Researchers tested 26 large language models from eight developers on their ability to produce accurate biomedical citations. The authors report that fabricated or incorrect references remain a persistent problem across the models evaluated. The work is a preprint and characterises the frequency of this failure mode rather than proposing a fix.
A new arXiv paper introduces the concept of "efficiency hallucination," describing how large language models sometimes propose code changes that do not alter functionality while claiming performance gains that are not actually supported. The authors formalize and measure this behavioral calibration problem to assess how reliably LLMs can justify their optimizations. The work aims to give researchers a way to quantify when model claims about speedups diverge from verified results.
A new arXiv paper examines how well audio-language models can turn long doctor-patient conversations into structured SOAP clinical notes. The authors compare lightweight and heavyweight end-to-end approaches, noting that while cascaded speech recognition pipelines remain strong, end-to-end models tend to lose information or produce hallucinations. The work targets the modality gap in long-form clinical audio.
A new arXiv paper introduces E2A-Bench, a benchmark aimed at measuring how reliably financial vision-language models convert chart evidence into action recommendations. The authors argue that existing hallucination tests focus on whether individual claims are supported, rather than whether the underlying evidence actually drives the recommended action. The benchmark is designed to close that gap in evaluating financial chart reasoning.
Researchers present a hierarchical hybrid model that combines one-dimensional convolutional networks with bidirectional LSTMs to produce extractive summaries of biomedical and clinical text. The approach avoids text generation entirely, sidestepping the factual hallucination risks that make abstractive large language models unreliable in medical settings. The work is posted as a preprint on arXiv.
A new arXiv paper examines what tool-augmented language models do when a tool call fails to return usable information, rather than focusing only on whether they reach the correct answer. The authors built a benchmark of 1,024 items spanning 16 internal systems to isolate this post-failure decision point. They find that agents tend to assert values their tools never returned instead of reporting the failure honestly.
Researchers assembled a corpus of 150 academic papers with deliberately planted errors to test how well large language models can act as automated document-quality auditors. They report that detection reliability worsens as processing batch sizes increase, and that models sometimes fabricate audit findings with high confidence. The results cast doubt on deploying LLMs unsupervised for contamination-detection tasks.
Researchers have introduced LexAgentHallu, a new benchmark for measuring how tool-augmented legal AI agents hallucinate. It uses a hierarchical structure to trace how errors in tool calls and reasoning cascade into fabricated case holdings and miscited legal authority. The benchmark aims to fill a gap left by existing legal evaluations that do not capture agentic workflows.