arXiv subjects

Nicolas Ramos

Publications and source records attributed to Nicolas Ramos.

3 recordsLinked to original sources

arXiv2025

SafeTy Reasoning Elicitation Alignment for Multi-Turn Dialogues

Malicious attackers can exploit large language models (LLMs) by engaging them in multi-turn dialogues to achieve harmful objectives, posing significant safety risks to society. To address this challenge, we propose a novel defense mechanism: SafeTy Reasoning Elicitation Alignment for Multi-Turn Dialogues (STREAM). STREAM defends LLMs against multi-turn attacks while preserving their functional capabilities. Our approach involves constructing a human-annotated dataset, the Safety Reasoning Multi-turn Dialogues dataset, which is used to fine-tune a plug-and-play safety reasoning moderator. This model is designed to identify malicious intent hidden within multi-turn conversations and alert the target LLM of potential risks. We evaluate STREAM across multiple LLMs against prevalent multi-turn attack strategies. Experimental results demonstrate that our method significantly outperforms existing defense techniques, reducing the Attack Success Rate (ASR) by 51.2%, all while maintaining comparable LLM capability.

cs.CL↗

arXiv2021

Rationalizing the influence of tunable energy levels on quantum efficiency to design optimal non-fullerene acceptor-based ternary organic solar cells

Non-fullerene acceptor (NFA)-based ternary bulk heterojunction solar cells (TSC) are the most efficient organic solar cells (OSCs) today due to their broader absorption and quantum efficiencies (QE) often surpassing those of corresponding binary blends. We study how the energetics driving charge transfer at the electron donor:electron acceptor (D/A) interfaces impact the QE in blends of PBDB-T-2F donor with several pairs of lower bandgap NFAs. As in binary blends, the ionization energy offset between donor and acceptor ({\Delta}IE) controls the QE and maximizes for {\Delta}IE > 0.5 eV. However, {\Delta}IE is not controlled by the individual NFAs IEs but by their average, weighted for their blending ratio. Using this property, we improved the QE of a PBDB-T-2F:IEICO binary blend that had an insufficient {\Delta}IE for charge generation by adding a deep IE third component: IT-4F. Combining two NFAs enables to optimize the D/A energy alignment and cells' QE without molecular engineering.

physics.app-ph↗

arXiv2018

Mid-to-far infrared tunable perfect absorption by a sub - λ/100 nanofilm in a fractal phasor resonant cavity

Integrating an absorbing thin film into a resonant cavity is the most practical way to achieve perfect absorption of light at a selected wavelength in the mid-to-far infrared, as required to target blackbody radiation or molecular fingerprints. The cavity is designed to resonate and enable perfect absorption in the film at the chosen wavelength λ. However, in current state-of-the-art designs, a still large absorbing film thickness (λ/50) is needed and tuning the perfect absorption wavelength over a broad range requires changing the cavity materials. Here, we introduce a new resonant cavity concept to achieve perfect absorption of infrared light in much thinner and thus really nanoscale films, with a broad wavelength tunability using a single set of cavity materials. It requires a nanofilm with giant refractive index and small extinction coefficient (found in emerging semi-metals, semi-conductors and topological insulators) backed by a transparent spacer and a metal mirror. The nanofilm acts both as absorber and multiple reflector for the internal cavity waves, which after escaping follow a fractal phasor trajectory. This enables a totally destructive optical interference for a nanofilm thickness more than 2 orders of magnitude smaller than λ. With this remarkable effect, we demonstrate angle-insensitive perfect absorption in sub - λ/100 bismuth nanofilms, at a wavelength tunable from 3 to 20 μm.

physics.optics↗