analiza aiqbench
This research advances the theoretical robustness of split conformal prediction by quantifying how training and calibration contamination affects interval length under Huber models. By establishing matching finite-sample upper and minimax lower bounds, the work provides rigorous guarantees for uncertainty quantification in non-ideal data environments. The findings are particularly relevant for practitioners deploying conformal methods where data integrity cannot be assumed.
Split Conformal Interval Length under Huber Contamination
arXiv:2610.15458v1 Announce Type: new Abstract: Existing analyses of interval length in split conformal prediction largely focus on clean data. However, how contamination at the training and calibration stages alters interval length relative to the corresponding clean population oracle remains understudied. In this paper, we consider Huber contamination and derive a finite-sample upper bound that holds uniformly over contaminating distributions, with leading terms scaling as $L(\delta_1+n_1^{-1/2}) +(n_2^{-1/2}+\delta_2)/((1-\delta_2)r_0)$, in addition to a clean-sample estimation term. We further establish a minimax lower bound matching the upper-bound rate up to constant factors on a regular asymmetric location submodel. For calibration contamination, we obtain an exact worst-case characterization of the population interval-length deviation. We also extend the upper-bound results to conformalized quantile regression and adaptive residual scores. Experiments on synthetic and real-world datasets provide empirical support for our theoretical findings on training and calibration contamination.
Abstract:Existing analyses of interval length in split conformal prediction largely focus on clean data. However, how contamination at the training and calibration stages alters interval length relative to the corresponding clean population oracle remains understudied. In this paper, we consider Huber contamination and derive a finite-sample upper bound that holds uniformly over contaminating distributions, with leading terms scaling as $L(\delta_1+n_1^{-1/2}) +(n_2^{-1/2}+\delta_2)/((1-\delta_2)r_0)$, in addition to a clean-sample estimation term. We further establish a minimax lower bound matching the upper-bound rate up to constant factors on a regular asymmetric location submodel. For calibration contamination, we obtain an exact worst-case characterization of the population interval-length deviation. We also extend the upper-bound results to conformalized quantile regression and adaptive residual scores. Experiments on synthetic and real-world datasets provide empirical support for our theoretical findings on training and calibration contamination.
Machine Learning (cs.LG)
arXiv:2610.15458 [cs.LG]
(or arXiv:2610.15458v1 [cs.LG] for this version)
https://doi.org/10.48550/arXiv.2610.15458 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
experiment_code/MANIFEST.json
experiment_code/README.md
experiment_code/config/experiment.json
experiment_code/config/input_data.json
experiment_code/config/plotting.json
experiment_code/config/synthetic_population_oracle.json
experiment_code/data_generation/core.py
experiment_code/data_generation/fetch_real.py
experiment_code/data_generation/replacement.py
experiment_code/data_generation/utils.py
experiment_code/environment.json
experiment_code/experiments/aggregate.py
experiment_code/experiments/compute_population_oracle.py
experiment_code/experiments/conformal.py
experiment_code/experiments/density.py
experiment_code/experiments/engine.py
experiment_code/experiments/fixed.py
experiment_code/experiments/fixed_io.py
experiment_code/experiments/learner.py
experiment_code/experiments/metrics.py
experiment_code/experiments/nn_learner.py
experiment_code/experiments/population_oracle.py
experiment_code/experiments/protocol.py
experiment_code/experiments/real.py
experiment_code/experiments/replay.py
experiment_code/experiments/run.py
experiment_code/experiments/score_engine.py
experiment_code/experiments/synthetic.py
