BVSR
Collection
Improving Creative Idea Generation in AI Scientists • 3 items • Updated • 2
Dataset Preview
The full dataset viewer is not available (click to read why). Only showing a preview of the rows.
The dataset generation failed
Error code: DatasetGenerationError
Exception: CastError
Message: Couldn't cast
cell: string
id: string
reward: double
c: double
code: string
split_sr: string
run: string
score: double
to
{'run': Value('string'), 'score': Value('float64'), 'split_sr': Value('string'), 'code': Value('string')}
because column names don't match
Traceback: Traceback (most recent call last):
File "/usr/local/lib/python3.14/site-packages/datasets/builder.py", line 1835, in _prepare_split_single
for key, table in generator:
^^^^^^^^^
File "/src/services/worker/src/worker/job_runners/config/parquet_and_info.py", line 621, in wrapped
for item in generator(*args, **kwargs):
~~~~~~~~~^^^^^^^^^^^^^^^^^
File "/usr/local/lib/python3.14/site-packages/datasets/packaged_modules/json/json.py", line 361, in _generate_tables
self._cast_table(pa_table, json_field_paths=json_field_paths),
~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/usr/local/lib/python3.14/site-packages/datasets/packaged_modules/json/json.py", line 132, in _cast_table
pa_table = table_cast(pa_table, self.info.features.arrow_schema)
File "/usr/local/lib/python3.14/site-packages/datasets/table.py", line 2381, in table_cast
return cast_table_to_schema(table, schema)
File "/usr/local/lib/python3.14/site-packages/datasets/table.py", line 2309, in cast_table_to_schema
raise CastError(
...<3 lines>...
)
datasets.table.CastError: Couldn't cast
cell: string
id: string
reward: double
c: double
code: string
split_sr: string
run: string
score: double
to
{'run': Value('string'), 'score': Value('float64'), 'split_sr': Value('string'), 'code': Value('string')}
because column names don't match
The above exception was the direct cause of the following exception:
Traceback (most recent call last):
File "/src/services/worker/src/worker/job_runners/config/parquet_and_info.py", line 1379, in compute_config_parquet_and_info_response
parquet_operations, partial, estimated_dataset_info = stream_convert_to_parquet(
~~~~~~~~~~~~~~~~~~~~~~~~~^
builder, max_dataset_size_bytes=max_dataset_size_bytes
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
)
^
File "/src/services/worker/src/worker/job_runners/config/parquet_and_info.py", line 956, in stream_convert_to_parquet
builder._prepare_split(split_generator=splits_generators[split], file_format="parquet")
~~~~~~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/usr/local/lib/python3.14/site-packages/datasets/builder.py", line 1702, in _prepare_split
for job_id, done, content in self._prepare_split_single(
~~~~~~~~~~~~~~~~~~~~~~~~~~^
gen_kwargs=gen_kwargs, job_id=job_id, **_prepare_split_args
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
):
^
File "/usr/local/lib/python3.14/site-packages/datasets/builder.py", line 1887, in _prepare_split_single
raise DatasetGenerationError("An error occurred while generating the dataset") from e
datasets.exceptions.DatasetGenerationError: An error occurred while generating the datasetNeed help to make the dataset viewer work? Make sure to review how to configure the dataset viewer, and open a discussion for direct support.
run string | score float64 | split_sr string | code string |
|---|---|---|---|
svg_ar_kimi_45214125 | 0.168146 | train | def rank_svg(adata):
# Transplant of the Lo-MacKinlay variance-ratio test (heteroskedasticity-robust
# random-walk specification test) into spatial ranking.
#
# Finance mechanism: a price series with genuine structure fails the random-walk
# identity Var(k-period return) = k * Var(1-period return). ... |
svg_ar_kimi_45214125 | 0.762165 | train | def rank_svg(adata):
# Transplanted mechanism: NONPARAMETRIC BOTTOM-UP CENTER-SURROUND SALIENCY.
# In the paper, saliency is a structure-vs-background discriminator estimated
# without committing to a fixed filter bank: the maximally excitatory stimulus
# that emerges is a center-surround operator. Here... |
svg_ar_kimi_45214125 | 0.705736 | train | def rank_svg(adata):
