ICML 2026 #30204 Reproduction
Collection
Source paper, Trackio logbook, and complete artifacts for Networked Information Aggregation for Binary Classification. • 3 items • Updated
paper dict | mode string | method string | quadrature_order int64 | k_values list | mc dict | invariants dict | lower_bound_fits list | idealized_lower_bound_fit dict | upper_bound_check dict | environment dict | wall_seconds float64 |
|---|---|---|---|---|---|---|---|---|---|---|---|
{
"title": "Networked Information Aggregation for Binary Classification",
"arxiv_id": "2605.01082",
"openreview_id": "mrtg4NmvAe"
} | substantive | deterministic population sequential-logit simulation plus independent Monte Carlo lower-bound check | 140 | [
4,
8,
12,
16,
24
] | {
"samples_per_seed": 500000,
"seeds": [
11,
23,
37
],
"device": "cpu",
"fit_log_excess_vs_log_p": {
"slope": -0.9350206812289982,
"intercept": -2.6686461943304782,
"r2": 0.9968147767574095
}
} | {
"max_agent_stationarity_abs": 1.0819462638311794e-8,
"max_end_of_pass_off_relevant_l2": 0,
"max_loss_decomposition_abs_error": 6.465142761424847e-10,
"min_pinsker_slack": -7.200367990754695e-17
} | [
{
"k": 4,
"n_points": 3,
"slope": -0.5878402635155259,
"intercept": -3.009358647585332,
"r2": 0.9986958337218635
},
{
"k": 8,
"n_points": 7,
"slope": -0.6076749374430602,
"intercept": -3.0013797439977976,
"r2": 0.999417157745313
},
{
"k": 12,
"n_points": 11,
... | {
"all_points": {
"slope": -0.9352501163874916,
"intercept": -2.6695775957507704,
"r2": 0.9968323922882505
},
"tail_p_ge_128": {
"slope": -0.9977674521988302,
"intercept": -2.287549895506178,
"r2": 0.9999989096931335
},
"tail_p_excess_range": [
0.10242379279748093,
0.1032825393... | {
"all_excess_below_theorem_bound": true,
"max_excess_to_bound_ratio": 0.004337979502507641,
"note": "The theorem bound is valid but loose on its own hard instance; this check does not establish worst-case tightness."
} | {
"python": "3.11.15",
"platform": "Linux-6.17.0-23-generic-x86_64-with-glibc2.39",
"numpy": "2.4.3",
"scipy": "1.17.1",
"torch": "2.13.0+cu130",
"cuda_available": false,
"cuda_device": null
} | 8.304809 |
ICML 2026 #30204 reproduction artifacts
Artifacts for Networked Information Aggregation for Binary Classification (arXiv:2605.01082; OpenReview mrtg4NmvAe).
Downloads
icml-2026-30204-repro-bundle.tar.gz— complete 50-file reproduction bundleMANIFEST.sha256— checksums of files inside the unpacked bundleposter.pdf— gate-verified 24x36 inch reproduction postersummary.json— authoritative aggregate numerical resultsEvidence_Memo.md— claim-by-claim source and numerical auditrepro.py— deterministic population simulation and Monte Carlo cross-check
Outer archive SHA-256:
ef4324252697ab531e88335f225ca7e2de418f8aebfbf6775a1829fb4f6c6837
The paper PDF and TeX source inside the bundle remain under their original authors' terms; the independent reproduction scripts and generated results are provided for challenge reproducibility.
- Downloads last month
- 26