Tag
AI Historian is an AI agent system that helps historians organize and verify person-centred temporal clues from dispersed historical narratives, reducing the cost of historical research while achieving high accuracy in temporal localization.
The article demonstrates a Python/Flask example using Telnyx Call Control and AI inference to create a natural language IVR system, allowing callers to verbally state their needs instead of navigating fixed menus.
This paper presents the first application of data science to evaluate the UK Honours system using natural language processing, introducing a novel sentiment analysis algorithm called Minos to assess public opinion on honours recipients.
This paper presents a production-grade framework that uses large language models to convert natural-language pricing policies into executable decisions for tourism pricing, achieving significant efficiency gains and auditability in real-world deployment.
This paper investigates the interpretability of DAPF-based models for dementia detection, revealing that while DAPF achieves strong performance, its token-level explanations lack faithfulness.
This paper introduces a causal graph-based attention mechanism to enhance retrieval precision in Retrieval-Augmented Generation (RAG) systems, showing improvements in keyword-stuffing regimes of proprietary knowledge bases.
The paper introduces Peony, the first benchmark for evaluating large language models on comprehending poetic logic in modern Chinese poetry, and evaluates six mainstream LLMs, revealing their limitations in this specialized task.
The paper introduces an ontology-driven framework to quantify and enforce structural consistency in document-level relation extraction datasets, reducing logical contradictions and improving model generalization when using distant supervision data.
This paper introduces ImmigrationReason, a large-scale structured dataset of U.S. immigration appeals for legal reasoning research, addressing the gap in administrative adjudication data for NLP studies.
This study develops an ambiguity taxonomy to evaluate large language model performance on clinical registry abstraction from unprocessed EMR data, finding that LLM accuracy is significantly lower than human abstractors and declines as task ambiguity increases.
The article discusses the finding that AI agents are currently utilizing five times more tokens than humans, indicating a notable trend in AI capabilities and efficiency.
A comparative study of BART, BERT, and RoBERTa for text summarization, examining their architectures and suitability for extractive and abstractive summarization tasks.
The paper introduces a failure-aware adversarial retrieval-augmented framework using contextual bandits to improve robustness in natural language understanding, with significant improvements on benchmarks like SNLI, ANLI, and MultiNLI.
This paper explores how LLMs can act as persona-guided agents in task-oriented dialogue, analyzing the effects of personality adaptation on task performance and user satisfaction through evaluations with models like GPT-4o and Gemini 2.0 Flash.
LongNovel introduces a multi-scale benchmark for evaluating hallucinations in long-context novel summarization, built from Chinese and English novels to improve LLM assessment.
Polaris trains LLMs to generate table descriptions directly from retrieval feedback using DPO, significantly outperforming previous methods like AutoDDG in table retrieval tasks.
This study demonstrates that large language models with institution-specific prompting can recover protected health information missed by existing de-identification systems, enhancing data privacy compliance in electronic health records.
The paper proposes a neurosymbolic architecture for learning to assemble novel structures by incorporating semantic constraints from natural language and visual demonstrations, achieving more data-efficient online adaptation.
Sol AI can understand and answer questions about large codebases in one reading, but it struggles to write meaningful comments for code.
A user tweets about their experience asking Opus 5, an AI model, to teach them about Buddhism, commenting on the concept of 'seamslop' which likely refers to AI-generated content quality.