Search for: All records

Smart contracts underpin decentralized applications but face significant security risks from vulnerabilities, while traditional analysis methods have limitations. Large Language Models (LLMs) offer promise for vulnerability detection, yet adapting these powerful models efficiently, particularly generative ones, remains challenging. This paper investigates two key strategies for the efficient adaptation of LLMs for Solidity smart contract vulnerability detection: (1) replacing token-level generation with a dedicated classification head during fine-tuning, and (2) selectively freezing lower transformer layers using Low-Rank Adaptation (LoRA). Our empirical evaluation demonstrates that the classification head approach enables models like Llama 3.2 3B to achieve high accuracy (77.5%), rivaling the performance of significantly larger models such as the fine-tuned GPT-3.5. Furthermore, we show that selectively freezing bottom layers reduces training time and memory usage by approximately 10-20% with minimal impact on accuracy. Notably, larger models (3B vs. 1B parameters) exhibit greater resilience to layer freezing, maintaining high accuracy even with a large proportion of layers frozen, suggesting a localization of general code understanding in lower layers versus task-specific vulnerability patterns in upper layers. These findings present practical insights for developing and deploying performant LLM-based vulnerability detection systems efficiently, particularly in resource-constrained settings. 

W e present a scalable solution-processing method for fabricating high-quality graphene and graphene/1T-MoS 2 heterostructure films. The process begins with the synthesis of potassium-intercalated graphite (KC 8 ), which is exfoliated in tetrahydrofuran (THF) to produce stable dispersions of negatively charged (electron rich) graphene sheets. The graphene is subsequently transferred to water, forming a surfactant-free aqueous dispersion suitable for creating homogenous graphene films via vacuum filtration and stamping. Additionally, graphene is combined with 1T-MoS 2 nanosheets to fabricate graphene/1T-MoS 2 bulk heterostructure films. Comprehensive characterization, including X-ray diffraction (XRD), absorption spectroscopy, scanning electron microscopy (SEM), transmission electron microscopy ( TEM), Raman spectroscopy, and X-ray photon emission spectroscopy (XPS), reveals that the heterostructure films exhibit enhanced optical and electronic properties, including improved light absorption, which could lead to novel photo-responsive devices. Raman spectroscopy shows significant changes in the graphene’s structural a nd electronic properties upon interaction with MoS 2 , indicating strong interlayer coupling and potential charge transfer between the layered components. The g raphene films demonstrate highly sensitive detection of dopamine (DA), while the graphene/1T-MoS 2 b ulk heterostructure films exhibit capacitance values up to 3 8.3 Fg − 1 at 5 mV/s in non-aqueous electrolytes. These results highlight the potential of these films for advanced applications in molecular sensing and energy storage. 

We report a facile method to prepare polymer-grafted plasmonic metal nanoparticles (NPs) that exhibit pH-responsive surface-enhanced Raman scattering (SERS). The concept is based on the use of pH- responsive polymers, such as poly(acrylic acid) (PAA) and poly(allylamine hydrochloride) (PAH), as multi- dentate ligands to wrap around the surface of NPs instead of forming polymer brushes. Upon changing the solvent quality, the grafted pH-responsive polymers would drive reversible aggregation of NPs, leading to a decreased interparticle distance. This creates numerous hot spots, resulting in a secondary enhancement of SERS as compared to the SERS from discrete NPs. For negatively charged PAA-grafted NPs, the SERS response at pH 2.5 showed a secondary enhancement of up to 104-fold as compared to the response for discrete NPs at pH 12. Similarly, positively charged PAH-grafted AuNPs showed an oppo- site response to pH. We demonstrated that enhanced SERS with thiol-containing and charged molecular probes was indeed from the pH-driven solubility change of polymer ligands. Our method is different from the conventional SERS sensors in the solid state. With pH-responsive polymer-grafted NPs, SERS can be performed in solution with high reproducibility and sensitivity but without the need for sample pre-con- centration. These findings could pave the way for innovative designs of polymer ligands for metal NPs where polymer ligands do not compromise interparticle plasmon coupling. 

Immune checkpoint inhibitors can stimulate antitumor immunity but can also induce toxicities termed immune-related adverse events (irAEs). Colitis is a common and severe irAE that can lead to treatment discontinuation. Mechanistic understanding of gut irAEs has been hampered because robust colitis is not observed in laboratory mice treated with checkpoint inhibitors. We report here that this limitation can be overcome by using mice harboring the microbiota of wild-caught mice, which develop overt colitis following treatment with anti-CTLA-4 antibodies. Intestinal inflammation is driven by unrestrained activation of IFNγ-producing CD4⁺T cells and depletion of peripherally induced regulatory T cells through Fcγ receptor signaling. Accordingly, anti-CTLA-4 nanobodies that lack an Fc domain can promote antitumor responses without triggering colitis. This work suggests a strategy for mitigating gut irAEs while preserving antitumor stimulating effects of CTLA-4 blockade.