id
int64
1
546k
paper_arxiv_id
stringlengths
10
13
path
stringlengths
1
249
caption
stringlengths
9
170k
label
stringlengths
6
192
name
float64
199
2508.00721
fig/visualization/gd-afhq-flowdps.png
\caption{FlowDPS}
null
null
200
2508.00721
fig/visualization/gd-afhq-dflow.png
\caption{D-Flow}
null
null
201
2508.00721
fig/visualization/gd-afhq-fmplug-w.png
\caption{FMPlug-W}
null
null
202
2508.00721
fig/visualization/gd-afhq-fmplug-w-r.png
\caption{FMPlug-W-R}
null
null
203
2508.00721
fig/visualization/gd-div-measurement.png
\caption{Measurement}
null
null
204
2508.00721
fig/visualization/gd-div-gt.png
\caption{Ground Truth}
null
null
205
2508.00721
fig/visualization/gd-div-dip.png
\caption{DIP}
null
null
206
2508.00721
fig/visualization/gd-div-flowchef.png
\caption{FlowChef}
null
null
207
2508.00721
fig/visualization/gd-div-flowdps.png
\caption{FlowDPS}
null
null
208
2508.00721
fig/visualization/gd-div-dflow.png
\caption{D-Flow}
null
null
209
2508.00721
fig/visualization/gd-div-fmplug-w.png
\caption{FMPlug-W}
null
null
210
2508.00721
fig/visualization/gd-div-fmplug-r.png
\caption{FMPlug-W-R}
null
null
211
2508.00721
fig/visualization/gd-div-measurement.png
\caption{Measurement}
null
null
212
2508.00721
fig/visualization/gd-div-gt.png
\caption{Ground Truth}
null
null
213
2508.00721
fig/visualization/gd-div-dip.png
\caption{DIP}
null
null
214
2508.00721
fig/visualization/gd-div-flowchef.png
\caption{FlowChef}
null
null
215
2508.00721
fig/visualization/gd-div-flowdps.png
\caption{FlowDPS}
null
null
216
2508.00721
fig/visualization/gd-div-dflow.png
\caption{D-Flow}
null
null
217
2508.00721
fig/visualization/gd-div-fmplug-w.png
\caption{FMPlug-W}
null
null
218
2508.00721
fig/visualization/gd-div-fmplug-r.png
\caption{FMPlug-W-R}
null
null
219
2508.00700
images/architecture_server.pdf
\caption{Overview of the experimentation architecture.}
\label{fig:architecture}
null
220
2508.00700
images/ResearchMethodOverview-llmq-workflow.drawio.pdf
\caption{Overview of the experimentation workflow.}
\label{fig:method}
null
221
2508.00700
images/questionSolution.jpg
\caption{Question (Problem) and Answer (Solution).}
\label{fig:questionAnswer}
null
222
2508.00700
images/count_issues.png
\caption{Normalized issue \% per dataset. }
\label{fig:count_issue}
null
223
2508.00700
images/newFigure5_all.pdf
\caption{Normalized issues per dataset. }
\label{fig:severity_distribution}
null
224
2508.00700
images/effort_min.png
\caption{Average Effort in Minutes to Fix Bug or Code Smell.}
\label{fig:effort_average}
null
409
2508.00380
figs/vis/vis_toy_main_1-2.pdf
null
null
null
225
2508.00288
fig1_uav-on_outline.pdf
\caption{ Task definition in the UAV-ON benchmark. The aerial agent receives a semantic instruction and is required to navigate in a 3D environment to locate the described target. }
\label{fig:overview}
null
226
2508.00288
fig2_prompt_statistics.pdf
\caption{Statistics of target semantic information. (a) Word cloud of object category names. (b) Word cloud of descriptive attributes. (c) Distribution of target object sizes.}
\label{fig:wordcloud}
null
227
2508.00288
fig3_onair_outline.pdf
\caption{System overview of the Aerial ObjectNav Agent (AOA). Multi-view RGB-D observations, target information, and pose history are encoded into a structured input and processed by LLM to predict UAV actions.}
\label{fig:baseline}
null
228
2508.00288
apdx1_scene_object_prompt.pdf
\caption{ (a) Task prompt used for scene-object co-occurrence inference. (b) Co-occurrence mapping visualized as a Sankey diagram linking scenes to likely object categories.}
\label{fig:scene_object}
null
229
2508.00288
apdx2_gpt_description.pdf