# Transplanted mechanism: BAO Template Fitting with Mock-Catalog Significance Calibration.
# Per feature ("tracer"), we measure the spatial two-point correlation function xi(r)
# (monopole) and its cos(2*theta) quadrupole moment (anisotropy), then project them onto a
# bank of dampe... |
svg_ar_kimi_45214125 | 0.656692 | train | def rank_svg(adata):
# Transplanted mechanism: Common Spatial Patterns (CSP) spatial-filter learning with
# eigenvalue ranking (EEG spatial filter selection). For EACH feature we treat its
# expression as a "condition" weighting over samples and jointly diagonalize two
# covariance matrices over a bank ... |
svg_ar_kimi_45214125 | 0.762544 | train | def rank_svg(adata):
# Transplant of Kulldorff's Spatial Scan Statistic (SaTScan) into feature ranking.
#
# Mechanism, committed to the statistic:
# 1) WINDOW BANK: circular scan windows at many centers x radii (multi-scale, any location),
# restricted to the classic Kulldorff range (<=50% of t... |
svg_ar_kimi_45214125 | 0 | train | def rank_svg(adata):
# Transplanted mechanism: DBSCAN CORE-POINT STATISTIC for spatial randomness testing
# (Koo & Hwang, wafer-map defect pattern analysis). Each feature's high-expression
# cells are treated as "defects" on the wafer (tissue); the number of DBSCAN core
# points among them is the projec... |
svg_ar_kimi_45214125 | 0.829956 | train | def rank_svg(adata):
# Transplant of Rossmo's Criminal Geographic Targeting (CGT) jeopardy surface:
# each feature's expression values play the role of a crime series; we overlay a
# grid on the tissue and score every grid cell as a candidate "anchor point" via
# Rossmo's distance-decay function WITH A ... |
svg_ar_kimi_45214125 | 0 | train | def rank_svg(adata):
# Transplanted mechanism: INHOMOGENEOUS PAIR CORRELATION FUNCTION (point-process ecology).
# Each feature is treated as a marked point pattern over the tissue: samples are points,
# expression is the mark. We estimate the scale-resolved pair correlation g(r) with the
# first-order i... |
svg_ar_kimi_45214125 | 0.859117 | train | def rank_svg(adata):
# Mechanism transplanted from Zeng, Liang & Zhang's BLINDLY COMBINED ENERGY DETECTION
# (spectrum sensing under unknown noise power):
#
# * Each spatial kernel is a "channel". Per channel we measure an ENERGY statistic:
# the projection energy of the feature's count vector... |
svg_ar_kimi_45214125 | 0.738685 | train | def rank_svg(adata):
# Transplanted mechanism: ELLIPTIC SPATIAL SCAN STATISTIC (Kulldorff-style,
# elliptic extension). For every feature we scan a bank of anisotropic windows
# (center x semi-major scale x eccentricity x rotation angle), compute the
# Poisson likelihood-ratio statistic contrasting the ... |
svg_ar_kimi_45214125 | 0.327397 | train | def rank_svg(adata):
# Cyclic Correlation Bank (Gardner spectral-correlation transplant).
# Each feature's spatial field is treated as a 2-D signal; we compute its
# spectral correlation density at a bank of nonzero cyclic frequencies
# (frequency shifts alpha). Stationary sampling noise has ~zero spect... |
svg_ar_kimi_45214125 | 0.799848 | train | def rank_svg(adata):
# Transplant of Mokken Scale Analysis (Loevinger's H) into spatial ranking.
#
# Mokken H for an item = cov(item, rest-score) / cov_max(item, rest-score),
# where cov_max is the Frechet upper bound: the covariance attainable under a
# comonotone (monotone-homogeneity) rearrangeme... |
svg_ar_kimi_45214125 | 0.186135 | train | def rank_svg(adata):
# Transplant: LEAST TRIMMED SQUARES array processing (Bishop et al., robust
# infrasound beamforming). In seismology, plane-wave parameters across a sensor
# array are fit by LTS: iteratively refit on the h sensors with smallest
# residuals, so corrupted elements (dead channels, clo... |
svg_ar_kimi_45214125 | 0.797197 | train | def rank_svg(adata):
# 4PL-IRT transplant: every feature is an "examinee", every spatial kernel is an "item".
# The 4PL model bounds the response curve away from 0 and 1 with a GUESSING floor (structure
# a pure-noise feature fakes by chance) and a CARELESSNESS ceiling (structure a truly
# organized fea... |
svg_ar_kimi_45214125 | 0.856277 | train | def rank_svg(adata):