experiment_code/output/paper/Figure_1.pdf
experiment_code/output/paper/Figure_1.png
experiment_code/output/paper/Figure_10.pdf
experiment_code/output/paper/Figure_10.png
experiment_code/output/paper/Figure_11.pdf
experiment_code/output/paper/Figure_11.png
experiment_code/output/paper/Figure_12.pdf
experiment_code/output/paper/Figure_12.png
experiment_code/output/paper/Figure_13.pdf
experiment_code/output/paper/Figure_13.png
experiment_code/output/paper/Figure_14.pdf
experiment_code/output/paper/Figure_14.png
experiment_code/output/paper/Figure_15a.pdf
experiment_code/output/paper/Figure_15a.png
experiment_code/output/paper/Figure_15b.pdf
experiment_code/output/paper/Figure_15b.png
experiment_code/output/paper/Figure_16.pdf
experiment_code/output/paper/Figure_16.png
experiment_code/output/paper/Figure_2.pdf
experiment_code/output/paper/Figure_2.png
experiment_code/output/paper/Figure_3a.pdf
experiment_code/output/paper/Figure_3a.png
experiment_code/output/paper/Figure_3b.pdf
experiment_code/output/paper/Figure_3b.png
experiment_code/output/paper/Figure_4a.pdf
experiment_code/output/paper/Figure_4a.png
experiment_code/output/paper/Figure_4b.pdf
experiment_code/output/paper/Figure_5.pdf
experiment_code/output/paper/Figure_5.png
experiment_code/output/paper/Figure_6.pdf
experiment_code/output/paper/Figure_6.png
experiment_code/output/paper/Figure_7.pdf
experiment_code/output/paper/Figure_7.png
experiment_code/output/paper/Figure_8.pdf
experiment_code/output/paper/Figure_8.png
experiment_code/output/paper/Figure_9.pdf
experiment_code/output/paper/Figure_9.png
experiment_code/output/paper/figure_index.json
experiment_code/plotting/__init__.py
experiment_code/plotting/contour.py
experiment_code/plotting/contour_triptych.py
experiment_code/plotting/diagnostics.py
experiment_code/plotting/figure_index.py
experiment_code/plotting/figures.py
experiment_code/plotting/fixed_panels.py
experiment_code/plotting/manuscript_layout.py
experiment_code/plotting/panels.py
experiment_code/plotting/paper.py
experiment_code/plotting/placement_style.py
experiment_code/plotting/reference_compat.py
experiment_code/plotting/render.py
experiment_code/plotting/style.py
experiment_code/plotting/surface.py
experiment_code/reference/README.md
experiment_code/reference/checks/fixed/joint/synthetic/rep00/joint.csv
experiment_code/reference/checks/fixed/real/abalone/rep00/correction.csv
experiment_code/reference/checks/fixed/real/abalone/rep00/dirac.csv
experiment_code/reference/checks/fixed/real/abalone/rep00/trends.csv
experiment_code/reference/checks/fixed/real/cadata/rep00/correction.csv
experiment_code/reference/checks/fixed/real/cadata/rep00/dirac.csv
experiment_code/reference/checks/fixed/real/cadata/rep00/trends.csv
experiment_code/reference/checks/fixed/real/cpusmall/rep00/correction.csv
experiment_code/reference/checks/fixed/real/cpusmall/rep00/dirac.csv
experiment_code/reference/checks/fixed/real/cpusmall/rep00/trends.csv
experiment_code/reference/checks/fixed/real/msd/rep00/correction.csv
experiment_code/reference/checks/fixed/real/msd/rep00/dirac.csv
experiment_code/reference/checks/fixed/real/msd/rep00/trends.csv
experiment_code/reference/checks/fixed/synthetic/synthetic/rep00/correction.csv
experiment_code/reference/checks/fixed/synthetic/synthetic/rep00/dirac.csv
experiment_code/reference/checks/fixed/synthetic/synthetic/rep00/trends.csv
experiment_code/reference/checks/manifest.json
experiment_code/reference/checks/original/joint/synthetic/rep00/joint.csv
experiment_code/reference/checks/original/learner/synthetic/rep00/learner.csv
experiment_code/reference/checks/original/real/abalone/rep00/correction.csv
experiment_code/reference/checks/original/real/abalone/rep00/dirac.csv
experiment_code/reference/checks/original/real/abalone/rep00/trends.csv
experiment_code/reference/checks/original/real/cadata/rep00/correction.csv
experiment_code/reference/checks/original/real/cadata/rep00/dirac.csv
experiment_code/reference/checks/original/real/cadata/rep00/trends.csv