\caption{Illustration of the object prompt generation process. Multi-view images are used to generate an instance-level object prompt via GPT-based object description and size estimation.}
\label{fig:gpt_description}
null
230
2508.00288
apdx3_max_move_count.pdf
\caption{Geometric basis for determining the maximum exploration step limit. (a) Visualization of the UAV’s downward-facing FOV at 5 meters altitude; (b) Grid-based search region constrained within a 50-unit radius circle centered at the initialization point; (c) Zig-zag traversal policy covering all 88 valid grid cells.}
\label{fig:move_count}
null
231
2508.00288
apdx4_method_trajectories.pdf
\caption{Representative 3D trajectories of four baseline methods across three outcome types: \textit{Failure/Collision} (top row), \textit{Oracle Success} (middle row), and \textit{Success} (bottom row). Methods include: (a) Random, (b) CLIP-H, (c) AOA-V, and (d) AOA-F. All examples are drawn from the same scene (\textit{BrushifyUrban}) for comparability.}
\label{fig:baseline_trajectories}
null
232
2508.00668
generalizedCM.jpg
\caption{Generalized conjecture map for educational design research (Sandoval, 2014)}
\label{fig:conjecture}
null
233
2508.00668
conjecture_initial.jpg
\caption{Initial conjecture map of a design to promote argumentation in elementary science (Sandoval, 2014)}
\label{fig:conjecture2}
null
234
2508.00725
Scalar_EFT.pdf
\caption{SN~1987A cooling constraints for scalar (left) and vector (right) interactions compared to DD searches by PANDA4T-X and prospects from DARWIN. The dotted line covers the estimated uncertainty from using different SN models. Other upper bounds include cosmological overproduction (see text and SupM for details) and those from ID derived in Ref.~\cite{Ge:2022ius}. \label{fig:money}}
null
null
235
2508.00725
Scalar_light.pdf
\caption{Bounds on the DD cross-section for the absorption of DM with $m_\chi=20$ keV and sub-GeV scalar mediator. For $y_e=y_\chi$, the SN~1987A limits from SFHo18.80 are compared to the limit from DM overproduction %using the same interaction and sensitivity limits achieved by PANDAX-4T and the prospects for DARWIN. We also show the SN~1987A bound obtained using $y_\chi=10^5 y_e$ or vice-versa. For such hierarchical couplings the overproduction bound would only get stronger. }
\label{fig:light_scalar}
null
236
2508.00725
ScalarR_EFT.pdf
\caption{SN~1987A cooling bound for scalar (left) and vector (right) interactions and sterile neutrinos. See caption of Fig.~\ref{fig:money}. \label{fig:moneyR}}
null
null
237
2508.00725
light_vectorR.pdf
\caption{Bounds on the DD cross-section for the absorption of DM with $m_\chi=100$ keV, a sub-GeV vector mediator and a sterile neutrino. See caption of Fig.~\ref{fig:light_scalar}. }
\label{fig:light_vector}
null
238
2508.00587
figures/colorbar.pdf
null
null
null
239
2508.00587
figures/teaser.png
null
null
null
240
2508.00587
figures/teaser2.png
null
null
null
241
2508.00587
figures/teaser3.png
null
null
null
242
2508.00587
figures/teaser5.png
null
null
null
243
2508.00587
figures/teaser8.png
null
null
null
244
2508.00587
figures/teaser6.png
null
null
null
245
2508.00587
figures/teaser7.png
null
null
null
246
2508.00587
figures/teaser4.png
null
null
null
247
2508.00587
figures/teaser.png
null
null
null
248
2508.00587
figures/teaser2.png
null
null
null
249
2508.00587
figures/teaser3.png
null
null
null
250
2508.00587
figures/teaser5.png
null
null
null
251
2508.00587
figures/teaser8.png
null
null
null
252
2508.00587
figures/teaser6.png
null
null
null
253
2508.00587
figures/teaser7.png
null
null
null
254
2508.00587
figures/teaser4.png
null
null
null
255
2508.00587
figures/combined.pdf
