{"id":"4290de40-64d7-4a92-bb00-91beb52b6a37","arxiv_id":"1908.04292","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"A total order on braid representatives powers a skein-relation recursion that computes HOMFLY-type link polynomials, with an empirical growth of about 1.42^c.","lead":"Knots can be drawn as braids, and this paper orders all braid pictures so that a knot's polynomial can be computed by always simplifying the picture. The result is a practical recipe for HOMFLY-type polynomials with a claimed speedup over naive methods, plus a way to tell links apart that is complete but computationally hard.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Lemma 3.5 is asserted, not proved: termination of Algorithm 3.4 and strict decrease in <br at every step are left unverified, so the induction in Theorem 3.6 rests on an unproved well-foundedness claim.","rationale":"After checking the definitions and the worked example, the reduction steps are individually plausible: Step 1's rotation brings a new minimum to the front and so decreases in <br; Step 2 shortens length; Step 3 shortens length; Cases 4.1 and 4.3 compare favorably in the first absolute-value entry or weight vector; the terminal cases in Case 4.2 and 4.3.1 appear to match Lemma 3.5. The algebraic equivalence in Theorem 3.6 case (3) uses the braid relation inverted for the relevant sign pattern, so no counterexample emerged there. However, the paper never supplies the induction that these local decreases imply termination: Lemma 3.5's proof is a restatement, and the finiteness claim before Theorem 2.3 is false because [n;-] is smaller than any nonempty braid for every n. These observations do not refute the algorithm, but they leave the central well-foundedness claim unproved. A machine-checked proof of Lemma 3.5, or a finite exhaustive search for a transition that is not a strict <br decrease, would settle whether the concern lands. Until then, the reader's CONDITIONAL verdict is appropriate and no change is needed.","tokens_in":6835,"tokens_out":19595,"duration_ms":207014,"concrete_test":"Formalize Lemma 3.5 in a proof assistant (Lean, Coq, or similar): define <br as in Definition 2.2, prove it is well-founded by showing the tuple (length, n, weight vector, absolute-value word, signed word) decreases lexicographically, and prove for each branch of Steps 1-4 that Algorithm 3.4 either stops in one of the three terminal forms or produces a representative strictly smaller in <br under a decreasing well-founded measure. If the formal proof cannot be completed, it will identify the exact transition whose decrease or termination is false; if it succeeds, the gap in Lemma 3.5 is closed and Theorem 3.6 is supported.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Theorem 3.6's induction is the central claim; its well-foundedness depends entirely on Lemma 3.5, which asserts that Algorithm 3.4 always reaches one of three terminal states. The proof of Lemma 3.5 is a one-sentence restatement of the algorithm's stop rules and gives no argument that any run actually reaches such a state. The surrounding claim that 'in each step braid representatives decrease' and 'ordered leading tag length is becoming longer' is not checked case by case. This matters because Step 1 and parts of Step 4 are equivalence moves (cyclic rotations), not obvious decreases; e.g. after Case 4.3.2 the algorithm restarts at Step 1 and must be shown not to loop. Also, the paper's stated finiteness justification is false: for any nonempty braid [n; b1,...,bk], the empty representative [n';-] is smaller for every n'≥1, so there are infinitely many smaller representatives. Well-foundedness of <br therefore requires a separate argument that is not supplied. If any transition cycles, or stops outside the three cases, Theorem 3.6 collapses. The reader's concern identifies exactly this gap.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper proposes a total order <br on the set of all braid representatives and uses it to build an algorithm (Algorithm 3.4) that repeatedly replaces a braid representative by an equivalent one that is smaller in this order. The author claims that this algorithm terminates in one of three normal forms and that, in the non-unknot cases, applying the skein relation at a distinguished crossing reduces the computation of the HOMFLY polynomial to strictly smaller braid representatives. Theorem 3.6 states that the HOMFLY polynomial of any link can therefore be computed inductively. The paper also claims, as a by-product, that the minimal braid representative in each equivalence class is a complete link invariant. A sample computation for a six-crossing knot and a Mathematica experiment for knots up to 12 crossings are reported.","tokens_in":6948,"tokens_out":10628,"duration_ms":109331,"significance":"If the main theorem were fully proved, the paper would provide a concrete, order-driven recursion for skein-defined link polynomials and a potentially practical computational scheme; the reported average branching factor of 1.42 is suggestive. The paper is honest that the method is exponential in the worst case, consistent with known hardness results. However, the proposed complete invariant is definitional: choosing the minimum