experiment_code/reference/checks/original/real/cpusmall/rep00/correction.csv
experiment_code/reference/checks/original/real/cpusmall/rep00/dirac.csv
experiment_code/reference/checks/original/real/cpusmall/rep00/trends.csv
experiment_code/reference/checks/original/real/msd/rep00/correction.csv
experiment_code/reference/checks/original/real/msd/rep00/dirac.csv
experiment_code/reference/checks/original/real/msd/rep00/trends.csv
experiment_code/reference/checks/original/synthetic/synthetic/rep00/correction.csv
experiment_code/reference/checks/original/synthetic/synthetic/rep00/dirac.csv
experiment_code/reference/checks/original/synthetic/synthetic/rep00/trends.csv
experiment_code/reference/fixed/results/correction_summary.csv
experiment_code/reference/fixed/results/density_summary.csv
experiment_code/reference/fixed/results/dirac_summary.csv
experiment_code/reference/fixed/results/joint_summary.csv
experiment_code/reference/fixed/results/learner_slopes.csv
experiment_code/reference/fixed/results/learner_summary.csv
experiment_code/reference/fixed/results/trends_summary.csv
experiment_code/reference/original/results/correction_summary.csv
experiment_code/reference/original/results/density_summary.csv
experiment_code/reference/original/results/dirac_summary.csv
experiment_code/reference/original/results/joint_summary.csv
experiment_code/reference/original/results/learner_slopes.csv
experiment_code/reference/original/results/learner_summary.csv
experiment_code/reference/original/results/trends_summary.csv
experiment_code/reference/original/runs/real/abalone/rep00/baseline.npy
experiment_code/reference/original/runs/real/abalone/rep00/clean.npy
experiment_code/reference/original/runs/real/abalone/rep00/correction_candidate.npy
experiment_code/reference/original/runs/real/abalone/rep00/correction_mask_u.npy
experiment_code/reference/original/runs/real/abalone/rep00/dirac_mask_u.npy
experiment_code/reference/original/runs/real/abalone/rep00/dirac_point_x.npy
experiment_code/reference/original/runs/real/abalone/rep00/dirac_point_y.npy
experiment_code/reference/original/runs/real/abalone/rep00/dirac_score.npy
experiment_code/reference/original/runs/real/abalone/rep00/heteroscedastic_0_scores.npy
experiment_code/reference/original/runs/real/abalone/rep00/heteroscedastic_1_scores.npy
experiment_code/reference/original/runs/real/abalone/rep00/heteroscedastic_2_scores.npy
experiment_code/reference/original/runs/real/abalone/rep00/heteroscedastic_3_scores.npy
experiment_code/reference/original/runs/real/abalone/rep00/heteroscedastic_mask_u.npy
experiment_code/reference/original/runs/real/abalone/rep00/shift_0_scores.npy
experiment_code/reference/original/runs/real/abalone/rep00/shift_1_scores.npy
experiment_code/reference/original/runs/real/abalone/rep00/shift_2_scores.npy
experiment_code/reference/original/runs/real/abalone/rep00/shift_3_scores.npy
experiment_code/reference/original/runs/real/abalone/rep00/shift_mask_u.npy
experiment_code/reference/original/runs/real/abalone/rep00/t_noise_0_scores.npy
experiment_code/reference/original/runs/real/abalone/rep00/t_noise_1_scores.npy
experiment_code/reference/original/runs/real/abalone/rep00/t_noise_2_scores.npy
experiment_code/reference/original/runs/real/abalone/rep00/t_noise_3_scores.npy
experiment_code/reference/original/runs/real/abalone/rep00/t_noise_mask_u.npy
experiment_code/reference/original/runs/real/abalone/rep00/target_scale.npy
experiment_code/reference/original/runs/real/abalone/rep00/test_mu.npy
experiment_code/reference/original/runs/real/abalone/rep00/test_q_hi.npy
experiment_code/reference/original/runs/real/abalone/rep00/test_q_lo.npy
experiment_code/reference/original/runs/real/abalone/rep00/test_sigma.npy
experiment_code/reference/original/runs/real/abalone/rep00/test_y_test.npy
experiment_code/reference/original/runs/real/abalone/rep00/variance_inflate_0_scores.npy
experiment_code/reference/original/runs/real/abalone/rep00/variance_inflate_1_scores.npy
experiment_code/reference/original/runs/real/abalone/rep00/variance_inflate_2_scores.npy