\caption{\textbf{Visualisation of univariate Gaussian classification.} \textbf{Top:} Comparison between an evidential classifier (EDL) and a standard binary classifier trained with binary cross-entropy (BCE), with corresponding uncertainty measures, vacuity and entropy, respectively. Shaded blue area indicates the training data density. \textbf{Bottom:} Predicted Dirichlet distributions $\text{Dir}(p|\boldsymbol{\alpha})$ at $x \in \{-6, -0.4, 0, 0.4, 6 \}$. The vertical dashed lines indicate the mean probability $p_{\theta}^{\text{EDL}}(y^{\text{out}}=1|x)$.}
\label{fig:exp:univariate_gaussian}
null
256
2508.00587
figures/layer_ablation_study.pdf
\caption{\textbf{DINOv2 layer ablation study.} Comparison of models trained using features from different layers $l \in \{1, 6, 12, 16, 21, 24\}$ of DINOv2. Performance is reported in terms of average precision ($\uparrow$) and false positive rate ($\downarrow$) on the SMIYC Anomaly Track validation set.}
\label{fig:exp:ablation_study}
null
257
2508.00587
figures/Static_min_dist.pdf
\caption{\textbf{Visualisation of overconfident extrapolation.} Comparison between an evidential classifier (EDL) and a standard binary classifier (BCE) on the Fishyscapes Static dataset. The plot shows the average probability $\bar{p}_{\theta}(y^{\text{out}}=1|\mathbf{x}_i)$ over the cosine distance $d_{\mathrm{cos}}(\boldsymbol{\mu}_k, \mathbf{x}_i)$ between a feature vector $\mathbf{x}_i$ and the nearest class mean feature vector $\boldsymbol{\mu}_k$ from the training set. For reference, shaded blue and red areas indicate the density of distances for \gls{id} and \gls{ood} features, respectively.}
\label{fig:exp:overconfidence}
null
258
2508.00607
Figures/Figure1.eps
\caption{Five-level system at Raman resonance ($\delta\omega=6B_X$): On the left, energy levels in the rotating wave frame with couplings from lasers $L_1$ (blue) and $L_2$ (pink). On the right, dressed energy levels after diagonalizing the asymptotic Hamiltonian \eqref{eq:H5States-2}. Horizontal dashed lines mark the energy origin in both panels.}
\label{fig:5-level-system}
null
259
2508.00607
Figures/Figure2.eps
\caption{Dressed potential-energy curves for $\Omega_1 = 2\pi\times 50$~MHz, $\Omega_2 = 2\pi\times 200$~MHz, $\Delta = 2\pi\times 1$~GHz. Each panel corresponds to a different energy scale, with progressive zoom-ins from the upper to the lower panel.}
\label{dressed_PECs}
null
410
2508.00380
figs/vis/vis_toy_main_1-3.pdf
null
null
null
260
2508.00607
Figures/Figure3.eps
\caption{Adiabatic elimination: dressed PECs calculated by diagonalizing the full Hamiltonian (Eq.(\ref{H_total}), solid black lines), and the effective Hamiltonian (Eq. \ref{effective_hamiltonian_5level_sys}, in dashed red lines), for laser parameters $\Omega_1 = 2\pi\times 50$~MHz, $\Omega_2 = 2\pi\times 200$~MHz, and $\Delta = 2\pi\times 1$~GHz. The upper panel shows the PECs on a scale of 1400~MHz including the excited state manifold, and the lower panel shows a zoom-in on a scale of 11~MHz around the zero energy. \label{adiab_elimination_fig} }
null
null
261
2508.00607
Figures/Figure4.eps
\caption{Invalid adiabatic elimination: same as Figure \ref{adiab_elimination_fig}, but for $\Delta = 2\pi\times 300$~MHz. \label{adiab_not_elim} }
null
null
262
2508.00607
Figures/Figure5.eps
\caption{Dressed PECs in the energy range of the entrance channel (marked with a red arrow) for $\Omega_1 = 2\pi\times 50$~MHz, $\Omega_2 = 2\pi\times 200$~MHz, and $\Delta = 2\pi\times 1$~GHz at Raman resonance ($\hbar \delta\omega = 6B_X$). The main components of each channel are listed on the right, with colors corresponding to $(|m_1|, |m_2|)$ subspaces: red for $(0,0)$, yellow for $(0,1)$, green for $(0,2)$, blue for $(1,1)$ and $(1,2)$, and purple for $(2,2)$ (see Appendix for a detailed discussion).}