# Transplanted mechanism (ecology: aggregated species distributions / quadrat analysis):
# NEGATIVE-BINOMIAL-CALIBRATED QUADRAT AGGREGATION ACROSS SCALES.
# Space is partitioned into quadrat grids at many resolutions. For each gene and each
# grid, quadrat counts are tested agains... |
svg_ar_kimi_45214125 | 0.823896 | train | def rank_svg(adata):
# Transplanted mechanism: TEST-RETEST REPRODUCIBILITY RANKING (radiomics/RIDER).
# Each feature's spatial-organization PROFILE (projection statistics across a bank of
# multi-scale anisotropic kernels) is treated like a radiomic feature vector measured on a
# tumor. "Repeat acquisit... |
svg_ar_kimi_45214125 | 0.293312 | train | def rank_svg(adata):
import numpy as np
import scipy.sparse as sp
from sklearn.neighbors import NearestNeighbors
X = np.asarray(adata.obsm['spatial'], dtype=np.float64)
Y = adata.layers['normalized']
Y = Y.toarray() if sp.issparse(Y) else np.asarray(Y, dtype=np.float64)
n, g = Y.shape
Y... |
svg_ar_kimi_45214125 | 0.163601 | train | def rank_svg(adata):
# Transplant of "Nonparametric Spatial Ordinal Pattern (SOP) control charts"
# (Adämmer, Wittenberg & Weiß) into a per-feature spatial-dependence score.
#
# Mechanism, faithfully moved into the statistic itself:
# 1. SPATIAL ORDINAL PATTERNS: each sample's local spatial configu... |
svg_ar_kimi_45214125 | 0.782995 | train | def rank_svg(adata):
# Transplant of Kulldorff's Spatial Scan Statistic (SaTScan) into feature ranking.
#
# Mechanism, faithfully implemented per feature:
# 1. MULTI-SCALE CIRCULAR WINDOW BANK: overlapping circles centered on a grid over
# the tissue, with radii spanning ~5%..50% of the spott... |
svg_ar_kimi_45214125 | 0.776178 | train | def rank_svg(adata):
# Transplanted mechanism: TWO-POINT CORRELATION FUNCTION statistics from random-media
# microstructure characterization. For each feature we measure the spatial two-point
# correlation S2(r, theta) = <y_i * y_j> over all inter-point pairs, resolved jointly over
# distance lags r AND... |
svg_ar_kimi_45214125 | 0.768982 | train | def rank_svg(adata):
# MUSIC transplanted to spatial variation ranking.
#
# In direction-of-arrival MUSIC, one forms the spatial covariance of array
# snapshots, eigendecomposes it into a SIGNAL subspace (top eigenvectors =
# coherent modes) and a NOISE subspace (small-eigenvalue eigenvectors), then... |
svg_ar_kimi_45214125 | 0.708387 | train | def rank_svg(adata):
# HPS transplant: build a 2-D spatial periodogram per feature (projection power onto
# cos/sin plane waves over a grid of directions x frequencies), then score each feature
# by a comb-template statistic: for every candidate (direction, fundamental), multiply
# (geometric-mean) the ... |
svg_ar_kimi_45214125 | 0.415638 | train | def rank_svg(adata):
# Transplanted mechanism: WEIGHTED LOCAL-LIKELIHOOD FRONT DETECTION with
# uncertainty quantification (ocean SST front detection).
#
# For each feature we treat its spatial field as a noisy "sea surface" and ask:
# is there a front — a smooth curve across which the field makes a... |
svg_ar_kimi_45214125 | 0.741904 | train | def rank_svg(adata):
import numpy as np
import scipy.sparse as sp
from sklearn.metrics import pairwise_distances
Y = adata.layers['normalized']
Y = Y.toarray() if sp.issparse(Y) else np.asarray(Y)
Y = np.asarray(Y, dtype=np.float64)
coords = np.asarray(adata.obsm['spatial'], dtype=np.float6... |
svg_ar_kimi_45214125 | 0.700813 | train | def rank_svg(adata):
