and the summation rule Σ_e s_e = -1 (Eq. (3)). The proof maps the absorbing chain to an irreducible chain by redirecting absorption edges to the source, so the reciprocal MFPT becomes a stationary current; the channel-flux response identity (Eq. (5)) from the authors' prior work, together with triangle and Kac inequalities, bounds each channel response in [-1,1], and convex averaging yields the local bound. The paper then interprets the two laws as a conserved control budget (B_- = 1 + B_+), derives consequences for barrier and state perturbations, and applies the budget to kinetic proofreading, finding a discrimination cap Δ ≤ m and a concentration-sensitivity tradeoff |s_a| ≥ max(0, 1 + Δ - m). Appendices prove the channel-average identity and a same-source group bound, and a final appendix documents exact linear-algebra numerical methods.","tokens_in":12774,"tokens_out":8885,"duration_ms":89289,"significance":"If correct, the result is a remarkably general constraint on kinetic control: no individual rate can move the MFPT by more than its own fractional change, and uniform rescaling always leaves one unit of net speeding budget after cancellations. The proof combines a clean redirection argument with exact linear-algebra checks and random-network numerics, and the proofreading application gives a falsifiable bound (D ≤ f^m) and a sharp concentration-robustness tradeoff. The paper ships reproducible numerical procedures and closed-form formulas, and the central derivation is mostly self-contained. The main caveat is that the load-bearing response identity Eq. (5) is quoted from an overlapping-author preprint rather than proved here; this is a completeness issue for a standalone publication, not a demonstrated error in the central claim.","major_comments":[{"comment":"The unit bound -1 ≤ s_e ≤ 1 rests entirely on the channel-flux response identity (5), which is imported from Ref. [9], an arXiv preprint by the same authors, without proof or a precise statement of hypotheses. Since this is the only externally loaded step and the central claim collapses if the identity fails, the manuscript should either prove Eq. (5) in an appendix or quote a published version with a full derivation; a citation to an unreviewed preprint is not sufficient for a load-bearing identity in a standalone paper. This is a completeness issue rather than an observed error, so I expect it to be fixable by adding a self-contained proof.","section":"Section II, Eq. (5)"}],"minor_comments":[{"comment":"The word 'timming' should be 'timing'.","section":"Introduction, first paragraph"},{"comment":"The notation 'N± count the edges with s_e ≷ 0' is ambiguous; please define N_+ and N_- explicitly as the numbers of edges with positive and negative sensitivity.","section":"Section III, Eq. (8)"},{"comment":"The sentence 'a firing leaves the process at the destination and it must return to the source before firing again' is imprecise, because the inter-firing interval includes both the return and the subsequent wait for the edge to fire. Please state the decomposition explicitly as j_e(τ_{n_e→m_e} + τ_{m_e→n_e}) ≤ 1, with the strong-Markov-property argument spelled out.","section":"Section II, traffic-MFPT inequality"},{"comment":"The orientation of the generator (column versus row convention) should be declared at the start of the appendix; as written, the Kac equation ∑_z W_{zy} τ̃_{z→y} = 1/π̃_y - 1 can be confusing without this convention.","section":"Appendix A"},{"comment":"When deriving |s_a| ≥ max(0, 1 + Δ - m), state explicitly that this uses the unit bound on each forward-edge sensitivity |s_fi| ≤ 1, so the remaining budget must be carried by the binding edge.","section":"Section V, Eq. (16)"}],"recommendation":"major_revision","confidential_remarks":"The paper's central theorem is significant and the derivation is sound apart from the imported identity Eq. (5). I recommend asking the authors to provide a self-contained proof of Eq. (5) or a published reference before acceptance; this is a completeness issue arising from reliance on the authors' own unreviewed preprint, not a circularity. The manuscript is well within the scope of the journal and the numerical and analytical checks are convincing."