experiment_code/reference/original/runs/real/abalone/rep00/variance_inflate_3_scores.npy
experiment_code/reference/original/runs/real/abalone/rep00/variance_inflate_mask_u.npy
experiment_code/reference/original/runs/real/cadata/rep00/baseline.npy
experiment_code/reference/original/runs/real/cadata/rep00/clean.npy
experiment_code/reference/original/runs/real/cadata/rep00/correction_candidate.npy
experiment_code/reference/original/runs/real/cadata/rep00/correction_mask_u.npy
experiment_code/reference/original/runs/real/cadata/rep00/dirac_mask_u.npy
experiment_code/reference/original/runs/real/cadata/rep00/dirac_point_x.npy
experiment_code/reference/original/runs/real/cadata/rep00/dirac_point_y.npy
experiment_code/reference/original/runs/real/cadata/rep00/dirac_score.npy
experiment_code/reference/original/runs/real/cadata/rep00/heteroscedastic_0_scores.npy
experiment_code/reference/original/runs/real/cadata/rep00/heteroscedastic_1_scores.npy
experiment_code/reference/original/runs/real/cadata/rep00/heteroscedastic_2_scores.npy
experiment_code/reference/original/runs/real/cadata/rep00/heteroscedastic_3_scores.npy
experiment_code/reference/original/runs/real/cadata/rep00/heteroscedastic_mask_u.npy
experiment_code/reference/original/runs/real/cadata/rep00/shift_0_scores.npy
experiment_code/reference/original/runs/real/cadata/rep00/shift_1_scores.npy
experiment_code/reference/original/runs/real/cadata/rep00/shift_2_scores.npy
experiment_code/reference/original/runs/real/cadata/rep00/shift_3_scores.npy
experiment_code/reference/original/runs/real/cadata/rep00/shift_mask_u.npy
experiment_code/reference/original/runs/real/cadata/rep00/t_noise_0_scores.npy
experiment_code/reference/original/runs/real/cadata/rep00/t_noise_1_scores.npy
experiment_code/reference/original/runs/real/cadata/rep00/t_noise_2_scores.npy
experiment_code/reference/original/runs/real/cadata/rep00/t_noise_3_scores.npy
experiment_code/reference/original/runs/real/cadata/rep00/t_noise_mask_u.npy
experiment_code/reference/original/runs/real/cadata/rep00/target_scale.npy
experiment_code/reference/original/runs/real/cadata/rep00/test_mu.npy
experiment_code/reference/original/runs/real/cadata/rep00/test_q_hi.npy
experiment_code/reference/original/runs/real/cadata/rep00/test_q_lo.npy
experiment_code/reference/original/runs/real/cadata/rep00/test_sigma.npy
experiment_code/reference/original/runs/real/cadata/rep00/test_y_test.npy
experiment_code/reference/original/runs/real/cadata/rep00/variance_inflate_0_scores.npy
experiment_code/reference/original/runs/real/cadata/rep00/variance_inflate_1_scores.npy
experiment_code/reference/original/runs/real/cadata/rep00/variance_inflate_2_scores.npy
experiment_code/reference/original/runs/real/cadata/rep00/variance_inflate_3_scores.npy
experiment_code/reference/original/runs/real/cadata/rep00/variance_inflate_mask_u.npy
experiment_code/reference/original/runs/real/cpusmall/rep00/baseline.npy
experiment_code/reference/original/runs/real/cpusmall/rep00/clean.npy
experiment_code/reference/original/runs/real/cpusmall/rep00/correction_candidate.npy
experiment_code/reference/original/runs/real/cpusmall/rep00/correction_mask_u.npy
experiment_code/reference/original/runs/real/cpusmall/rep00/dirac_mask_u.npy
experiment_code/reference/original/runs/real/cpusmall/rep00/dirac_point_x.npy
experiment_code/reference/original/runs/real/cpusmall/rep00/dirac_point_y.npy
experiment_code/reference/original/runs/real/cpusmall/rep00/dirac_score.npy
experiment_code/reference/original/runs/real/cpusmall/rep00/heteroscedastic_0_scores.npy
experiment_code/reference/original/runs/real/cpusmall/rep00/heteroscedastic_1_scores.npy
experiment_code/reference/original/runs/real/cpusmall/rep00/heteroscedastic_2_scores.npy
experiment_code/reference/original/runs/real/cpusmall/rep00/heteroscedastic_3_scores.npy
experiment_code/reference/original/runs/real/cpusmall/rep00/heteroscedastic_mask_u.npy
experiment_code/reference/original/runs/real/cpusmall/rep00/shift_0_scores.npy