# Transplanted mechanism: COMPLEX (HILBERT) EOF ANALYSIS for propagating wave-train
# detection (meteorology: complex PCA of Hilbert-transformed station series).
#
# For each of several propagation directions, the 2-D field is collapsed into an
# ordered 1-D "station series" (bi... |
svg_ar_kimi_45214125 | 0.73717 | train | def rank_svg(adata):
# Transplant: Frequency-Wavenumber (f-k) delay-and-sum BEAMFORMING with a steering-vector
# bank (RETREAT / ObsPy array seismology). Each sample is a seismometer at a known location;
# a feature's expression field is the array's amplitude snapshot. We steer the array over a
# bank o... |
svg_ar_kimi_45214125 | 0 | train | def rank_svg(adata):
# Transplanted mechanism: Spatio-Spectral Decomposition (SSD; Nikulin et al.,
# "reliable and fast extraction of neuronal oscillations via spatio-spectral
# decomposition"). In EEG, SSD builds filters whose TARGET-BAND power is
# maximized relative to FLANKING-BAND power (generalize... |
svg_ar_kimi_45214125 | 0.854004 | train | def rank_svg(adata):
# Transplanted mechanism: GABOR FILTER BANK + GRATING CELL OPERATOR (texture vision).
# Each gene's spatial count field is probed with a bank of localized, oriented,
# quadrature Gabor filters (Gaussian envelope x cos/sin carrier) spanning multiple
# positions, orientations and spat... |
svg_ar_kimi_45214125 | 0.34728 | train | def rank_svg(adata):
# Transplant of Kelly's GLRT adaptive detector (radar multichannel adaptive
# signal detection) into spatially-variable-feature ranking.
#
# Radar view: each feature's expression vector over the 2500 spots is a
# PRIMARY snapshot; the spatial "clutter" (unstructured sampling noi... |
svg_ar_kimi_45214125 | 0.806475 | train | def rank_svg(adata):
import numpy as np
import scipy.sparse as sp
rng = np.random.RandomState(0)
counts = adata.layers['counts']
counts = counts.tocsr() if sp.issparse(counts) else sp.csr_matrix(counts)
coords = np.asarray(adata.obsm['spatial'], dtype=np.float64)
n_obs, n_gene = counts.shap... |
svg_ar_kimi_45214125 | 0.740389 | train | def rank_svg(adata):
# Transplant of FieldTrip-style cluster-based permutation testing (Maris & Oostenveld)
# to spatial transcriptomics:
# 1) sample-level statistic: locally-smoothed (neighbor-averaged) z-scored expression,
# which is sensitive to coherent structure of ANY shape/scale but not to... |
svg_ar_kimi_45214125 | 0.473393 | train | def rank_svg(adata):
# Transplanted mechanism: 2PL/3PL IRT orthogonalization of discrimination vs.
# difficulty vs. pseudo-guessing, applied to spatial-coherence measurement.
#
# Each of K anisotropic / multi-scale spatial kernels is an "examinee" with a
# latent sensitivity theta_j to true spatial ... |
svg_ar_kimi_45214125 | 0.658586 | train | def rank_svg(adata):
# Transplant of Kulldorff's Space-Time Permutation Scan Statistic (case-only).
# Mechanism committed to the statistic:
# (1) CASE-ONLY modeling: each feature's counts are the "cases"; no expression-level
# covariate enters — a highly expressed feature gets no a-priori advantag... |
svg_ar_kimi_45214125 | 0 | train | def rank_svg(adata):
import numpy as np
import scipy.sparse as sp
from scipy.spatial import cKDTree
X = adata.layers['normalized']
X = X.toarray() if sp.issparse(X) else np.asarray(X)
X = np.ascontiguousarray(X, dtype=np.float32)
coords = np.asarray(adata.obsm['spatial'], dtype=np.float64)
... |
svg_ar_kimi_45214125 | 0.793788 | train | def rank_svg(adata):
# Transplant of Ramsay's TestGraf / KernSmoothIRT mechanism: kernel-smoothed
# nonparametric item characteristic curves.
#
# In IRT terms: each spatial direction u defines a "latent continuum" t = coords @ u
# (the ability axis theta). Each feature is an "item" whose spatial res... |
svg_ar_kimi_45214125 | 0.713879 | train | def rank_svg(adata):
# Transplant of rsPCA (Rotated Spectral PCA for climate modes) into spatial-feature ranking.
#
# The borrowed mechanism, committed to the statistic itself:
# * Each feature's spatial field is rasterized onto a 2-D grid (the "climate field").
# * Its cross-spectral content is d... |
svg_ar_kimi_45214125 | 0.772959 | train | def rank_svg(adata):
