Browse Articles
Discover research articles across all indexed journals
Enhanced adsorptive removal of malachite green dye from aqueous solution using raw and acid modified rosemary biomass
Author Correction: Inactivating SnRK1β1A promotes broad-spectrum disease resistance in rice
Risk factor analysis and construction of risk nomogram for severe fever with thrombocytopenia syndrome based on clinical classification
Ambiphilic cross-coupling with aryl-bismuth reagents
Abstract Cross-coupling reactions traditionally permit the formation of Ar-Ar bonds between an aryl nucleophile and an aryl electrophile under transition-metal catalysis 1,2 . The high selectivity of the myriad of couplings known to date relies on a tailored combination of nucleophilic and electrophilic coupling partners, enabled by the mechanistic distinction between nucleophiles and electrophiles, which undergo fundamentally different catalytic steps 3 . Here we report ambiphilic aryl-bismuth reagents that can behave as either nucleophiles or electrophiles in transition-metal-catalysed cross-couplings, fundamentally breaking from this dichotomy in reactivity. Their ambiphilic reactivity arises from their ability to engage in both oxidative addition and transmetalation processes with transition-metal complexes, as demonstrated by stoichiometric and mechanistic studies. By demonstrating that a single aryl reagent can engage in both canonical elementary steps, this work challenges the long-standing assumption that intrinsic bond polarity rigidly dictates the mechanistic role in cross-coupling chemistry.
Early cerebral palsy risk screening through wearable accelerometry in infants under nine weeks
Validation of the ultra short term heart rate variability measures under changing psychophysiological state of the healthy participants at rest
AI-Assisted segmentation and volumetric reconstruction of radiographs through multi-angular scintillation imaging
Perceived exertion reflects exercise-induced pain rather than psychological characteristics in recreational endurance athletes
The vagus nerve promotes memory in rats via nutrient-induced septo-hippocampal acetylcholine signaling
Tough peer-review process? Your paper might end up being more highly cited
Effects of oropharyngeal exercises on orofacial function and corticomotor excitability in healthy individuals
Imaging foundation model for universal enhancement of non-ideal measurement CT
Mitigating multimodal hallucinations through visual attention tracing and origin-point regeneration
Abstract Multimodal large language models (MLLMs) exhibit impressive prowess in vision-language understanding, but their utility is often compromised by hallucinations—instances where generated narratives diverge significantly from visual evidence. Current remedial strategies largely struggle to pinpoint the genesis of these errors, relying either on resource-intensive retraining or indiscriminate global penalties that fail to address the specific locus of the discrepancy. Addressing this limitation, we introduce hallucination backtracking (HB), a training-free decoding framework designed to effectively detect and mitigate errors by monitoring visual attention dynamics during generation. This approach is grounded in the observation that hallucinations are not random; rather, they stem from specific pivotal tokens where the model’s focus precipitously shifts from image features to its own prior textual generations. By quantifying this drift through a novel visual attention score (VAS), our origin-point detection mechanism successfully isolates the source of errors, achieving a 41.8% exact match and 84.1% before-first accuracy in localization. Once an attentional anomaly is detected, the system autonomously backtracks to the divergence point, triggering a regeneration process reinforced by stricter visual grounding constraints. Rigorous evaluations across diverse architectures—LLaVA-1.5, InstructBLIP, MiniGPT-4, and Shikra—confirm that HB consistently surpasses state-of-the-art baselines; notably, on LLaVA-1.5, our method elevates the F1 score on the POPE benchmark to 91.4% while reducing the CHAIR $$_S$$ metric to 40.2%, yielding improvements of 1.5 and 4.4 points over OPERA, respectively, though a residual false negative rate of 15.9% indicates that inference-driven hallucinations remain an open challenge. Beyond quantitative gains, we provide a granular dissection of the hallucination phenomenon through analyses of VAS trajectory patterns and failure modes, ultimately advocating for precise localization and targeted correction as a promising paradigm for reliable multimodal generation.