of an arbitrary total order within each equivalence class defines an invariant tautologically, and the paper does not address its computability or topological content. The central value of the paper therefore rests entirely on the termination and strict-descent claims for Algorithm 3.4.","major_comments":[{"comment":"The proof of Lemma 3.5 does not prove termination; it only lists the three stopping rules and says that the three cases are exactly those terminated at Step 1, Case 4.2, and Case 4.3.1. The surrounding claim that every step decreases in <br is not literally true: Step 1 and parts of Step 4 are cyclic rotations, which are equivalence moves and can increase the representative in the order. For example, [2;1,2] rotates to [2;2,1], and [2;1,2] <br [2;2,1]. Since Theorem 3.6 relies on strict <br descent at every recursive branch, the well-foundedness of the algorithm's transition relation is the load-bearing premise, and it is neither proved nor checked case by case.","section":"Section 3, Lemma 3.5"},{"comment":"The statement 'For any given braid representative β, there are finitely many braid representatives which are smaller than β' is false. For any nonempty representative β=[n;b1,...,bk] with k>0, every empty representative [r;-] with r>=1 is smaller by condition (1) of Definition 2.2, since 0<k. Hence there are infinitely many smaller representatives. The proof of Theorem 2.3 is therefore invalid as written; if the intended claim is well-foundedness of the restriction of <br to each equivalence class, that must be proved separately and is not supplied.","section":"Section 2, before Theorem 2.3"},{"comment":"The terminal condition 'cq+1 = cq−1, and sgn(cq+1) = −sgn(cq)' is impossible as written: if cq>0 then cq−1 has positive sign, and if cq<0 then cq−1 has negative sign, so in both cases the signs are equal, not opposite. The intended condition, matching Subcase 4.3.1, appears to be |cq+1|=|cq|−1 with opposite signs. As printed, the subsequent verification that γ' is smaller than γ in Theorem 3.6 cannot be checked, and the displayed equivalence to δ in that proof needs to be re-verified after correcting the condition.","section":"Lemma 3.5(3) and Theorem 3.6, case (3)"},{"comment":"In case (3) of Lemma 3.5, the proof asserts that γ' is equivalent to δ by elementary relations (5) and (2), citing the braid relation (2.4), and then asserts a weight comparison w_{|c_{q-1}|}(δ)=w_{|c_{q-1}|}(γ)−1. No details of this braid-relation manipulation are given, and because the sign condition in the lemma is currently inconsistent, the claimed strict decrease is not demonstrated. This is the branch that must establish well-foundedness for the induction, so a complete, correct verification is required.","section":"Theorem 3.6, final paragraph"}],"minor_comments":[{"comment":"There are typographical and grammatical errors: 'meth od' should be 'method', and 'As by-product a new complete link invariant are obtained' should be 'As a by-product, a new complete link invariant is obtained.'","section":"Abstract and Section 1"},{"comment":"The tail of the rotated representative is written as 'b1, . . . , bk−1', which appears to be a typo for 'b1, . . . , b_{m−1}'. If m=3 and k=5, the rotation should end with b1,b2, not b1,...,b4.","section":"Algorithm 3.4, Step 1"},{"comment":"The condition 'bq = −cq < 0' is confusing; it should be stated as 'bq = −cq and bq < 0 < cq', since the intended meaning is that when the first sign difference occurs, the negative entry is smaller.","section":"Definition 2.2(5)"},{"comment":"The sentence 'The complicities is about 1.42c' should read 'The complexity is about 1.42^c'. Also, the geometric average over knots with crossing number at most 12 is an empirical observation, not a worst-case complexity bound, and the text should not imply otherwise.","section":"Section 4"},{"comment":"The worked example jumps between representatives without indicating which step of Algorithm 3.4 or which braid relation is used at each equality. Annotating the steps would make the example much easier to verify.","section":"Example in Section 3"}],"recommendation":"major_revision","confidential_remarks":"The central issue is that Lemma 3.5, on which Theorem 3.6 depends, is asserted rather than proved. The termination of Algorithm 3.4 and the strict descent property of its transitions are not established, and the current proof text even contains a false claim about the order <br in the lead-up to Theorem 2.3. The sign condition in Lemma 3.5(3) also appears internally inconsistent. These are fixable in principle if the author can supply a genuine termination measure and correct the statement, but without that the main theorem is unsupported. The computational experiments are encouraging but do not substitute for a proof."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague — The gist: this note proposes an order-driven way to compute HOMFLY-type polynomials on braid representatives, and it's the first place I've seen the 'both branches go down' trick made explicit. But the termination lemma is asserted, not proved, and the paper's own justification for well-foundedness is false. I wouldn't rely on it yet.