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague,\n\nHere's the take: this paper has a genuinely new and useful result—the per-edge unit bound on MFPT sensitivities—and the proof is sound. The summation rule sum s_e = -1 is just Euler homogeneity, as they admit. The unit bound is the real thing. They get it by redirecting absorption edges to the source, turning first passage into a stationary current, then applying a response identity from their earlier preprint (arXiv:2412.19602) and squeezing hitting-time differences with triangle and Kac inequalities. I walked through the two places where that imported identity could fail in this setting—the Kronecker-delta channel case and a direct A->B edge becoming a self-loop—and both hold. The appendices fill in the channel-average identity and the same-source bound; the machinery is coherent.\n\nWhat's new: the local bound, the budget accounting B_- = 1 + B_+, and the proofreading consequences Delta <= m and the concentration-sensitivity tradeoff |s_a| >= max(0, 1+Delta-m). These are real additions; metabolic control analysis has summation theorems but no per-edge scale. The proofreading application is a direct use of the budget, not a forced one.\n\nSoft spots: the load-bearing identity Eq (5) is quoted, not proved. The stress-test says the extension is sound, and I agree, but a referee should ask for the proof to be included or at least sketched in an appendix. As written, a skeptical reader must chase the self-citation. That's a completeness matter, not a fatal one. Also, the numerics are exact linear algebra with clear provenance, but no code is shipped; that's acceptable for a theory paper, though it would be nice.\n\nThe paper is clearly written, the citations are appropriate, and nothing looks post-hoc. I'd send it to a serious referee. If I were refereeing, I'd recommend acceptance after the response identity is made self-contained. I'll cite this in my own work on first-passage control.","headline":"A clean, genuinely new unit bound on MFPT sensitivities, with a proof that holds up once you supply the missing proof of the imported response identity.","tokens_in":13346,"tokens_out":4079,"would_cite":true,"duration_ms":47112,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["60J27","60J75"],"pacs":[],"model":"deepseek-v4-flash","headline":"A universal control budget for first-passage kinetics: each rate sensitivity lies in $[-1,1]$ and the sensitivities sum to $-1$.","keywords":["first-passage time","Markov chain","logarithmic sensitivity","control budget","kinetic proofreading","rate-limiting step","recycled current","response theory"],"falsifier":"Solve the linear equations $Q\\nu=-\\delta_A$ for a candidate absorbing network, compute each $s_e=(k_e/\\tau)\\mathbf{1}^\\top \\partial\\nu/\\partial k_e$ by exact linear algebra, and search for an edge with $|s_e|>1$; one such edge in any finite chain with positive rates would falsify the unit bound. A cheaper check is the minimal chain $A\\to M\\to B$ with backward rate $k_w$: the closed form $\\tau=(k_a+k_w+k_b)/(k_a k_b)$ implies $s_w\\to +1$ and $s_a+s_b\\to -2$ as $k_w\\to\\infty$, so the predicted approach to those limits is directly testable by arithmetic.","tokens_in":12373,"feed_emoji":"⏱️","tokens_out":10063,"duration_ms":109459,"temperature":0.7,"pith_summary":"This paper claims that the mean first-passage time of any finite Markov chain with positive rates and an absorbing target obeys two universal sensitivity laws: each edge rate's logarithmic derivative $s_e = \\partial \\log \\tau / \\partial \\log k_e$ is bounded by $|s_e| \\le 1$, and the sum over all edges is $\\sum_e s_e = -1$. The two laws together act as a conserved control budget: speeding up some transitions must be paid for by delaying others, with exactly one extra unit of speeding sensitivity left after cancellation. The authors prove the local bound by redirecting every absorption event back to the source, turning first passage into a stationary recycled current and bringing steady-state response inequalities to bear on it. They then use the budget to cap kinetic-proofreading discrimination at the number of checkpoints and to show that near-maximal discrimination forces near-total sensitivity to substrate concentration. If correct, the result is a universal constraint on any experiment or design question that asks how much a completion time can be moved by changing rates.","feed_headline":"No single rate shifts completion time by more than its own change","feed_subtitle":"Every rate's grip on a reaction's mean time is capped at one, and sensitivities sum to -1.","key_machinery":"The