Ricardo Pinho
@rpinho
- Affiliation
- Independent
Publications
- rstackPublic
By Research tools contributors
0 stars
- ai physicsPublic
By @rpinho
0 stars
- ai biologyPublic
By @rpinho
0 stars
By @rpinho
0 stars
- math certificates paperPublic
By @rpinho
0 stars
Claims 58
rstack
- In an earlier line of work, running about 38 workers for two days produced 528 claims of which the reviewer passed 98%, yet none of it was any good despite every check being green.Unverified
- Out of 12 guards examined in the author's system, 5 crashed instead of refusing, which allowed hooks to let them go through.Unverified
- The author opened 129 PRs on other people's repositories, of which 15 are merged (6 bounds in Tao's optimization constants table and 9 bug fixes), 106 are still open, and 8 were closed, as of a November 18, 2026 snapshot.Unverified
- Planted false claims caught a failure mode on 11-09 where GPT referees found the real bug and still said yes to the fake claim, which was fixed by adding one line: judge the claim as it is written.Unverified
- Four reports based on NASA TESS data sent to Research Notes of the American Astronomical Society (RNAAS) were all corrected after re-auditing the numbers, with corrected versions on Zenodo crediting two earlier public sources for some of the signals.Unverified
- For the MecCog pipeline on APOE4 hypotheses (H1, H2, H3), a second model checked every row without human checking, and in a 5-row check of H2 by Claude, all 5 matched while also finding 2 bad rows that the code now catches.Unverified
- 131 of the 356 rules in rstack have something enforcing them, while 225 are still prose.Unverified
ai physics
- For the de Sitter bootstrap bound at and principal-series gap , a concrete four-term rational Legendre functional yields the certified upper bound , improving on the printed value in arXiv:2107.13871 eq. (5.53).Verified
- In the leading-power double-light-cone sector of Stampedes I (arXiv:2111.12131v2) Appendix E, retaining all inverse-log corrections with and , the single-light-cone seed satisfies , or equivalently with .Verified
- For every and every prescribed , there exists a strongly regular static quasi-linear Signorini ball with finite radius, positive density, and positive interior principal pressures, satisfying the virial radius bound .Verified
- For monomial scalar field inflation (arXiv:1503.06994v2), the Padé approximant in the inverse center-manifold variable has certified relative Hubble error at the first endpoint, and added rows for have certified upper bounds , , and , respectively.Verified
- For tracking quintessence (arXiv:2407.14378v3) with and , , and , the analytically generated Padé approximant with self-consistent normalization achieves certified uniform relative errors and .Verified
- For the fused-link map at and , the fused-link image has exact dimension , whereas ordinary invariants () have dimension in strict integer weight zero and over all sectors; an explicit witness satisfies while annihilates all link columns, generalizing to an explicit obstruction family for every odd at .Verified
- KITE (commit a1e1f7ce6e48505ce2655315c2565446315a67e0) contains four confirmed correctness defects: KITE-CFG-001 (nonzero energy shift invents interorbital couplings in multi-orbital onsite matrices), KITE-TOOLS-001 (file temperature is ignored in both optical conductivity branches), KITE-TOOLS-002 (Green LDOS omits a factor of ), and KITE-TOOLS-003 (threaded LDOS silently drops remainder moments for non-divisible counts).Verified
- In Guedes, Oriti, and Raasakka (arXiv:1301.7750, Sec. 5.2), the displayed symmetric cubic coordinate words disagree with the map's definition in of cases (calculated minus displayed: ) and the Duflo cubics disagree in of cases (calculated minus displayed: ), while all quadratics per map and all FLM cubics agree.Verified
- The exact inverse images for lengths ( words per map, total) for the symmetric, Duflo, and FLM quantization maps under match the computed coordinate-word table.Verified
ai biology
- The maximum number of hyperbolic asymptotically stable nonnegative equilibria in the printed Garcia, Bonhoeffer, and Fu (2020) NK-CTL model over all strictly positive constants is 3, and with non-coupling constants fixed to printed Table 3, the sharper maximum is 2.Verified
- With every non-coupling constant fixed to printed Table 3, for all , , and every finite nonnegative initial state, the printed three-dimensional Garcia, Bonhoeffer, and Fu (2020) NK-CTL system converges to one equilibrium, excluding nonconstant periodic or other recurrent non-equilibrium limit sets.Verified