\label{entrance_states_figure}
null
263
2508.00607
Figures/Figure6.eps
\caption{Collision rate coefficients (in cm$^3$/s) of two ultracold $^{23}$Na$^{39}$K molecules exposed to two lasers, as function of their Rabi frequencies $\Omega_1$ and $\Omega_2$ for a detuning $\Delta = 2\pi\times 1$~GHz, at the Raman resonance ($\delta\omega = 0$), and for a temperature of $T=300$~nK: (a) elastic, (b) inelastic and (c) reactive rates.}
\label{fig:rates2-os}
null
264
2508.00607
Figures/Figure7.eps
\caption{The coupling scheme of two molecules exposed to two linearly polarized $\pi$ laser fields. Solid red arrows represent the five-level coupling scheme in the $(|m_1|,|m_2|)=(0,0)$ subspace. Solid yellow arrows represent the four-level scheme in the $(|m_1|,|m_2|)=(0,1)$ subspace. Solid green arrows represent the three-level system, and a single solid yellow arrow represents a two-level scheme in the $(|m_1|,|m_2|)=(0,2)$ subspace. Solid blue arrows represent the two-level schemes in the $(|m_1|,|m_2|)=(1,1)$ and $(|m_1|,|m_2|)=(1,2)$ subspaces. The purple dot corresponds to the state that is not coupled to lasers in the $(|m_1|,|m_2|)=(2,2)$ subspace.}
\label{full_level_scheme_compact}
null
265
2508.00607
Figures/Figure8.eps
\caption{The six coupling schemes for the different $(|m_1|+|m_2|)$ values in the rotating wave frame in the case of two linearly polarized lasers ($\Delta m_1=0$, $\Delta m_2=0$). The inner five rectangles of each panel correspond to the different $j_1+j_2$ values. The arrows represent the laser couplings between the different $(|m_1|,|m_2|)$ levels. Panels a, b, c, d, e, and f represent the five-, four-, three-level systems, single field coupling, and the uncoupled state, respectively.}
\label{full_level_scheme}
null
266
2508.05616v1
old_figures/catchy_figure4.pdf
\caption{Motivation for our \our{} framework. Traditional manual heuristic design (left) relies on human expertise and trial-and-error processes. Deep learning approaches (center) produce more accurate predictions but demand substantial computational resources, yield black-box models, and often struggle with generalization. \our{} (right) automates the heuristic design process using evolutionary algorithms, enabling the generation of novel and effective trajectory prediction heuristics.}
\label{fig:motivation}
null
267
2508.05616v1
new_figures/overview.pdf
\caption{An illustration of the \our{} evolutionary process. \textbf{(Left)} The framework continuously discovers and evaluates a variety of heuristic strategies, such as those based on constant velocity or social forces. \textbf{(Middle)} As the evolution progresses, the performance of these heuristics steadily improves, evidenced by the decreasing objective value from a simple seed to a final, optimized solution. \textbf{(Right)} The overall \our{} pipeline orchestrates this entire discovery and optimization process, using an LLM-driven evolutionary algorithm to automatically generate and refine the heuristics.}
\label{fig:trajevo-catchy}
null
268
2508.05616v1
new_figures/cges.pdf
\caption{Cross-Generation Elite Sampling (CGES) helps escape local optima by sampling elite individuals from past generations (left), which greatly helps achieve much better objective values (right).}
\label{fig:cross-generation-elite-sampling}
null
269
2508.05616v1
old_figures/f1-cropped.pdf
null
\label{fig:stats_analysis}
null
270
2508.05616v1
old_figures/f2-cropped.pdf
null
\label{fig:stats_refinement}
null
271
2508.05616v1
old_figures/f1-cropped.pdf
null
\label{fig:stats_analysis}
null
272
2508.05616v1
old_figures/f2-cropped.pdf