central object is the redirection map: delete the absorbing target $B$ and send each absorption edge back to the source $A$ at the same rate, creating an irreducible network whose stationary absorption current equals the reciprocal MFPT, $J^{\\rm red}_{A\\to B}=1/\\tau_{A\\to B}$. This converts a transient first-passage quantity into a steady-state current, so the response identity for Markov jump processes, Eq. (5), can be applied to each channel flux. The unit bound for a single rate follows from two inequalities on the redirected graph: the MFPT triangle inequality squeezes the hitting-time difference between the two legs of the perturbed edge, and the traffic–MFPT (Kac) inequality $\\tilde j_e(\\tilde\\tau_{n_e\\to\\tilde m_e}+\\tilde\\tau_{\\tilde m_e\\to n_e})\\le 1$ bounds that commute by the edge traffic. For the source/edge bookkeeping, Eq. (9), $s_e=\\tilde j_e(\\tau_{m_e\\to B}-\\tau_{n_e\\to B})$, re-expresses each sensitivity as expected traversals per completion times remaining wait removed, from which barrier, state, and group bounds follow.","core_discovery":"On the paper's own terms, the discovery is a pair of exact constraints on how the mean first-passage time $\\tau_{A\\to B}$ responds to microscopic rate changes in any finite Markov chain with a source $A$, an absorbing target $B$, and positive rates. For each directed edge $e$, the logarithmic sensitivity $s_e$ satisfies $-1 \\le s_e \\le 1$, and the full set satisfies $\\sum_e s_e = -1$. The summation rule follows from Euler's theorem because $\\tau$ is homogeneous of degree $-1$ in the rates under a uniform clock rescaling; the unit bound is obtained by redirecting all absorption edges back to $A$, so the reciprocal MFPT becomes the stationary current of a recycled process, and then applying a channel-flux response identity together with triangle and Kac-type inequalities to show each channel flux responds within $[-1,1]$. The paper further derives structured consequences: a two-way barrier perturbation has sensitivity bounded by $1-\\rho_{uv}$, where $\\rho_{uv}$ is the traversal ratio of the edge; a state-energy perturbation has sensitivity exactly $-\\tilde\\pi_v$, the stationary occupancy share, and hence partitions the unit budget across states; and in kinetic proofreading the discrimination gain equals the delaying budget, so it is capped by the number of checkpoints $m$, with the binding step forced to absorb the remaining speeding budget.","pith_inferences":["Editorial inference: the summation rule should extend to any observable homogeneous in the rates, such as higher first-passage moments or splitting probabilities, but the unit bound likely does not; computing sensitivities of the variance on the minimal chain would test whether the local law is special to the mean.","Editorial inference: the budget gives a parameter-free diagnostic for kinetic assays: a measured sensitivity near $+1$ on a reset edge signals that the remaining speeding budget is nearly exhausted, so a fitted model must show the compensating $-2$ across productive edges; checking this on published kinetic schemes for motors or channel gating is a direct test.","Editorial inference: the proofreading result suggests a general tradeoff for any proofreading cascade, namely that each checkpoint buys at most one power of selectivity and the price in concentration sensitivity grows as the cap is approached; a testable extension would be to measure the discrimination gain and binding-edge sensitivity in a reconstituted proofreading system and compare with Eq. (1"],"forward_implications":["For any single rate perturbation, the mean completion time can move by at most the fractional change of that rate; no one microscopic step is ever more than fully rate-limiting.","Uniformly rescaling all rates leaves $\\sum_e s_e=-1$, so any speeding sensitivity in excess of the first unit must be paid for by an equal delaying budget on other edges.","A barrier that scales both directions of an edge controls completion only through the net current: when forward and backward traffic balance, the barrier's sensitivity vanishes.","A state-energy shift has sensitivity equal to minus that state's stationary occupancy share, so the completion time is controlled by a partition of one across states.","Kinetic proofreading can discriminate by at most the number of checkpoints, and operating near that cap forces the binding step to carry nearly