\n\nWhat's actually new: the weight-refined lexicographic order on braid representatives (over Gittings' order, which the paper honestly says doesn't work for this), and Theorem 3.6's observation that if the algorithm terminates, the two skein branches have smaller representatives. That's a real algorithmic idea, clearly stated. The worked example is checkable; I traced the skein reductions as far as the OCR'd text allows and they hang together. The reported Mathematica run over 2977 knots with a 1.42^c empirical node growth is a nice signal, though no code or data are given.\n\nSoft spots: Lemma 3.5 is the load-bearing piece. Its proof is one sentence restating the three stop conditions. No argument is given that a run must hit one of them. The claim that every step decreases in <br is asserted, not checked. The reader's concern about Step 1's rotation is actually off the mark: moving the minimal absolute value to the front makes the word strictly smaller by condition (4) of Definition 2.2, so that step is a descent. But that doesn't repair the missing termination proof for the other steps, especially the restart in Case 4.3.2. Moreover, the paper's basis for well-foundedness is wrong: for any nonempty word, [n';-] is smaller for every n', so infinitely many representatives are smaller. You need a separate measure—say, lexicographic on (length, ordered leading tag length)—proved to decrease. That proof is absent.\n\nThe 'complete invariant' of Theorem 2.3 is also unsupported: it relies on the same false finiteness claim. It's definitional in spirit, but as stated it's not established.\n\nWho's this for: people working on computational skein algorithms and braid orders. If the termination question gets settled, this could be a useful tool. Right now it's an interesting but incomplete note. I'd send it to review—a good referee could push the author to either prove the invariant or find the counterexample. It shouldn't be cited as a proved result yet.","headline":"A real algorithmic idea for order-driven skein computation, but the termination proof is missing and the paper's finiteness justification is false.","tokens_in":7613,"tokens_out":5699,"would_cite":false,"duration_ms":57647,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["57M25"],"pacs":[],"model":"deepseek-v4-flash","headline":"By ordering all braid representatives, the paper reduces the HOMFLY skein recursion to a terminating induction and obtains a complete link invariant from the unique minimal representative.","keywords":["braid representatives","total order","skein relation","HOMFLY polynomial","link invariant","link polynomial","closed braid","algorithm"],"falsifier":"Run Algorithm 3.4 on every braid representative with, say, at most five strands and twelve generators; if any input produces a cycle, or terminates at a braid $\\gamma=[m;c_1,\\ldots,c_l]$ with $\\mathrm{ol}(\\gamma)=q>0$ that is not covered by $c_{q+1}=c_q$ or by $c_{q+1}=c_q-1$ with $\\operatorname{sgn}(c_{q+1})=-\\operatorname{sgn}(c_q)$, then Lemma 3.5 and Theorem 3.6 fail.","tokens_in":6478,"feed_emoji":"🧶","tokens_out":13487,"duration_ms":126828,"temperature":0.7,"pith_summary":"The paper claims that any link polynomial defined by a skein relation can be computed by a simple, deterministic algorithm that works directly on closed-braid representatives. The key move is a new total order on all braid representatives; the algorithm replaces a representative only by equivalent representatives that are smaller in this order, so the recursion cannot loop. When the algorithm stops, the remaining braid is either the empty braid, whose closure is an unknot, or has a distinguished crossing whose sign flip and deletion both give strictly smaller representatives; applying the skein relation there reduces the calculation to smaller links. As a by-product, the same order selects a unique minimal braid representative for every link, which the paper presents as a complete link invariant.","feed_headline":"New braid order calculates HOMFLY polynomials by induction","feed_subtitle":"A deterministic reduction to unknots replaces exponential guesswork, and each link gains a unique minimal braid.","key_machinery":"The load-bearing object is the total order $<_{\\mathrm{br}}$ on braid representatives (Definition 2.2), which compares representatives first by word length, then by number of strands, then by the vector of weights $w_m$ counting how many generators of each absolute value appear, then by the absolute values of the entries, and finally by their signs. The algorithm also uses the ordered leading tag length $\\mathrm{ol}(\\beta)$, the length of the initial run of entries whose absolute values increase by one, to decide which of three stopping cases applies. The order is what makes the recursion well-founded: each allowed replacement lowers the representative, and the terminal forms are exactly the places where the skein move sends both branches to still-lower representatives.","core_discovery":"The central claim is Theorem 3.6: given a braid representative of a link $L$, the HOMFLY polynomial of $L$ can be computed inductively by the skein relation $\\ell P(L_+)+\\ell^{-1}P(L_-)+mP(L_0)=0$ together with the normalization $P(\\mathrm{unknot})=1$, using Algorithm 3.4 as the reduction engine. At a terminal representative $\\gamma$, the two skein branches are $\\gamma'$, obtained by changing the sign of the $q$-th generator, and $\\gamma''$, obtained by deleting it; the proof shows both are strictly smaller than $\\gamma$ in the total order $<_{\\mathrm{br}}$, and the empty braid represents the unknot. Because each representative has only finitely many smaller representatives, the induction is well-founded. The paper further claims that the unique minimal representative under this order is a complete link invariant.","pith_inferences":["Because Step 1 of Algorithm 3.4 is a cyclic rotation rather than an order-decrease, a fully formal termination proof needs an invariant that treats rotations as neutral; searching for such a measure could either complete the proof or expose a hidden cycle.","If Algorithm 3.4 can be shown to compute the minimal representative itself, the total order would turn the complete invariant into a practical classification tool for links.","The measured growth factor of $1.42^c$ is an average over 2977 knots, not a worst-case bound; testing on longer braid words or on links with many components would show whether the practical improvement persists."],"forward_implications":["For any braid input, HOMFLY computation becomes a deterministic induction: at each nontrivial terminal braid exactly one crossing is flipped or deleted, and both resulting representatives are smaller in $<_{\\mathrm{br}}$.","The algorithm is not specific to HOMFLY: the proof uses only the skein relation and the unknot normalization, so the same reduction computes any link polynomial defined by that skein relation.","Every link gets a unique minimal braid representative under $<_{\\mathrm{br}}$, so the order supplies a complete link invariant; whether that representative can be found efficiently is a separate question the paper does not settle.","The recorded computation of HOMFLY polynomials for 2977 knots with braid crossing number at most 12 took 430 seconds in total, with an average node-growth factor of about $1.42^c$, suggesting the branching is much gentler than the generic $2^c$ of naive skein expansion."],"supporting_citations":[{"why":"Supplies the closed-braid representation of links and the corollary that the elementary moves of Definition 3.1 preserve the link type; this is what lets Algorithm 3.4 replace a braid representative by an equivalent one.","marker":"[1]"},{"why":"Defines the HOMFLY polynomial through the skein relation and unknot normalization that the algorithm evaluates recursively.","marker":"[5]"},{"why":"Introduces the lexicographic order on braid representatives that Definition 2.2 extends with weight data; the new order is the paper's central machinery.","marker":"[6]"}],"fun_headline_variants":["Braid order tames skein relations for HOMFLY","Inductive braid reduction computes HOMFLY fast","Unique minimal braid gives complete link invariant","Simple algorithm: braid order yields HOMFLY","New total order on braids computes HOMFLY"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The argument rests on Algorithm 3.4 always terminating in one of the three listed terminal forms, with every nonterminal step strictly lowering the representative in the order; the paper's proof of this is a restatement of the cases, and Step 1 is a rotation rather than an order-decrease, so the decreasing measure is not fully exhibited.","fun_headline_variants_meta":{"raw":{"variants":["Braid order tames skein relations for HOMFLY","Inductive braid reduction computes HOMFLY fast","Unique minimal braid gives complete link invariant","Simple algorithm: braid order yields HOMFLY","New total order on braids computes HOMFLY"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000372,"raw_usage":{"total_tokens":1896,"prompt_tokens":761,"completion_tokens":1135,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":377,"completion_tokens_details":{"reasoning_tokens":1055}},"tokens_in":377,"tokens_out":1135,"duration_ms":8848,"temperature":1.0,"reasoning_tokens":1055,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T14:08:56.979654+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Run Algorithm 3.4 on every braid representative with, say, at most five strands and twelve generators; if any input produces a cycle, or terminates at a braid $\\gamma=[m;c_1,\\ldots,c_l]$ with $\\mathrm{ol}(\\gamma)=q>0$ that is not covered by $c_{q+1}=c_q$ or by $c_{q+1}=c_q-1$ with $\\operatorname{sgn}(c_{q+1})=-\\operatorname{sgn}(c_q)$, then Lemma 3.5 and Theorem 3.6 fail.","supporting_citations":[{"cited_title":"Braids, links, and mapping class groups","cited_arxiv_id":null,"evidence_quote":"Supplies the closed-braid representation of links and the corollary that the elementary moves of Definition 3.1 preserve the link type; this is what lets Algorithm 3.4 replace a braid representative by an equivalent one."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Defines the HOMFLY polynomial through the skein relation and unknot normalization that the algorithm evaluates recursively."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Introduces the lexicographic order on braid representatives that Definition 2.2 extends with weight data; the new order is the paper's central machinery."}],"review_version":1}