null
\label{fig:stats_refinement}
null
273
2508.05616v1
old_figures/final_with_legend_aaai.pdf
\caption{Trajectory prediction results for CVM-S \citep{scholler2020constant} and \our{} across different datasets. Each row illustrates a distinct pattern: (top) linear trajectories, (middle) non-linear trajectories, and (bottom) collision-avoidance cases. We visualize the single best trajectory out of $K=20$ samples based on the optimization objective.}
\label{fig:generated-trajectories-image}
null
274
2508.05616v1
old_figures/main_figure.pdf
\caption{A qualitative illustration of the \our{} evolutionary process. The optimization begins with a simple seed heuristic and iteratively discovers more complex and effective strategies by exploring different concepts like acceleration with noise, social repulsion, and weighted velocity averaging. The process converges on a final, highly-optimized heuristic that combines multiple discovered strategies, achieving a significantly lower objective value.}
\label{fig:appendix_evo_example}
null
275
2508.05616v1
new_figures/performance_curve.pdf
\caption{Performance sensitivity to the exploration ratio for crossover parent selection. The x-axis represents the proportion of parents selected uniformly at random (exploration), while the y-axis shows the resulting MSE score (lower is better). The optimal performance is achieved at an exploration ratio of 0.7, effectively balancing exploration and exploitation.}
\label{fig:exploration_ratio}
null
276
2508.05616v1
new_figures/evolution_resources_comparison.pdf
\caption{ \textbf{Evolution Resources Comparison.} Compared to SOTA neural network methods, \our{} shows a significant advantage in both evolution time and cost (e.g., 5 minutes vs. 1 day, ~$0.05 vs. ~$4.00). Both metrics are displayed on a logarithmic scale. }
\label{fig:evo_resources}
null
277
2508.05616v1
new_figures/inference_resources_comparison.pdf
\caption{ \textbf{Inference Resources Comparison.} \our{} achieves a fast 0.65ms inference time on a single CPU core, while baseline methods require more time even on a GPU. All times are displayed on a logarithmic scale. }
\label{fig:inf_resources}
null
278
2508.05616v1
old_figures/selection_ablation.pdf
\caption{Visual comparison of the evolutionary process with and without CGES. The plot shows the objective value (lower is better) over the number of function evaluations, averaged over multiple runs. The \our{} framework with CGES consistently converges to better solutions than the ablated version, highlighting the effectiveness of CGES in improving the search process.}
\label{fig:cges_ablation_appendix}
null
279
2508.00466
H_rare_decays_cartoon.png
\caption{Schematic diagrams of rare and exclusive two- and three-body decays of the Higgs boson into (i) two or three gauge bosons (V = Z, W, $\gamma$) or into a gauge boson plus a neutrino pair $(\nunubar$) through virtual loops (grey circle) (ii) a gauge boson %(mostly a photon) plus a difermion bound state (meson or leptonium), and (iii) two onium states.\label{fig:rare_decays_diags}}
null
null
280
2508.00466
rare_H_2_3_4body_decays.png
\caption{Representative diagrams of rare 2-, 3-, and 4-body decays of the H boson into photons and/or neutrinos, and into Z bosons plus gluons or photons.}
\label{fig:H_rare}
null
281
2508.00466
plots/H_3_4body_rare_decays.pdf
\caption{Theoretical branching fractions of the Higgs boson into rare two-, three-, or four- gauge bosons and/or neutrinos shown as a function of the Higgs boson mass (left) and as blue bars in negative log scale (right). In the left panel, the dashed lines for $m_\mathrm{H}<m_\mathrm{Z}$ show the $\rm H \to Z^*gg,\,Z^*\gaga$ decays with offshell Z bosons. In the right panel, the red vertical line indicates the minimum $\BR$ value reachable at the HL-LHC given just by the total number of H bosons expected to be produced. \label{fig:H_rare_BR}}