experiment_code/reference/original/runs/real/cpusmall/rep00/shift_1_scores.npy
experiment_code/reference/original/runs/real/cpusmall/rep00/shift_2_scores.npy
experiment_code/reference/original/runs/real/cpusmall/rep00/shift_3_scores.npy
experiment_code/reference/original/runs/real/cpusmall/rep00/shift_mask_u.npy
experiment_code/reference/original/runs/real/cpusmall/rep00/t_noise_0_scores.npy
experiment_code/reference/original/runs/real/cpusmall/rep00/t_noise_1_scores.npy
experiment_code/reference/original/runs/real/cpusmall/rep00/t_noise_2_scores.npy
experiment_code/reference/original/runs/real/cpusmall/rep00/t_noise_3_scores.npy
experiment_code/reference/original/runs/real/cpusmall/rep00/t_noise_mask_u.npy
experiment_code/reference/original/runs/real/cpusmall/rep00/target_scale.npy
experiment_code/reference/original/runs/real/cpusmall/rep00/test_mu.npy
experiment_code/reference/original/runs/real/cpusmall/rep00/test_q_hi.npy
experiment_code/reference/original/runs/real/cpusmall/rep00/test_q_lo.npy
experiment_code/reference/original/runs/real/cpusmall/rep00/test_sigma.npy
experiment_code/reference/original/runs/real/cpusmall/rep00/test_y_test.npy
experiment_code/reference/original/runs/real/cpusmall/rep00/variance_inflate_0_scores.npy
experiment_code/reference/original/runs/real/cpusmall/rep00/variance_inflate_1_scores.npy
experiment_code/reference/original/runs/real/cpusmall/rep00/variance_inflate_2_scores.npy
experiment_code/reference/original/runs/real/cpusmall/rep00/variance_inflate_3_scores.npy
experiment_code/reference/original/runs/real/cpusmall/rep00/variance_inflate_mask_u.npy
experiment_code/reference/original/runs/real/msd/rep00/baseline.npy
experiment_code/reference/original/runs/real/msd/rep00/clean.npy
experiment_code/reference/original/runs/real/msd/rep00/correction_candidate.npy
experiment_code/reference/original/runs/real/msd/rep00/correction_mask_u.npy
experiment_code/reference/original/runs/real/msd/rep00/dirac_mask_u.npy
experiment_code/reference/original/runs/real/msd/rep00/dirac_point_x.npy
experiment_code/reference/original/runs/real/msd/rep00/dirac_point_y.npy
experiment_code/reference/original/runs/real/msd/rep00/dirac_score.npy
experiment_code/reference/original/runs/real/msd/rep00/heteroscedastic_0_scores.npy
experiment_code/reference/original/runs/real/msd/rep00/heteroscedastic_1_scores.npy
experiment_code/reference/original/runs/real/msd/rep00/heteroscedastic_2_scores.npy
experiment_code/reference/original/runs/real/msd/rep00/heteroscedastic_3_scores.npy
experiment_code/reference/original/runs/real/msd/rep00/heteroscedastic_mask_u.npy
experiment_code/reference/original/runs/real/msd/rep00/shift_0_scores.npy
experiment_code/reference/original/runs/real/msd/rep00/shift_1_scores.npy
experiment_code/reference/original/runs/real/msd/rep00/shift_2_scores.npy
experiment_code/reference/original/runs/real/msd/rep00/shift_3_scores.npy
experiment_code/reference/original/runs/real/msd/rep00/shift_mask_u.npy
experiment_code/reference/original/runs/real/msd/rep00/t_noise_0_scores.npy
experiment_code/reference/original/runs/real/msd/rep00/t_noise_1_scores.npy
experiment_code/reference/original/runs/real/msd/rep00/t_noise_2_scores.npy
experiment_code/reference/original/runs/real/msd/rep00/t_noise_3_scores.npy
experiment_code/reference/original/runs/real/msd/rep00/t_noise_mask_u.npy
experiment_code/reference/original/runs/real/msd/rep00/target_scale.npy
experiment_code/reference/original/runs/real/msd/rep00/test_mu.npy
experiment_code/reference/original/runs/real/msd/rep00/test_q_hi.npy
experiment_code/reference/original/runs/real/msd/rep00/test_q_lo.npy
experiment_code/reference/original/runs/real/msd/rep00/test_sigma.npy
experiment_code/reference/original/runs/real/msd/rep00/test_y_test.npy
experiment_code/reference/original/runs/real/msd/rep00/variance_inflate_0_scores.npy
experiment_code/reference/original/runs/real/msd/rep00/variance_inflate_1_scores.npy
experiment_code/reference/original/runs/real/msd/rep00/variance_inflate_2_scores.npy
