Scollr found 1,597 eligible XAI papers in the most recent 30-day window, compared with 812 in the previous 30 days. The current evidence is broad but concentrated in a single OpenAlex topic cluster, with many representative items appearing as arXiv preprints.
Explainable Artificial Intelligence (XAI) shows strong short-term publication momentum: the recent 30-day window contains 1,597 eligible papers, nearly double the 812 eligible papers in the prior 30-day window. The available cluster evidence is not fine-grained—only the parent XAI topic appears among top descendants—so the safest reading is a field-wide surge rather than a clearly separable subtopic shift. Representative recent papers emphasize explanation methods and evaluation issues, including feature association maps, saliency decorrelation, partial dependence plots, personalized rule explanations, explanation stability, counterfactual explanations, local class differentiation with Shapley-style explanations, and interpretation-as-evaluation.
The recent window includes 1,597 eligible papers versus 812 in the prior 30-day window, an increase of 785 papers, or roughly a 97% rise. This supports a short-term momentum signal for the topic, though it does not by itself establish longer-term growth.
Window-level countsThe top-descendant evidence lists only the parent topic, Explainable Artificial Intelligence (XAI), with 1,597 recent eligible papers. Because the evidence threshold indicates the cluster criterion is not met, this brief avoids claiming specific subfields are independently trending.
Top-cluster evidenceThe selected recent papers span proposed explanation techniques and interpretability evaluation topics: feature association maps, saliency decorrelation, partial dependence plots for tree ensembles, preference-based rule explanations, stability under perturbations, robust counterfactuals, Shapley-style local class differentiation, and rigorous interpretation as evaluation. Several are preprints, so their findings should be treated as provisional until peer review or replication.
Representative paper titles and metadataThe only top descendant surfaced in the evidence packet is the parent XAI topic itself, accounting for all 1,597 recent eligible papers. This indicates strong topic-level activity but does not provide enough cluster resolution to rank separate XAI subthemes.
1,597 recent eligible papers in the parent XAI topicThe representative set points to active work on explanation generation, local and global interpretability, stability, counterfactual robustness, and evaluation framing. Because this cluster is inferred from representative titles rather than a separate top-descendant cluster, it should be read as a qualitative snapshot, not a measured subtopic count.
Representative recent papersRecent XAI preprint proposing a feature-association-map approach; useful as a provisional example of method-building activity in the window.
Recent preprint in the representative set focused on saliency and feature decorrelation for neural network interpretability; treat findings as not yet settled.
Represents work on partial dependence plots for decision-tree ensembles, highlighting continued interest in efficient global explanation tools.
Representative preprint on personalized rule explanations for black-box models, showing user-preference-oriented explanation work in the recent window.
Peer-reviewed article in the representative set addressing explanation stability under perturbations, a key evaluation concern for XAI methods.
Recent preprint representing counterfactual-explanation work, specifically framed around robustness and probabilistic consensus.
Peer-reviewed article focused on local explainability for class differentiation using Shapley additive explanations, illustrating model-specific interpretability refinement.
Representative preprint that frames interpretation as evaluation, reflecting methodological scrutiny around what XAI explanations are meant to validate.