null
null
282
2508.00466
excl_H_decays_gamma_meson_diags.png
\caption{Schematic diagrams of exclusive decays of the H boson into a meson plus a gauge boson in direct (left), indirect (center), and W-loop (right) processes. The solid fermion lines represent quarks, and the gray blob represents the mesonic bound state.}
\label{fig:H_gauge_meson_diags}
null
283
2508.00466
plots/H_gamma_meson_limits.pdf
\caption{Branching ratios (in negative log scale) of exclusive $\rm H \to \gamma +vector$-meson decays. Most recent theoretical predictions (blue bars) compared to current experimental limits (violet) and expected conservative HL-LHC bounds (orange). The red vertical line indicates the minimum $\BR$ value reachable at the HL-LHC given just by the total number of H bosons expected to be produced. }
\label{fig:H_gamma_meson_limits}
null
284
2508.00466
plots/H_Z_meson_limits.pdf
\caption{Branching ratios (in negative log scale) of exclusive $\rm H \to Z\,+\, meson$ decays. The most recent theoretical predictions (blue bars) are compared to current experimental limits (violet) and conservatively expected HL-LHC bounds (orange). The red vertical line indicates the minimum $\BR$ value reachable at the HL-LHC given just by the total number of H bosons expected to be produced.}
\label{fig:H_Z_meson_limits}
null
285
2508.00466
Higgs_FCNC_cartoon1.png
\caption{Left: Schematic diagram of exclusive flavour-changing Higgs decays into a meson plus a photon or a Z boson. %, $\rm H \to \gamma+M,\,Z+M$. The solid fermion lines represent quarks, the wavy ones bosons, and the gray blob a neutral meson. Right: Branching fractions (in negative log scale) of exclusive $\rm H\to\gamma+VM^*,\,Z+VM^*$ decays, where $\rm VM^*$ are excited flavoured neutral mesons. The SM predictions (blue bars) from Eqs.~(\ref{eq:BR_H_gamma_m}) and (\ref{eq:BR_H_Z_m}) are compared with current experimental limits (violet) and expected conservative HL-LHC bounds (orange). The red vertical line indicates the minimum $\BR$ value reachable at the HL-LHC given just by the total number of H bosons expected to be produced.}
\label{fig:H_boson_flavouredmeson_limits}
null
286
2508.00466
plots/H_W_meson_limits.pdf
\caption{Theoretical branching fractions (blue bars, in negative log scale) of exclusive $\rm H \to W^{\pm} + meson$ decays. No experimental limits exist to date. The red vertical line indicates the minimum $\BR$ value reachable at the HL-LHC given just by the total number of H bosons expected to be produced.}
\label{fig:H_W_meson_limits}
null
287
2508.00466
excl_H_decays_gamma_leptonium_diags.png
\caption{Schematic diagrams of exclusive H boson decays into a photon or a Z boson plus a leptonium state in the direct (left) and indirect (right) channels. The solid fermion lines represent leptons, the gray blob represents the $(\lele)$ bound state.}
\label{fig:H_leptonium_diags}
null
288
2508.00466
plots/H_boson_leptonium_limits.pdf
\caption{Theoretical branching fractions (blue bars, in negative log scale) of exclusive $\rm H \to \gamma,Z + leptonium$ decays. No experimental limits exist to date. The red vertical line indicates the minimum $\BR$ value reachable at the HL-LHC given just by the total number of H bosons expected to be produced.}
\label{fig:H_boson_leptonium_limits}
null
289
2508.00466
excl_H_decays_meson_meson_diags.png
\caption{Schematic diagrams of exclusive decays of the H boson into two mesons. The wavy lines indicate gauge bosons, the solid fermion lines represent quarks and the gray blobs are the meson bound states.}