- Fixing the Table 3 immune rates in the printed Garcia, Bonhoeffer, and Fu (2020) NK-CTL model, every finite nonnegative trajectory converges to one equilibrium for every and every .Verified
- For Figure 2B of Garcia et al. (2025), Kruskal NMDS stress on the same 23 sequences, 253 pairs, two dimensions, Euclidean raw Shah-Gilchrist Table S2 Rc rate profiles, and MASS Stress-1 objective is improved from the printed 34.8 percent to 16.9693901688774 percent.Verified
- For Figure 2C of Garcia et al. (2025), Kruskal NMDS stress on the same 23 sequences, 253 pairs, two dimensions, Euclidean raw Rudorf-Lipowsky Table S2 rate profiles, and MASS Stress-1 objective is improved from the printed 23.4 percent to 13.4830828456807 percent.Verified
- In a retrospective empirical test against later Danish SSI (2021-2022) and national MiBa (2021-2024) macrolide-resistance surveillance fractions, the unchanged Cadosch et al. (2020) Danish forecast achieves lower unweighted MAE (10.8641 and 16.8311 percentage points) than both the trivial constant baseline (24.8732 and 21.0295) and a historical linear-time binomial logistic regression baseline (13.9194 and 19.9468).Verified
- Independent refereeing confirmed multiple reproducible software correctness defects across published scientific repositories, including ExOrthist (27 confirmed/minor across A/B/C/D and M011), vast-tools (6 confirmed/minor), pyglmnet (11 confirmed/minor), structSSI/miniLNM/multimedia/scDesigner (7 confirmed/minor), GO::TermFinder (6 confirmed/minor), and COVID-Twitter-BERT/FoodRepo/PlantVillage (5 confirmed/minor).Verified
research software bug hunt
Show 14 major claimsHide 14 major claims
- In glm-tools/pyglmnet (base commit 9339c5c71c908ec31cf4bebda5d9fe47dfed4f1c), three correctness defects were confirmed and reported via pull requests #399 (merged), #400 (open), and #401 (merged): Bernoulli log likelihood NaN/overflow at probability endpoints and large logits, gamma deviance and pseudo-R2 omitting the saturated log likelihood, and simulate_glm(sample=True) for gamma returning exp(mu) instead of gamma samples.Verified
- In omarwagih/rmotifx (commit f13775f07679b86df173ccc2a9112263a1a9052f), upper-tail subtraction in R/motifx-functions.R:96 rounds representable binomial tails to zero, inflating the default motif score on twenty ASA strings against 400 aSb strings from 52.041200 to 615.305311.Verified
- In evocellnet/funscoR (commit 1ed6c74f112fd381626295245da327453d70cbe2), passing non-NULL parameters to R/predict_score.R:58 crashes with an undefined param error because param is never initialized from parameters.Verified
- In evocellnet/ksea (commit 126a64541a11df9d8dc16a00cd9ea3ab9f16147f), two defects were confirmed: (1) null distribution permutation scores in R/ksea.R:120 ignore the power parameter p and use default p=1 (shifting p=2 empirical significance from 0.277 to 0.154), and (2) classic enrichment (p=0) in R/ksea.R:56 counts nonmembers as hits via 0^0 (shifting ES from 1/2 to 1/6).Verified
- In evocellnet/ptm_hotspots (commit 15b4b256f7f2945eebe72f622d68115d4bdd5e70), one minor defect was confirmed: ptm_hotspots.py:203 omits the last permutation when summarizing draws, using 99/100 draws and producing NaN p-values on two-draw runs (and shifting default raw p at the first reported position from 0.005557604 to 0.005796981).Verified
- In bentoayr/Efficient-Projection-onto-the-Perfect-Phylogeny-Model (commit 769ccd389a559085df8fa628047eba2f2d850c15), two defects were confirmed: (1) project_onto_PPM.c:970 misreads documented column-major input tokens as row-major (inflating equality cost from 0 to 0.562501), and (2) project_onto_PPM.c:917 adds unscaled absolute jitter after D^2 F scaling, breaking small weights (yielding invalid child fraction -49.374222 instead of 0.250000).Verified
- In edbosne/pyfdd (commit 6aba44c5c4aaed278e08691f28abe79c724cf1da), seven defects (including 3 minor) were confirmed: angular grid index scaling distortion (B01), JSON reload discarding backing masks (B02), orientation transforms leaving backing masks behind (B03), in-place background mutation retaining earlier exclusions (B04), direct-input degrees of freedom miscount and np.Array typo (B05, minor), omitted zero-probability costs (B06, minor), and zero-minimum uncertainty collapse (B07, minor).Verified
- In borenstein-lab/MUSiCC (commit 478c9244cf406935e67abdf7622bdb05421ab1fb), musicc/core.py:177 assigns corrected abundances into integer storage when given integer input, truncating abundances (e.g., K01866=1, K00001=1, K01867=2 becomes [0, 1, 2] instead of [0.9914541785, 1.0066761508, 2.0133523016]).Verified
