HCMA Paper 4 Extension: Mondrian Conformal Correction for Conditional Coverage

Extension of HCMA Paper 4 (OSF GHVMZ): a Mondrian (group-conditional) conformal correction that restores conditional coverage in versioned retrieval. Paper 4's conformal abstention guarantee was marginal; stratifying by query difficulty exposed a conditional coverage gap (medium-margin group loss 0.226 vs α=0.10). This extension partitions queries by their retrieval margin (low/medium/high, tertiles of the calibration split) and calibrates a separate conformal threshold per group. Result: the medium-margin group loss drops from 0.226 to 0.043 (α=0.10), restoring conditional validity at the cost of coverage in the hardest group. The margin stratifies difficulty cleanly (accuracy 0.365 / 0.774 / 1.000 across groups), confirming that margin-based conditioning is sufficient — no version-count signal is required. Artifacts: mondrian_conformal.py + mondrian_conformal_results.json. Requires: HCMA core (Paper 1, OSF VQX95) and Paper 4 (OSF GHVMZ). License: CC-BY 4.0 (docs); Apache-2.0 (code, SPDX headers inline).

Authors

Publication Details

Journal
Open Science Framework
Published
2026-10-30
DOI
https://doi.org/10.17605/osf.io/3k5mq
Primary Topic
Information Retrieval and Search Behavior
Type
article
Field-Weighted Citation Impact
0.00
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
article

HCMA Paper 4 Extension: Mondrian Conformal Correction for Conditional Coverage

Mauro Javier Silva
2 links
Open Science Framework
Information Retrieval and Search Behavior
article

HCMA Paper 4 Extension: Mondrian Conformal Correction for Conditional Coverage

Mauro Javier Silva
article en

Abstract

Extension of HCMA Paper 4 (OSF GHVMZ): a Mondrian (group-conditional) conformal correction that restores conditional coverage in versioned retrieval. Paper 4's conformal abstention guarantee was marginal; stratifying by query difficulty exposed a conditional coverage gap (medium-margin group loss 0.226 vs α=0.10). This extension partitions queries by their retrieval margin (low/medium/high, tertiles of the calibration split) and calibrates a separate conformal threshold per group. Result: the medium-margin group loss drops from 0.226 to 0.043 (α=0.10), restoring conditional validity at the cost of coverage in the hardest group. The margin stratifies difficulty cleanly (accuracy 0.365 / 0.774 / 1.000 across groups), confirming that margin-based conditioning is sufficient — no version-count signal is required. Artifacts: mondrian_conformal.py + mondrian_conformal_results.json. Requires: HCMA core (Paper 1, OSF VQX95) and Paper 4 (OSF GHVMZ). License: CC-BY 4.0 (docs); Apache-2.0 (code, SPDX headers inline).

Open Science Framework
Openalex Percentile: Top 7%
Information Retrieval and Search Behavior
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon