We present HILGEN, a Hierarchically-Informed Data Generation approach that combines domain knowledge from the Unified Medical Language System (UMLS) with synthetic data generated by large language models (LLMs), specifically GPT-3.5. Our approach leverages UMLS's hierarchical structure to expand training data with related concepts, while incorporating contextual information from LLMs through targeted prompts aimed at automatically generating synthetic examples for sparsely occurring named entities. The performance of the HILGEN approach was evaluated across four biomedical NER datasets (MIMIC III, BC5CDR, NCBI-Disease, and Med-Mentions) using BERT-Large and DANN (Data Augmentation with Nearest Neighbor Classifier) models, applying various data generation strategies, including UMLS, GPT-3.5, and their best ensemble. For the BERT-Large model, incorporating UMLS led to an average F1 score improvement of 40.36%, while using GPT-3.5 resulted in a comparable average increase of 40.52%. The Best-Ensemble approach using BERT-Large achieved the highest improvement, with an average increase of 42.29%. DANN model's F1 score improved by 22.74% on average using the UMLS-only approach. The GPT-3.5-based method resulted in a 21.53% increase, and the Best-Ensemble DANN model showed a more notable improvement, with an average increase of 25.03%. Our proposed HILGEN approach improves NER performance in few-shot settings without requiring additional manually annotated data. Our experiments demonstrate that an effective strategy for optimizing biomedical NER is to combine biomedical knowledge curated in the past, such as the UMLS, and generative LLMs to create synthetic training instances. Our future research will focus on exploring additional innovative synthetic data generation strategies for further improving NER performance.

The emergence of powerful LLMs has led to a paradigm shift in Natural Language Understanding and Natural Language Generation. The properties that make LLMs so valuable for these tasks -- creativity, ability to produce fluent speech, and ability to quickly and effectively abstract information from large corpora -- also present new challenges to evaluating their outputs. The rush to market has led teams to fall back on quick, cost-effective automatic evaluations which offer value, but do not obviate the need for human judgments in model training and evaluation. This paper argues that in cases in which end users need to agree with the decisions made by ML models -- e.g. in toxicity detection or extraction of main points for summarization -- models should be trained and evaluated on data that represent the preferences of those users. We support this argument by explicating the role of human feedback in labeling and judgment tasks for model training and evaluation. First, we propose methods for disentangling noise from signal in labeling tasks. Then we show that noise in labeling disagreement can be minimized by adhering to proven methodological best practices, while signal can be maximized to play an integral role in model training and evaluation tasks. Finally, we illustrate best practices by providing a case study in which two guardrails classifiers are evaluated using human judgments to align final model behavior to user preferences. We aim for this paper to provide researchers and professionals with guidelines to integrating human judgments into their ML and generative AI evaluation toolkit, particularly when working toward achieving accurate and unbiased features that align with users' needs and expectations.

Large language model (LLM)-based tools such as ChatGPT seem useful for classical programming assignments. The more specialized the field, the more likely they lack reliability because of the lack of data to train them. In the case of quantum computing, the quality of answers of generic chatbots is low. C4Q is a chatbot focused on quantum programs that addresses this challenge through a software architecture that integrates specialized LLMs to classify requests and specialized question answering modules with a deterministic logical engine to provide trustworthy quantum computing support. This article describes the latest version (2.0) of C4Q, which delivers several enhancements: ready-to-run Qiskit code for gate definitions and circuit operations, expanded features to solve software engineering tasks such as the travelling salesperson problem and the knapsack problem, and a feedback mechanism for iterative improvement. Extensive testing of the backend confirms the system's reliability, while empirical evaluations show that C4Q 2.0's classification LLM reaches near-perfect accuracy. The evaluation of the result consists in a comparative study with three existing chatbots highlighting C4Q 2.0's maintainability and correctness, reflecting on how software architecture decisions, such as separating deterministic logic from probabilistic text generation impact the quality of the results.

Generative Large Language Models (LLMs) have shown promising results in text annotation using zero-shot and few-shot learning. Yet these approaches do not allow the model to retain information from previous annotations, making each response independent from the preceding ones. This raises the question of whether model memory -- the LLM having knowledge about its own previous annotations in the same task -- affects performance. In this article, using OpenAI's GPT-4o and Meta's Llama 3.1 on two political science datasets, we demonstrate that allowing the model to retain information about its own previous classifications yields significant performance improvements: between 5 and 25\% when compared to zero-shot and few-shot learning. Moreover, memory reinforcement, a novel approach we propose that combines model memory and reinforcement learning, yields additional performance gains in three out of our four tests. These findings have important implications for applied researchers looking to improve performance and efficiency in LLM annotation tasks.

Recently, large language models (LLMs) with hundreds of billions of parameters have demonstrated the emergent ability, surpassing traditional methods in various domains even without fine-tuning over domain-specific data. However, when it comes to financial sentiment analysis (FSA)$\unicode{x2013}$a fundamental task in financial AI$\unicode{x2013}$these models often encounter various challenges, such as complex financial terminology, subjective human emotions, and ambiguous inclination expressions. In this paper, we aim to answer the fundamental question: whether LLMs are good in-context learners for FSA? Unveiling this question can yield informative insights on whether LLMs can learn to address the challenges by generalizing in-context demonstrations of financial document-sentiment pairs to the sentiment analysis of new documents, given that finetuning these models on finance-specific data is difficult, if not impossible at all. To the best of our knowledge, this is the first paper exploring in-context learning for FSA that covers most modern LLMs (recently released DeepSeek V3 included) and multiple in-context sample selection methods. Comprehensive experiments validate the in-context learning capability of LLMs for FSA.