- In borenstein-lab/mimosa2 (commit 581b307dee48577117ac42599a07b33a51b0822d), two defects (including 1 minor) were confirmed: (1) R/core_mimosa2_funcs.R:1389 drops relAbund and remove_rev options when dispatching to get_cmp_scores_kos (shifting CMP from (-5,5) to (-400,400)), and (2) R/core_mimosa2_funcs.R:485 computes sparse-zero covariance shares over unequal sample sets (minor, shifting shares from (0,1,0) to (0,0,0)).Verified
- In borenstein-lab/MAAMOUL (commit 54f174c5a3a8d07ebac5b3431c43eb4bbfdc83b9), R/utils.R:102 crashes with a row-count error when either metabolites or ECs (or both) have complete coverage because scalar attributes cannot form a zero-row data.frame for imputed nodes.Verified
- In glarue/intronIC (commit 0db1d54274f1ffdbbc95dea6589a75a833eb831d), five defects (including 1 minor) were confirmed: undeclared pandas dependency in pyproject.toml (R01), strand-dependent branch-point window coordinate conversions (R02), empty 3' quality slice on short flanks (R03), --no-headers classification aborting (R04), and GenomeReader.close() AttributeError on generator exhaustion (R05, minor).Verified
- In feedzai/deccaf (commit b910b48d7e146d53b5f145224b819bdb0962d40a), three defects (including 2 minor) were confirmed: positional summation of capacity columns corrupting quotas when reordered (DECCAF-001, minor), batch label loops dropping cases or crashing on non-1 labels (DECCAF-002, minor), and FPR threshold selection exceeding the 5% budget due to -1 indexing and boundary ties (DECCAF-003).Verified
- In weissmanlab/magic (commit 87fd0923480d28399c5078ff4efd60c3205c22a0), four defects were confirmed: windower.py:61 loses the final window due to too few histogram edges (W01), windower.py:77 crashes with NameError on table inputs and merges scale 0 too early (W02), magic.py:503 omits the infinite tail in the piecewise-exponential mean (M01), and magic.py:427 crashes with TypeError when serializing a scalar zero-mass parameter (M02).Verified
- In GO::TermFinder 0.86, six defects (including 2 minor) were confirmed: compound NOT qualifiers treated as positive annotations (GTF-001), tied p-values receiving different FDR estimates at the same cutoff (GTF-004), duplicate background names/aliases biasing simulation sampling (GTF-005), obsolete term status depending on tag order (GTF-007, minor), Typedef stanzas terminating ontology parsing (GTF-009), and cyclic ontology input causing uncontrolled recursion (GTF-010, minor).Verified
math certificates paper
- Turán's pure power-sum constant satisfies , conditional on Bakeas's Theorem A, with two Arb verifiers giving and .Unverified
- For every , , and hence Turán's pure power-sum constant satisfies .Unverified
- The constant in for bounded-degree Boolean functions satisfies , with , , , , and .Unverified
- The dimension-free reverse Brunn–Minkowski constant for centrally symmetric convex bodies satisfies , with a separate finite witness certifying .Unverified
- The Schur–Siegel–Smyth trace constant satisfies .Unverified
- The essential minimum of the logarithmic Zhang–Zagier height satisfies .Unverified
- Korenblum's constant in the Bergman space satisfies .Unverified
- The average edit distance constant of two random binary strings satisfies .Unverified
- The Ramsey multiplicity constant of satisfies , with exact upper certificate density .Unverified
- The Ramsey multiplicity constant of satisfies .Unverified
- The univalent Bloch–Landau constant satisfies .Unverified
- The centred Hardy–Littlewood maximal constant over axis-parallel squares in dimension satisfies .Unverified
- The tight alternating knot constant satisfies ().Unverified
- The connective constant of the square lattice satisfies .Unverified
- The 3-sunflower-free capacity satisfies .Unverified
- For the Erdős–Ulam two-colouring sublattice function , , , , , , , , , , ( conditional on BMB `lem:window20`), , ( conditional on BMB `lem:window20`), , and .Unverified
- The growth rate of 1324-avoiding permutations satisfies , conditional on the cited BBEP results and the retained machinery of Norton's unrefereed v3.Unverified
- The four-letter LCS constant satisfies .Unverified
- There exists an explicit cover of the regular tetrahedron by 4,096 cylinders with compact triangular perpendicular bases whose normalized total base area satisfies .Unverified
- Nine explicit linear codes over and improve the lower bounds on minimum distance in Markus Grassl's codetables.de tables: , , , , , , , , and .Unverified