Despite achieving high accuracy on medical benchmarks, LLMs exhibit the Einstellung Effect in clinical diagnosis—relying on statistical shortcuts rather than patient-specific evidence, causing misdiagnosis in atypical cases. Existing benchmarks fail to detect this critical failure mode. We introduce MedEinst, a counterfactual benchmark with 5,383 paired clinical cases across 49 diseases. Each pair contains a control case and a “trap” case with altered discriminative evidence that flips the diagnosis. We measure susceptibility via Bias Trap Rate—probability of misdiagnosing traps despite correctly diagnosing controls. Evaluation shows frontier models achieve high baseline accuracy but severe bias trap rates. Thus, we propose ECR-Agent, aligning LLM reasoning with Evidence-Based Medicine via two components: (1) Dynamic Causal Inference (DCI) performs structured reasoning through dual-pathway perception, dynamic causal graph reasoning across three levels (association, intervention, counterfactual), and evidence audit for final diagnosis; (2) Critic-Driven Graph Memory Evolution (CGME) iteratively refines the system by storing validated reasoning paths in an exemplar base and consolidating disease-specific knowledge into evolving illness graphs. Source code is to be released.

Since it is difficult to obtain the number of sources accurately in practical situations, direction-of-arrival (DOA) estimation methods without the number of sources are very much needed for non-cooperative targets and maritime attack-defense pairings. In this paper, a DOA estimation method using acoustic vector sensors (AVS) without a known number of sources is proposed. First, a reference matrix is introduced, and the eigenvalues corresponding to the noise of this matrix converge to 0. Next, a scanning matrix containing the scanning sources is created, and the analysis of noise power invariance is used to determine the power selection method of the scanning sources. The ordering law of the eigenvalues of the source matrix scanned at different angles is applied, and the purpose of estimating DOA without the number of sources is accomplished. The proposed algorithm is capable of performing DOA estimation in situations where the number of sources is uncertain and it does not rely on the array structure, which means it can be used for arbitrary arrays. Finally, the correctness and effectiveness of the proposed algorithm are verified by simulation and experiment.