experiment_code/reference/original/runs/real/msd/rep00/variance_inflate_3_scores.npy
experiment_code/reference/original/runs/real/msd/rep00/variance_inflate_mask_u.npy
experiment_code/reference/original/runs/synthetic/synthetic/rep00/baseline.npy
experiment_code/reference/original/runs/synthetic/synthetic/rep00/clean.npy
experiment_code/reference/original/runs/synthetic/synthetic/rep00/dirac_mask_u.npy
experiment_code/reference/original/runs/synthetic/synthetic/rep00/dirac_point_x.npy
experiment_code/reference/original/runs/synthetic/synthetic/rep00/dirac_point_y.npy
experiment_code/reference/original/runs/synthetic/synthetic/rep00/dirac_score.npy
experiment_code/reference/original/runs/synthetic/synthetic/rep00/heteroscedastic_0_scores.npy
experiment_code/reference/original/runs/synthetic/synthetic/rep00/heteroscedastic_1_scores.npy
experiment_code/reference/original/runs/synthetic/synthetic/rep00/heteroscedastic_2_scores.npy
experiment_code/reference/original/runs/synthetic/synthetic/rep00/heteroscedastic_3_scores.npy
experiment_code/reference/original/runs/synthetic/synthetic/rep00/heteroscedastic_mask_u.npy
experiment_code/reference/original/runs/synthetic/synthetic/rep00/paired_clean_baseline.npy
experiment_code/reference/original/runs/synthetic/synthetic/rep00/reference_type.npy
experiment_code/reference/original/runs/synthetic/synthetic/rep00/reference_widths.npy
experiment_code/reference/original/runs/synthetic/synthetic/rep00/shift_0_scores.npy
experiment_code/reference/original/runs/synthetic/synthetic/rep00/shift_1_scores.npy
experiment_code/reference/original/runs/synthetic/synthetic/rep00/shift_2_scores.npy
experiment_code/reference/original/runs/synthetic/synthetic/rep00/shift_3_scores.npy
experiment_code/reference/original/runs/synthetic/synthetic/rep00/shift_mask_u.npy
experiment_code/reference/original/runs/synthetic/synthetic/rep00/t_noise_0_scores.npy
experiment_code/reference/original/runs/synthetic/synthetic/rep00/t_noise_1_scores.npy
experiment_code/reference/original/runs/synthetic/synthetic/rep00/t_noise_2_scores.npy
experiment_code/reference/original/runs/synthetic/synthetic/rep00/t_noise_3_scores.npy
experiment_code/reference/original/runs/synthetic/synthetic/rep00/t_noise_mask_u.npy
experiment_code/reference/original/runs/synthetic/synthetic/rep00/target_scale.npy
experiment_code/reference/original/runs/synthetic/synthetic/rep00/test_mu.npy
experiment_code/reference/original/runs/synthetic/synthetic/rep00/test_q_hi.npy
experiment_code/reference/original/runs/synthetic/synthetic/rep00/test_q_lo.npy
experiment_code/reference/original/runs/synthetic/synthetic/rep00/test_sigma.npy
experiment_code/reference/original/runs/synthetic/synthetic/rep00/test_y_test.npy
experiment_code/reference/original/runs/synthetic/synthetic/rep00/variance_inflate_0_scores.npy
experiment_code/reference/original/runs/synthetic/synthetic/rep00/variance_inflate_1_scores.npy
experiment_code/reference/original/runs/synthetic/synthetic/rep00/variance_inflate_2_scores.npy
experiment_code/reference/original/runs/synthetic/synthetic/rep00/variance_inflate_3_scores.npy
experiment_code/reference/original/runs/synthetic/synthetic/rep00/variance_inflate_mask_u.npy
experiment_code/requirements.txt
experiment_code/tests/test_core.py
experiment_code/tests/validate_runs.py
(299 additional files not shown) You must enabled JavaScript to view entire file list.
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Text integral sub licența deschisă a publicației · 547 cuvinte
Subiecte conexe
DI Series: Serving Kimi-K3 MXFP4 Wide-EP Disaggregated on AMD Instinct MI300X and MI325X
AMD · 8 oct. 2026
Mutual information across the commensurate--incommensurate crossover and near the Kosterlitz--Thouless point of the spin-$\tfrac12$ $J_1$--$J_2$ Heisenberg chain
arXiv · 14 oct. 2026
Conformalized Quantile Regression and Minimax Limits of Fixed-Score Calibration under Known Covariate Shift
arXiv · 22 sept. 2026
Questioning the Coverage-Length Metric in Conformal Prediction: When Shorter Intervals Are Not Better
arXiv · 17 iun. 2026