This paper reflects on the process of performing Thematic Analysis with Large Language Models (LLMs). Specifically, the paper deals with the problem of analytical saturation of initial codes, as produced by LLMs. Thematic Analysis is a well-established qualitative analysis method composed of interlinked phases. A key phase is the initial coding, where the analysts assign labels to discrete components of a dataset. Saturation is a way to measure the validity of a qualitative analysis and relates to the recurrence and repetition of initial codes. In the paper we reflect on how well LLMs achieve analytical saturation and propose also a novel technique to measure Inductive Thematic Saturation (ITS). This novel technique leverages a programming framework called DSPy. The proposed novel approach allows a precise measurement of ITS.

Large Visual Language Models (LVLMs) increasingly rely on preference alignment to ensure reliability, which steers the model behavior via preference fine-tuning on preference data structured as ``image - winner text - loser text'' triplets. However, existing approaches often suffer from limited diversity and high costs associated with human-annotated preference data, hindering LVLMs from fully achieving their intended alignment capabilities. We present \projectname, a self-supervised framework capable of transforming the already abundant supervised text-image pairs into holistic preference triplets for more effective and cheaper LVLM alignment, eliminating the need for human preference annotations. Our approach facilitates LVLMs in progressively enhancing alignment capabilities through iterative self-improvement. The key design rationale is to devise preference triplets where the winner text consistently improves in holisticness and outperforms the loser response in quality, thereby pushing the model to ``strive to the utmost'' of alignment performance through preference fine-tuning. For each given text-image pair, SHAPE introduces multiple visual augmentations and pairs them with a summarized text to serve as the winner response, while designating the original text as the loser response. Experiments across \textbf{12} benchmarks on various model architectures and sizes, including LLaVA and DeepSeek-VL, show that SHAPE achieves significant gains, for example, achieving +11.3\% on MMVet (comprehensive evaluation), +1.4\% on MMBench (general VQA), and +8.0\% on POPE (hallucination robustness) over baselines in 7B models. Notably, qualitative analyses confirm enhanced attention to visual details and better alignment with human preferences for holistic descriptions.

Despite extensive safety enhancements in large language models (LLMs), multi-turn "jailbreak" conversations crafted by skilled human adversaries can still breach even the most sophisticated guardrails. However, these multi-turn attacks demand considerable manual effort, limiting their scalability. In this work, we introduce a novel approach called Multi-turn-to-Single-turn (M2S) that systematically converts multi-turn jailbreak prompts into single-turn attacks. Specifically, we propose three conversion strategies - Hyphenize, Numberize, and Pythonize - each preserving sequential context yet packaging it in a single query. Our experiments on the Multi-turn Human Jailbreak (MHJ) dataset show that M2S often increases or maintains high Attack Success Rates (ASRs) compared to original multi-turn conversations. Notably, using a StrongREJECT-based evaluation of harmfulness, M2S achieves up to 95.9% ASR on Mistral-7B and outperforms original multi-turn prompts by as much as 17.5% in absolute improvement on GPT-4o. Further analysis reveals that certain adversarial tactics, when consolidated into a single prompt, exploit structural formatting cues to evade standard policy checks. These findings underscore that single-turn attacks - despite being simpler and cheaper to conduct - can be just as potent, if not more, than their multi-turn counterparts. Our findings underscore the urgent need to reevaluate and reinforce LLM safety strategies, given how adversarial queries can be compacted into a single prompt while still retaining sufficient complexity to bypass existing safety measures.

We rigorously establish a bipartite mutual information scaling law in natural language that governs long-range dependencies. This scaling law, which we show is distinct from and scales independently of the conventional two-point mutual information, is the key to understanding long-context language modeling. Using this scaling law, we formulate the Long-context Language Modeling (L$^2$M) condition, which relates a model's capacity for effective long context length modeling to the scaling of its latent state size for storing past information. Our results are validated through experiments on both transformers and state space models. This work establishes a theoretical foundation that guides the development of large language models toward longer context lengths.

Recent advancements in speech-to-speech dialogue systems leverage LLMs for multimodal interactions, yet they remain hindered by fine-tuning requirements, high computational overhead, and text-speech misalignment. Existing speech-enabled LLMs often degrade conversational quality by modifying the LLM, thereby compromising its linguistic capabilities. In contrast, we propose LLMVoX, a lightweight 30M-parameter, LLM-agnostic, autoregressive streaming TTS system that generates high-quality speech with low latency, while fully preserving the capabilities of the base LLM. Our approach achieves a significantly lower Word Error Rate compared to speech-enabled LLMs, while operating at comparable latency and UTMOS score. By decoupling speech synthesis from LLM processing via a multi-queue token streaming system, LLMVoX supports seamless, infinite-length dialogues. Its plug-and-play design also facilitates extension to various tasks with different backbones. Furthermore, LLMVoX generalizes to new languages with only dataset adaptation, attaining a low Character Error Rate on an Arabic speech task. Additionally, we have integrated LLMVoX with a Vision-Language Model to create an omni-model with speech, text, and vision capabilities, without requiring additional multimodal training. Our code base and project page is available at https://mbzuai-oryx.github.io/LLMVoX .