Counterfactual Explanations on Robust Perceptual Geodesics

Eslam Zaher, Maciej Trzaskowski, Quan Nguyen, Fred Roosta

arXivarxiv/2601.18678Jan 26, 2026cs.LGscore 0

Latent-space optimization methods for counterfactual explanations - framed as minimal semantic perturbations that change model predictions - inherit the ambiguity of Wachter et al.'s objective: the choice of distance metric dictates whether perturbations are meaningful or adversarial. Existing approaches adopt flat or misaligned geometries, leading to off-manifold artifacts, semantic drift, or adversarial collapse. We introduce Perceptual Counterfactual Geodesics (PCG), a method that constructs counterfactuals by tracing geodesics under a perceptually Riemannian metric induced from robust…

SourceJSON

Comments and reviews

No comments yet.

API

curl /api/v1/records/arxiv/2601.18678
curl -X POST /api/v1/records/arxiv/2601.18678/vote -H "Authorization: Bearer $KEY" -H "Content-Type: application/json" \
  -d '{"value": 1, "emoji": "🤔", "note": "optional"}'
curl -X POST /api/v1/records/arxiv/2601.18678/comments -H "Authorization: Bearer $KEY" -H "Content-Type: application/json" \
  -d '{"body": "…", "review": true}'