\label{fig:H_2meson_diags}
null
290
2508.00466
plots/H_meson_meson_limits.pdf
\caption{Branching ratios (in negative log scale) of exclusive $\rm H \to meson + meson$ decays. Most recent theoretical predictions (blue bars) are compared to current experimental limits (violet) and expected conservative HL-LHC bounds (orange). The red vertical line indicates the minimum $\BR$ value reachable at the HL-LHC given just by the total number of H bosons expected to be produced.}
\label{fig:H_2meson_limits}
null
291
2508.00667
sel_eff_0511.png
\caption{Distribution of the selected ICGs in the $L_\mathrm{c},L_\mathrm{gap}$ plane, where $L_\mathrm{c}$ represents the luminosity of the ICG and $L_\mathrm{gap}$ denotes the luminosity ratio between the ICG and its brightest satellite galaxy (or the SBG of the system). The color bar indicates the probability density of ICGs. Brown dash-dotted lines show the boundaries above which the SBGs are too faint to be observed at the corresponding redshift as labelled. Black dashed lines mark the boundaries of the eight $L_\mathrm{c},L_\mathrm{gap}$ bins used for stacked lensing analysis.}
\label{fig:sel_eff}
null
292
2508.00667
obs_z_dist.png
\caption{The redshift distribution of the observed ICG sample. The hatched region indicates that ICGs with $z\leq 0.03$ are removed from the lens sample for higher lensing efficiency.}
\label{fig:obs_z_dist}
null
293
2508.00667
ESD_obs_bestfit_0624.png
\caption{The measured lensing ESD profiles centered on ICGs in different bins of $L_\mathrm{c}$ (see the text in each panel) and $L_\mathrm{gap}$ (different color symbols in the same panel, see the legend). The bins are defined in Figure~\ref{fig:sel_eff}. The $x$-axis is the projected physical radius to the ICG. The symbols and associated errorbars show the estimated ESDs with the PDF-SYM method and their $1\sigma$ uncertainty estimated from 90 jackknife subsamples. The solid lines with corresponding colors show the best-fit ESD profiles within $[0.06-2.3R_\mathrm{200b}]$ based on the NFW model profile (see Section~\ref{sec:spf}).}
\label{fig:ESD-data}
null
294
2508.00667
logM_est_bias_2.png
\caption{Top: The bias of the lensing estimated $\log \hat{M}$ from the mean $\log M_\mathrm{200b}$ in the lightcone sample. Crosses with error bars show the average and $1\sigma$ scatter of $\Delta \log M$ in 100 lightcone samples in each bin of $L_\mathrm{c}$ and $L_\mathrm{gap}$. Bottom: The above bias relative to the mean $\log M_\mathrm{200b}$ in the lightcone sample. The solid lines together with the shaded regions show the mean and $1\sigma$ scatter of the relative bias in 100 lightcone samples. Colors represent different bins of $L_\mathrm{gap}$, as labeled in the legend.}
\label{fig:logM_est_bias}
null
5,734
2508.00687
2x2_1orient
null
null
null
296
2508.00667
logM_before_n_after_calibration_0624.png
\caption{Best fit halo mass ($\log \hat{M}$) from measured ESD profiles shown by the unfilled circles, compared with the bias-calibrated mean estimator of $\log M_\mathrm{200b}$, $\log \hat{M}_\mathrm{bc}$, marked by filled circles in each bin of $L_\mathrm{c}$ and $L_\mathrm{gap}$. The errorbars of the unfilled circles represent the $1\sigma$ uncertainty of the lensing best fit. The errorbars of the filled circles demonstrate the $1\sigma$ uncertainty of the mean estimator, which is a combination of uncertainties from both the lensing best fit and the bias estimator. The $x$ values show the mean $\log L_\mathrm{c}$ in each bin. For visual clarity, the lensing best fits are shifted left by 0.05 dex.}
\label{fig:logM_before_n_after_calibration}
null