AI
A Synthetic Ground-Truth Framework for the Evaluation of Explainable AI Methods
Publication: cs.AI updates on arXiv.org
Author: Miquel Mir\'o-Nicolau, Francesco Spinnato, Riccardo Guidotti
Source published: 28/09/2026, 04:00:00 UTC
Briefing created: 28/09/2026, 04:07:33 UTC
Based on a feed excerpt; full article content has not been acquired. Source evidence retrieved 28/09/2026, 04:03:24 UTC.
Briefings
A new framework proposes synthetic ground truth evaluation for XAI methods using controlled interventions to determine input component importance. The approach was tested across three domains (binary images, tabular data, and time series) using nine widely used XAI methods. Experimental results indicate significant limitations in current techniques, highlighting the need for intervention-based benchmarks for reliable assessment.
Key points
- The framework uses controlled interventions to generate synthetic datasets where input component importance can be determined by design.
- Nine widely used XAI methods were evaluated across three domains: binary images, tabular data, and time series.
Context and limitations
Experimental results are limited to nine methods and three data domains; uncertainty remains regarding the generalizability of findings across diverse XAI approaches.
Generated by Bittrees-hosted qwen35-2b sha256:aaf42c8b7c3c on 28/09/2026, 04:07:33 UTC. Based on publisher-provided feed evidence; may not cover the full original article. Check the source for details.
View archived briefing on IPFS
bafybeic2ex3uokn3bvcwyhfkpdmgorjn3ycxmeo2niogjh4k5vozlxcugi