Controllable Conversational Theme Detection Track at DSTC 12

Conversational analytics has been on the forefront of transformation driven by the advances in Speech and Natural Language Processing techniques. Rapid adoption of Large Language Models (LLMs) in the analytics field has taken the problems that can be automated to a new level of complexity and scale. In this paper, we introduce Theme Detection as a critical task in conversational analytics, aimed at automatically identifying and categorizing topics within conversations. This process can significantly reduce the manual effort involved in analyzing expansive dialogs, particularly in domains like customer support or sales. Unlike traditional dialog intent detection, which often relies on a fixed set of intents for downstream system logic, themes are intended as a direct, user-facing summary of the conversation's core inquiry. This distinction allows for greater flexibility in theme surface forms and user-specific customizations. We pose Controllable Conversational Theme Detection problem as a public competition track at Dialog System Technology Challenge (DSTC) 12 -- it is framed as joint clustering and theme labeling of dialog utterances, with the distinctive aspect being controllability of the resulting theme clusters' granularity achieved via the provided user preference data. We give an overview of the problem, the associated dataset and the evaluation metrics, both automatic and human. Finally, we discuss the participant teams' submissions and provide insights from those. The track materials (data and code) are openly available in the GitHub repository.

Interpretable Topic Extraction and Word Embedding Learning using row-stochastic DEDICOM

The DEDICOM algorithm provides a uniquely interpretable matrix factorization method for symmetric and asymmetric square matrices. We employ a new row-stochastic variation of DEDICOM on the pointwise mutual information matrices of text corpora to identify latent topic clusters within the vocabulary and simultaneously learn interpretable word embeddings. We introduce a method to efficiently train a constrained DEDICOM algorithm and a qualitative evaluation of its topic modeling and word embedding performance.

Taking the Garbage Out of Data-Driven Prediction Across Climate Timescales

Artificial intelligence (AI) -- and specifically machine learning (ML) -- applications for climate prediction across timescales are proliferating quickly. The emergence of these methods prompts a revisit to the impact of data preprocessing, a topic familiar to the climate community, as more traditional statistical models work with relatively small sample sizes. Indeed, the skill and confidence in the forecasts produced by data-driven models are directly influenced by the quality of the datasets and how they are treated during model development, thus yielding the colloquialism "garbage in, garbage out." As such, this article establishes protocols for the proper preprocessing of input data for AI/ML models designed for climate prediction (i.e., subseasonal to decadal and longer). The three aims are to: (1) educate researchers, developers, and end users on the effects that preprocessing has on climate predictions; (2) provide recommended practices for data preprocessing for such applications; and (3) empower end users to decipher whether the models they are using are properly designed for their objectives. Specific topics covered in this article include the creation of (standardized) anomalies, dealing with non-stationarity and the spatiotemporally correlated nature of climate data, and handling of extreme values and variables with potentially complex distributions. Case studies will illustrate how using different preprocessing techniques can produce different predictions from the same model, which can create confusion and decrease confidence in the overall process. Ultimately, implementing the recommended practices set forth in this article will enhance the robustness and transparency of AI/ML in climate prediction studies.

Why Synthetic Isn't Real Yet: A Diagnostic Framework for Contact Center Dialogue Generation

Synthetic transcript generation is critical in contact center domains, where privacy and data scarcity limit model training and evaluation. Unlike prior synthetic dialogue generation work on open-domain or medical dialogues, contact center conversations are goal-oriented, role-asymmetric, and behaviorally complex, featuring disfluencies, ASR noise, and compliance-driven agent actions. In deployments where transcripts are unavailable, standard pipelines still yield derived call attributes such as Intent Summaries, Topic Flow, and QA Evaluation Forms. We leverage these as supervision signals to guide generation. To assess the quality of such outputs, we introduce a diagnostic framework of 18 linguistically and behaviorally grounded metrics for comparing real and synthetic transcripts. We benchmark four language-agnostic generation strategies, from simple prompting to characteristic-aware multi-stage approaches, alongside reference-free baselines. Results reveal persistent challenges: no method excels across all traits, with notable deficits in disfluency, sentiment, and behavioral realism. Our diagnostic tool exposes these gaps, enabling fine-grained evaluation and stress testing of synthetic dialogue across languages.

