Study Reveals Methodological Flaws in ROAR Benchmark for Evaluating Feature Attribution in Neural Networks
Researchers have shown that the RemOve-And-Retrain (ROAR) benchmark, commonly used to evaluate how well AI explanation methods identify important features, can be gamed by simple post-processing transformations that add no real information. The flaw stems from a bias toward spatially blurry masks, which inflate ROAR scores without reflecting genuine explanatory quality. This undermines confidence in a key tool used to validate mechanistic understanding of neural networks.
The RemOve-And-Retrain (ROAR) benchmark is one of the most widely adopted methods for evaluating feature attribution techniques in machine learning, which aim to explain which parts of an input drive a model's decisions. A new study accepted at the 2026 ICML Workshop on Mechanistic Interpretability demonstrates that model- and data-agnostic post-processing of attribution maps — transformations that, by the data processing inequality, cannot add information about the underlying decision function — can nonetheless improve a method's ROAR score. This reveals that a higher ROAR ranking does not reliably indicate that an attribution map is more informative about the model. The researchers trace this failure to a systematic bias favoring spatially blurry masks, a pattern confirmed through experiments on three image datasets: CIFAR-10, SVHN, and CUB-200. The same bias was also observed in ROAD, a variant of the ROAR benchmark. The authors provide practical guidelines for more cautious use of removal-based benchmarks, with broader implications for how the field validates explainability and interpretability methods for neural networks.
What's missing
It is unclear how broadly the findings generalize beyond image classification tasks to other modalities or model architectures.
What different sources said
- arXiv cs.AICenter
On Pitfalls of $\textit{RemOve-And-Retrain}$: Data Processing Inequality Perspective
Related
Gut Bacteria Enzyme Found to Break Down Heat-Processed Food Compounds, Producing Novel Biogenic Amines
Researchers have discovered that an enzyme in common gut bacteria can degrade N-epsilon-carboxymethyllysine (CML), a compound formed during thermal food processing, producing previously unknown biogenic amines. The enzyme, ornithine decarboxylase SpeC from enterobacteria, acts on CML and related modified lysine derivatives through a low-level 'underground' catalytic activity. This finding suggests a previously unrecognized communication axis between thermally processed dietary compounds and gut microbial physiology, with potential implications for host health.
Full-Length Gene Sequencing Reveals Two Distinct Bacterial Communities in Black-Legged Ticks Expanding Into Canada
Researchers used Oxford Nanopore full-length 16S rRNA gene sequencing to characterize the microbiome of Ixodes scapularis black-legged ticks collected in Nova Scotia, Canada, distinguishing between tick-adapted bacteria and environmentally acquired bacteria. The study comes as I. scapularis — the primary vector of Lyme disease — is rapidly expanding northward into Canada due to climate change. The findings suggest that environmentally derived bacteria in tick microbiomes are not mere contamination, which has implications for how tick microbiome data is collected and interpreted across surveillance studies.
Study Identifies Metabolic Link Between Cell Envelope Stress and Biofilm Formation in Bacteria
Researchers have discovered that the metabolite acetyl-CoA directly inhibits enzymes that degrade the bacterial signaling molecule c-di-GMP, connecting cell envelope biosynthesis stress to biofilm formation in Pseudomonas aeruginosa. The study found that sub-inhibitory concentrations of antibiotics targeting early peptidoglycan biosynthesis — but not other antibiotic classes — elevate c-di-GMP levels by reducing phosphodiesterase activity, with acetyl-CoA competing for the enzyme active site. Because the relevant enzyme domain is broadly conserved across bacterial species, this checkpoint mechanism may be widespread and could have implications for understanding antibiotic-induced biofilm responses.