unit concentration sensitivity."],"supporting_citations":[{"why":"Supplies the channel-flux response identity, Eq. (5), on which the unit bound rests.","marker":"[9]"},{"why":"Introduces the redirection of absorption events to the source and the recycled-current representation $J=1/\\tau$.","marker":"[4]"},{"why":"Justifies deleting outgoing edges of the target so that $B$ may be treated as absorbing without changing first-passage statistics.","marker":"[33]"},{"why":"Provides the renewal-theory Kac recurrence relation used in the traffic–MFPT inequality.","marker":"[34]"},{"why":"Provides the Palm inversion formula for stationary point processes used in the Kac identities.","marker":"[35]"},{"why":"Defines the kinetic proofreading model whose reset checkpoints are the object of the discrimination cap.","marker":"[29]"},{"why":"Introduces the amplification mechanism that turns repeated reset chances into selectivity.","marker":"[30]"},{"why":"Establishes the energetic-discrimination limit in which only reset rates depend on substrate identity.","marker":"[31]"},{"why":"Supplies the discrimination-gain quantity $\\Delta$ and the speed-dissipation-error framing used to connect the budget to selectivity.","marker":"[41]"}],"fun_headline_variants":["First-passage kinetics obey a universal sensitivity budget","Sum of rate sensitivities is -1: a conserved control budget","Every rate's grip on completion time is bounded by one","Barrier shifts can't beat single-rate control in kinetics","Kinetic proofreading discrimination capped by checkpoint count"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The proof of the unit bound depends on a formula, taken from the authors' earlier paper and not proved here, describing how each recycled current changes when a rate changes; if that formula does not hold after absorption channels are redirected, the bound $|s_e|\\le1$ fails.","fun_headline_variants_meta":{"raw":{"variants":["First-passage kinetics obey a universal sensitivity budget","Sum of rate sensitivities is -1: a conserved control budget","Every rate's grip on completion time is bounded by one","Barrier shifts can't beat single-rate control in kinetics","Kinetic proofreading discrimination capped by checkpoint count"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000217,"raw_usage":{"total_tokens":1439,"prompt_tokens":952,"completion_tokens":487,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":568,"completion_tokens_details":{"reasoning_tokens":409}},"tokens_in":568,"tokens_out":487,"duration_ms":6290,"temperature":1.0,"reasoning_tokens":409,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-07T04:13:50.758038+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Solve the linear equations $Q\\nu=-\\delta_A$ for a candidate absorbing network, compute each $s_e=(k_e/\\tau)\\mathbf{1}^\\top \\partial\\nu/\\partial k_e$ by exact linear algebra, and search for an edge with $|s_e|>1$; one such edge in any finite chain with positive rates would falsify the unit bound. A cheaper check is the minimal chain $A\\to M\\to B$ with backward rate $k_w$: the closed form $\\tau=(k_a+k_w+k_b)/(k_a k_b)$ implies $s_w\\to +1$ and $s_a+s_b\\to -2$ as $k_w\\to\\infty$, so the predicted approach to those limits is directly testable by arithmetic.","supporting_citations":[{"cited_title":"Derivation of the First Passage Time Distribution for Markovian Process on Discrete Network","cited_arxiv_id":"2110.02216","evidence_quote":"Justifies deleting outgoing edges of the target so that $B$ may be treated as absorbing without changing first-passage statistics."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides the renewal-theory Kac recurrence relation used in the traffic–MFPT inequality."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides the Palm inversion formula for stationary point processes used in the Kac identities."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Defines the kinetic proofreading model whose reset checkpoints are the object of the discrimination cap."},{"cited_title":"Rao and L","cited_arxiv_id":null,"evidence_quote":"Establishes the energetic-discrimination limit in which only reset rates depend on substrate identity."},{"cited_title":"Murugan, D","cited_arxiv_id":null,"evidence_quote":"Supplies the discrimination-gain quantity $\\Delta$ and the speed-dissipation-error framing used to connect the budget to selectivity."}],"review_version":1}