A Hybrid AI Methodology for Generating Ontologies of Research Topics from Scientific Paper Corpora

Taxonomies and ontologies of research topics (e.g., MeSH, UMLS, CSO, NLM) play a central role in providing the primary framework through which intelligent systems can explore and interpret the literature. However, these resources have traditionally been manually curated, a process that is time-consuming, prone to obsolescence, and limited in granularity. This paper presents Sci-OG, a semi-auto\-mated methodology for generating research topic ontologies, employing a multi-step approach: 1) Topic Discovery, extracting potential topics from research papers; 2) Relationship Classification, determining semantic relationships between topic pairs; and 3) Ontology Construction, refining and organizing topics into a structured ontology. The relationship classification component, which constitutes the core of the system, integrates an encoder-based language model with features describing topic occurrence in the scientific literature. We evaluate this approach against a range of alternative solutions using a dataset of 21,649 manually annotated semantic triples. Our method achieves the highest F1 score (0.951), surpassing various competing approaches, including a fine-tuned SciBERT model and several LLM baselines, such as the fine-tuned GPT4-mini. Our work is corroborated by a use case which illustrates the practical application of our system to extend the CSO ontology in the area of cybersecurity. The presented solution is designed to improve the accessibility, organization, and analysis of scientific knowledge, thereby supporting advancements in AI-enabled literature management and research exploration.

Adaptive Backtracking for Privacy Protection in Large Language Models

The preservation of privacy has emerged as a critical topic in the era of artificial intelligence. However, current work focuses on user-oriented privacy, overlooking severe enterprise data leakage risks exacerbated by the Retrieval-Augmented Generation paradigm. To address this gap, our paper introduces a novel objective: enterprise-oriented privacy concerns. Achieving this objective requires overcoming two fundamental challenges: existing methods such as data sanitization severely degrade model performance, and the field lacks public datasets for evaluation. We address these challenges with several solutions. (1) To prevent performance degradation, we propose ABack, a training-free mechanism that leverages a Hidden State Model to pinpoint the origin of a leakage intention and rewrite the output safely. (2) To solve the lack of datasets, we construct PriGenQA, a new benchmark for enterprise privacy scenarios in healthcare and finance. To ensure a rigorous evaluation, we move beyond simple static attacks by developing a powerful adaptive attacker with Group Relative Policy Optimization. Experiments show that against this superior adversary, ABack improves the overall privacy utility score by up to 15\% over strong baselines, avoiding the performance trade-offs of prior methods.

Designing Object Detection Models for TinyML: Foundations, Comparative Analysis, Challenges, and Emerging Solutions

Object detection (OD) has become vital for numerous computer vision applications, but deploying it on resource-constrained IoT devices presents a significant challenge. These devices, often powered by energy-efficient microcontrollers, struggle to handle the computational load of deep learning-based OD models. This issue is compounded by the rapid proliferation of IoT devices, predicted to surpass 150 billion by 2030. TinyML offers a compelling solution by enabling OD on ultra-low-power devices, paving the way for efficient and real-time processing at the edge. Although numerous survey papers have been published on this topic, they often overlook the optimization challenges associated with deploying OD models in TinyML environments. To address this gap, this survey paper provides a detailed analysis of key optimization techniques for deploying OD models on resource-constrained devices. These techniques include quantization, pruning, knowledge distillation, and neural architecture search. Furthermore, we explore both theoretical approaches and practical implementations, bridging the gap between academic research and real-world edge artificial intelligence deployment. Finally, we compare the key performance indicators (KPIs) of existing OD implementations on microcontroller devices, highlighting the achieved maturity level of these solutions in terms of both prediction accuracy and efficiency. We also provide a public repository to continually track developments in this fast-evolving field: https://github.com/christophezei/Optimizing-Object-Detection-Models-for-TinyML-A-Comprehensive-Survey.