# Transplanted mechanism: ICA-dipolarity scoring ("Independent EEG Sources Are Dipolar").
# In EEG, each independent component's scalp map is fit with an equivalent-dipole template
# and components are ranked by the residual variance of that fit: a compact, physiologically
# coheren... |
svg_ar_kimi_45214125 | 0.743987 | train | def rank_svg(adata):
# Transplanted mechanism (wafer-map CNN defect-family recognition):
# each feature's spatial expression map is treated like a wafer map, and a
# bank of fixed "canonical defect family" matched filters (blobs/center,
# donut rings, edge steps, oriented stripe gratings — each at sever... |
svg_ar_kimi_45214125 | 0.836204 | train | def rank_svg(adata):
# Transplanted mechanism: Collaborative Cyclostationary Spectrum Sensing with
# SOFT STATISTIC COMBINING (Lunden, Koivunen, Huttunen & Poor).
#
# In cognitive radio, each sensor tests many cyclic hypotheses (unknown modulation
# patterns) and the fusion center aggregates the RAW... |
svg_ar_kimi_45214125 | 0.497063 | train | def rank_svg(adata):
# Transplanted mechanism: Multiscale Morphology Filter (MMF) from cosmic-web segmentation.
# Each feature's spatial field is smoothed with a bank of Gaussian scales; at every location
# and scale we compute the Hessian of the smoothed field analytically (second-derivative-of-
# Gaus... |
svg_ar_kimi_45214125 | 0.840749 | train | def rank_svg(adata):
# White's Reality Check (bootstrap data-snooping correction) transplanted to spatial
# pattern detection. A "universe of trading rules" becomes a bank of spatial pattern
# detectors (gradients, oriented Fourier stripes at several frequencies, isotropic and
# anisotropic Gaussian pat... |
svg_ar_kimi_45214125 | 0.706115 | train | def rank_svg(adata):
# Transplant of Hansen's SPA test (2005) into spatial-structure ranking.
#
# SPA anatomy, mapped onto this problem:
# * "model family" -> a large bank of spatial detector templates
# (anisotropic bumps, oriented stripes at several
# ... |
svg_ar_kimi_45214125 | 0.742472 | train | def rank_svg(adata):
# Transplant of Lo's Modified R/S test (long-run variance normalized rescaled range)
# into spatial structure ranking.
#
# Mechanism: for a series ordered along a spatial axis, form the partial-sum process
# of mean-centered expression (the "rescaled range" numerator R = max cum... |
svg_ar_kimi_45214125 | 0.859496 | train | def rank_svg(adata):
# Transplant of Christensen et al.'s "joint fundamental frequency and order
# estimation via optimal filtering" (speech multi-pitch analysis) into spatial
# transcriptomics:
#
# * COMB FILTER BANK <-> for each spatial direction u and candidate
# fundamental spatial f... |
svg_ar_kimi_45214125 | 0.774284 | train | def rank_svg(adata):
# Transplanted mechanism: Landy-Szalay two-point correlation estimator
# xi(r) = (DD - 2DR + RR) / RR
# For each feature, cells are weighted by expression; DD = expression-weighted
# pair counts per distance bin, RR = pair counts of a Poisson "random catalog"
# (all cells, wei... |
svg_ar_kimi_45214125 | 0.835771 | train | def rank_svg(adata):
# GP-SMART transplant (geographic profiling of suspects) into spatially-variable ranking.
#
# Crime-scene analogy mapped onto the computation:
# * Each kernel (raw coords, Gaussian bumps, cosines at 5 bandwidths) is a CLASS of
# "activity location" (home / family / prior o... |
svg_ar_kimi_45214125 | 0.202041 | train | def rank_svg(adata):
# MVDR / Capon beamforming transplanted to spatial transcriptomics.
#
# The tissue is treated as a sensor array: each sample is an element at a known
# 2-D position, and each feature's expression field is one snapshot across the
# array. For every candidate spatial wave (wavenum... |
End of preview.
BVSR data
Paper: Blind Variation, Selective Retention: Improving Creative Idea Generation in AI Scientists
Code: https://github.com/zou-group/BVSR
For the selective-retention selector, see Xkev/SR-judger.
Contents
bv_results/: verified scores and creativity labels of the blind-variation and prompting-baseline proposals on 12 scientific tasks, and the information content of each task context (Figure 2, Table 1).sr_data/: training and evaluation sets of selective retention, with the rankings of every compared method (Figure 3).bvsr_simpletes/: SimpleTES and SimpleTES + BVSR runs on five tasks (Figure 4).
Usage
bash reproduce.sh in the code repository downloads this dataset to data/ and reproduces Figures 2-4 and Table 1.
License
MIT.
- Downloads last month
- 43