When Verification Fails: How Compositionally Infeasible Claims Escape Rejection

TL;DR

This paper reveals that current scientific claim verification models often rely on shortcut reasoning, failing to properly evaluate compositional infeasibility, which undermines the reliability of verification benchmarks.

Contribution

It demonstrates the limitations of existing benchmarks in distinguishing rigorous verification from shortcut strategies and introduces compositionally infeasible claims to better evaluate model reasoning.

Findings

Models over-accept compositionally infeasible claims, indicating reliance on shortcut reasoning.

Existing benchmarks cannot differentiate between true verification and salient-constraint checking.

Verification thresholds vary across models, reflecting structural biases rather than reasoning ability.

Abstract

Scientific claim verification, the task of determining whether claims are entailed by scientific evidence, is fundamental to establishing discoveries in evidence while preventing misinformation. This process involves evaluating each asserted constraint against validated evidence. Under the Closed-World Assumption (CWA), a claim is accepted if and only if all asserted constraints are positively supported. We show that existing verification benchmarks cannot distinguish models enforcing this standard from models applying a simpler shortcut called salient-constraint checking, which applies CWA's rejection criterion only to the most salient constraint and accepts when that constraint is supported. Because existing benchmarks construct infeasible claims by perturbing a single salient element they are insufficient at distinguishing between rigorous claim verification and simple salient-constraint reliance. To separate the two, we construct compositionally infeasible claims where the salient constraint is supported but a non-salient constraint is contradicted. Across model families and modalities, models that otherwise saturate existing benchmarks consistently over-accept these claims, confirming the prevalence of such shortcut reasoning. Via model context interventions, we show that different models and prompting strategies occupy distinct positions on a shared ROC curve, indicating that the gap between model families reflects differences in verification threshold rather than underlying reasoning ability, and that the compositional inference bottleneck is a structural property of current verification behavior that strategy guidance alone cannot overcome.