Efficient Agent: Optimizing Planning Capability for Multimodal Retrieval Augmented Generation

Multimodal Retrieval-Augmented Generation (mRAG) has emerged as a promising solution to address the temporal limitations of Multimodal Large Language Models (MLLMs) in real-world scenarios like news analysis and trending topics. However, existing approaches often suffer from rigid retrieval strategies and under-utilization of visual information. To bridge this gap, we propose E-Agent, an agent framework featuring two key innovations: a mRAG planner trained to dynamically orchestrate multimodal tools based on contextual reasoning, and a task executor employing tool-aware execution sequencing to implement optimized mRAG workflows. E-Agent adopts a one-time mRAG planning strategy that enables efficient information retrieval while minimizing redundant tool invocations. To rigorously assess the planning capabilities of mRAG systems, we introduce the Real-World mRAG Planning (RemPlan) benchmark. This novel benchmark contains both retrieval-dependent and retrieval-independent question types, systematically annotated with essential retrieval tools required for each instance. The benchmark's explicit mRAG planning annotations and diverse question design enhance its practical relevance by simulating real-world scenarios requiring dynamic mRAG decisions. Experiments across RemPlan and three established benchmarks demonstrate E-Agent's superiority: 13% accuracy gain over state-of-the-art mRAG methods while reducing redundant searches by 37%.

Learning the Topic, Not the Language: How LLMs Classify Online Immigration Discourse Across Languages

Large language models (LLMs) are transforming social-science research by enabling scalable, precise analysis. Their adaptability raises the question of whether knowledge acquired through fine-tuning in a few languages can transfer to unseen languages that only appeared during pre-training. To examine this, we fine-tune lightweight LLaMA 3.2-3B models on monolingual, bilingual, or multilingual data sets to classify immigration-related tweets from X/Twitter across 13 languages, a domain characterised by polarised, culturally specific discourse. We evaluate whether minimal language-specific fine-tuning enables cross-lingual topic detection and whether adding targeted languages corrects pre-training biases. Results show that LLMs fine-tuned in one or two languages can reliably classify immigration-related content in unseen languages. However, identifying whether a tweet expresses a pro- or anti-immigration stance benefits from multilingual fine-tuning. Pre-training bias favours dominant languages, but even minimal exposure to under-represented languages during fine-tuning (as little as $9.62\times10^{-11}$ of the original pre-training token volume) yields significant gains. These findings challenge the assumption that cross-lingual mastery requires extensive multilingual training: limited language coverage suffices for topic-level generalisation, and structural biases can be corrected with lightweight interventions. By releasing 4-bit-quantised, LoRA fine-tuned models, we provide an open-source, reproducible alternative to proprietary LLMs that delivers 35 times faster inference at just 0.00000989% of the dollar cost of the OpenAI GPT-4o model, enabling scalable, inclusive research.

The Problem with Safety Classification is not just the Models

Studying the robustness of Large Language Models (LLMs) to unsafe behaviors is an important topic of research today. Building safety classification models or guard models, which are fine-tuned models for input/output safety classification for LLMs, is seen as one of the solutions to address the issue. Although there is a lot of research on the safety testing of LLMs themselves, there is little research on evaluating the effectiveness of such safety classifiers or the evaluation datasets used for testing them, especially in multilingual scenarios. In this position paper, we demonstrate how multilingual disparities exist in 5 safety classification models by considering datasets covering 18 languages. At the same time, we identify potential issues with the evaluation datasets, arguing that the shortcomings of current safety classifiers are not only because of the models themselves. We expect that these findings will contribute to the discussion on developing better methods to identify harmful content in LLM inputs across languages.