PolERo: Studying Political Evasion in Romanian

Political evasion refers to responses that engage with a question while withholding the requested information. Recent NLP work frames political evasion as a classification task using a two-level taxonomy of response clarity and fine-grained evasion strategies. Existing work on response clarity and evasion classification is limited to English, leaving open whether the taxonomy and model behavior transfer across languages and political contexts. We introduce PolERo, a dataset of 3,574 human-annotated question-answer pairs extracted from official transcripts of five Romanian presidents. We evaluate multiple classification approaches on both datasets under matched conditions, including TF-IDF baselines, fine-tuned encoder models, a proposed sliding-window encoder, and zero/few-shot LLM prompting. We study cross-lingual transfer through joint bilingual training and machine-translation-based data augmentation. Our results indicate that fine-tuned encoders are competitive, cross-lingual transfer is asymmetric, and ambivalent evasion categories involving pragmatic cues remain the main challenge across all model families.

Not All Fallbacks Are Failures: Understanding and Recovering from Fallbacks in Mobile Voice Assistants

Robust understanding of user input is a core requirement for voice assistants deployed in real-world environments. In practice, these systems encounter heterogeneous fallback situations caused by noisy audio input, transcription errors, ambiguous requests, incomplete utterances, or unintended activations. Existing systems typically respond with generic fallback messages, which do not resolve the underlying interaction failure and can degrade user experience. We study fallback handling in a deployed smartwatch-based voice assistant for general health support in everyday environments. Our analysis is based on six months of real-world usage data from more than 500 users, yielding a dataset of 3,030 anonymized, naturally occurring fallback-triggering utterances. We contribute (1) an operational taxonomy and the annotated VoxFallbacks dataset of these interactions, (2) a comparative evaluation of different models within a classification pipeline under practical deployment constraints, and (3) practical lessons for designing robust and cost-efficient fallback mechanisms. Results show that lightweight embedding-based classifiers outperform larger generative models on most classification tasks while requiring substantially fewer computational resources.

Knowing Before Answering: Decoding Language Models for Reliable RAG

In Retrieval-Augmented Generation (RAG), retrieval may provide insufficient or conflicting information needed to answer a question. The system should not only know when to answer but also be able to identify cases in which the documents provided in RAG are insufficient or contain conflicting information. This can be framed as a three-way classification problem, where we use the model's internal signals to determine whether the provided information in the input can be classified as sufficient, insufficient, or conflicting. We create a controlled benchmark dataset that replicates a RAG setup with fictitious information and labels each instance as answerable, insufficient, or conflicting. We use hidden activations and attention-derived features as inputs to train a lightweight linear model to distinguish among the three classes. Across 16 language models spanning different architectures and a range of model sizes, our feature-based router consistently outperforms prompting-based baselines and the performance of specialised RAG-models. We further conduct analyses into the information dynamics of the models. We show that the most informative signals for the classification are available in the middle layers, with hidden activation states being more effective than attention values or the MLP-feature outputs in most of the tested models. Overall, our results suggest that language models internally encode whether retrieved evidence is sufficient to support answering, and that this signal can be decoded reliably for RAG triage.

Evaluation Pitfalls and Sparsity Limitations in LLM-based Confidence Estimates for Classification

Confidence estimation is essential when LLMs are used for classification, indicating when predictions can be trusted. However, common approaches such as verbalization produce extremely sparse outputs. For instance, Qwen3-32B verbalizes only eight unique confidence values on SST-2, with over half being exactly 95%, a pattern we observe consistently across four datasets and two LLMs. Besides limiting practical utility, we show that this sparsity critically affects evaluation: the choice of interpolation in area under the accuracy-rejection curve (AUARC) dramatically alters rankings, with consistency sampling dropping from best to worst under stepwise versus linear interpolation. We advocate for standardizing stepwise interpolation for a fairer comparison. Under such a fair evaluation, we find that weighting verbalized digits by token probabilities, a method we term verbalization logprobs, addresses sparsity and achieves the best AUARC (+2.3 points over vanilla verbalization) without incurring additional inference cost.

Predicting Multilingual Classification and Translation Performance of LLMs with Cross-Lingual Alignment $\unicode{x2013}$ Is English Enough?

Multilingual large language models (LLMs) have been shown to perform better on non-English classification tasks when the representations of the given language are more aligned to English within the model. Several cross-lingual alignment (CLA) scores have been proposed for use with LLMs, along with multiple approaches for extracting embeddings from the models. We provide a comparative analysis of 27 CLA score variants, examining how they differ and how well each predicts downstream performance across three tasks. Crucially, while LLMs are widely used for generative tasks such as machine translation, prior work has focused almost exclusively on classification. We therefore investigate whether CLA scores are similarly predictive of translation performance. To enable computing correlations across target languages, we propose a PMI-based translation metric, which is less dependent on the target language and correlates strongly with chrF. We find that CLA with English predicts translation quality comparably to or better than source-target CLA, providing new evidence that LLMs use English as an internal pivot language.

Structured Data Extraction from Real Estate Documents using Clustering, Classification, and Large Language Models

Real estate property listings expose structured metadata through the API. Still, the richest property-level information (i.e., legal status, structural condition, utility supplies, heating systems) sits in attached questionnaire documents that no automated system currently processes at scale. These documents are heterogeneous. Some are digitally generated with selectable text, others are scanned physical forms. There are even more complex layouts that contain checkbox annotations that defeat conventional text extraction. In this paper, we present an end-to-end pipeline for acquiring, classifying, and extracting structured data from selectable text documents. The pipeline was applied to 3965 questionnaire documents collected from a live property platform via reverse-engineered REST APIs. First, we classified each document into one of three structural categories (text_only, scanned, and special_char), then extracted 35 predefined property attributes from eligible documents using DeepSeek R1 as the Large Language Model, prompted to return a structured JSON object. All 2781 submitted documents were processed successfully, producing a final dataset of 2766 unique property records. Downstream validation confirmed the data quality. Cosine similarity matching achieves a Jaccard consistency score of 0.82, and K-Means clustering produces interpretable market segments with a silhouette score of 0.2088. Results show that the proposed extraction from each property document is both feasible and reliable at this scale.

Comparing Architectures for Supervised Political Scaling

Text scaling, the task of positioning political actors on an ideological scale, is a fundamental task in political analysis. To ease the need for manual analysis, various NLP methods have been proposed for this task, including classification- and regression-based approaches, showing successes as well as limitations. The goal of our paper is to consolidate the state of the art in this area. We ask two questions: (a) Can the performance of scaling methods be improved by predicting scales not individually but jointly? (b) Is there a middle ground between classification and regression?

AnySimLite: A Lightweight Few-Shot Similarity Encoder for On-Device Speech-Adjacent Classification

To minimize privacy concerns and inference latency on edge devices like smartphones, lightweight on-device models remain important for end-user applications. Many of these applications involve natural language classification, but deploying multiple specialized models creates a memory footprint challenge. We investigate: Can a single lightweight architecture solve multiple Speech-Adjacent (SA) classification tasks through reduction to a nuanced text similarity formulation? We propose AnySimLite, a lightweight similarity encoder that combines word-level and character-level channels. Together with a dataset transformation strategy, we evaluate AnySimLite across multiple SA classification tasks and show that it consistently achieves state-of-the-art (SOTA) or SOTA-competitive performance in few-shot settings while maintaining a low memory footprint. Even in the worst case, the performance drop remains below 7% while using $<\frac{1}{250}^{\mathrm{th}}$ of the model size of the SOTA qLLaMA_LoRA-7B baseline.

Supervision versus Demonstration-Based In-Context Learning for Multiword Expression Classification

Turkish idiomatic light verb constructions (LVCs) are challenging for multiword expression processing because they often share the same surface form as fully literal verb-object combinations while functioning as a single, partially idiomatic predicate. We frame Turkish LVC detection as a binary classification task (literal meaning vs. idiomatic meaning) and evaluate on a manually created controlled set (N=147) with matched negatives: out-of-domain random sentences and in-domain literal controls (NLVC), alongside LVC positives. We compare a supervised Turkish encoder baseline (BERTurk with a classifier head) to three instruction-tuned LLMs from different families under zero-shot, one-shot, and few-shot prompting, and analyze how demonstrations shift error profiles. In zero-shot, LLMs perform well on negatives but show very low LVC recall. One-shot prompting sharply improves LVC detection but can induce strong, model-specific biases, leading models to overpredict or underpredict LVCs. A richer few-shot prompt improves calibration and yields robust overall performance for GPT-OSS-20B and Qwen 2.5-14B. Overall, the results highlight substantial prompt sensitivity in Turkish metalinguistic classification: the supervised baseline remains competitive, while prompted LLMs can match or exceed it on LVCs with carefully constructed demonstrations.