Are You Listening to Me? Fine-Tuning Chatbots for Empathetic Dialogue

Conversational agents have made significant progress since ELIZA, expanding their role across various domains, including healthcare, education, and customer service. As these agents become increasingly integrated into daily human interactions, the need for emotional intelligence, particularly empathetic listening, becomes increasingly essential. In this study, we explore how Large Language Models (LLMs) respond when tasked with generating emotionally rich interactions. Starting from a small dataset manually crafted by an expert to reflect empathic behavior, we extended the conversations using two LLMs: ChatGPT and Gemini. We analyzed the emotional progression of the dialogues using both sentiment analysis (via VADER) and expert assessments. While the generated conversations often mirrored the intended emotional structure, human evaluation revealed important differences in the perceived empathy and coherence of the responses. These findings suggest that emotion modeling in dialogues requires not only structural alignment in the expressed emotions but also qualitative depth, highlighting the importance of combining automated and humancentered methods in the development of emotionally competent agents.

Adapting Language Models to Indonesian Local Languages: An Empirical Study of Language Transferability on Zero-Shot Settings

In this paper, we investigate the transferability of pre-trained language models to low-resource Indonesian local languages through the task of sentiment analysis. We evaluate both zero-shot performance and adapter-based transfer on ten local languages using models of different types: a monolingual Indonesian BERT, multilingual models such as mBERT and XLM-R, and a modular adapter-based approach called MAD-X. To better understand model behavior, we group the target languages into three categories: seen (included during pre-training), partially seen (not included but linguistically related to seen languages), and unseen (absent and unrelated in pre-training data). Our results reveal clear performance disparities across these groups: multilingual models perform best on seen languages, moderately on partially seen ones, and poorly on unseen languages. We find that MAD-X significantly improves performance, especially for seen and partially seen languages, without requiring labeled data in the target language. Additionally, we conduct a further analysis on tokenization and show that while subword fragmentation and vocabulary overlap with Indonesian correlate weakly with prediction quality, they do not fully explain the observed performance. Instead, the most consistent predictor of transfer success is the model's prior exposure to the language, either directly or through a related language.

Is External Information Useful for Stance Detection with LLMs?

In the stance detection task, a text is classified as either favorable, opposing, or neutral towards a target. Prior work suggests that the use of external information, e.g., excerpts from Wikipedia, improves stance detection performance. However, whether or not such information can benefit large language models (LLMs) remains an unanswered question, despite their wide adoption in many reasoning tasks. In this study, we conduct a systematic evaluation on how Wikipedia and web search external information can affect stance detection across eight LLMs and in three datasets with 12 targets. Surprisingly, we find that such information degrades performance in most cases, with macro F1 scores dropping by up to 27.9\%. We explain this through experiments showing LLMs' tendency to align their predictions with the stance and sentiment of the provided information rather than the ground truth stance of the given text. We also find that performance degradation persists with chain-of-thought prompting, while fine-tuning mitigates but does not fully eliminate it. Our findings, in contrast to previous literature on BERT-based systems which suggests that external information enhances performance, highlight the risks of information biases in LLM-based stance classifiers. Code is available at https://github.com/ngqm/acl2025-stance-detection.

Emotionally Intelligent Task-oriented Dialogue Systems: Architecture, Representation, and Optimisation

Task-oriented dialogue (ToD) systems are designed to help users achieve specific goals through natural language interaction. While recent advances in large language models (LLMs) have significantly improved linguistic fluency and contextual understanding, building effective and emotionally intelligent ToD systems remains a complex challenge. Effective ToD systems must optimise for task success, emotional understanding and responsiveness, and precise information conveyance, all within inherently noisy and ambiguous conversational environments. In this work, we investigate architectural, representational, optimisational as well as emotional considerations of ToD systems. We set up systems covering these design considerations with a challenging evaluation environment composed of a natural-language user simulator coupled with an imperfect natural language understanding module. We propose \textbf{LUSTER}, an \textbf{L}LM-based \textbf{U}nified \textbf{S}ystem for \textbf{T}ask-oriented dialogue with \textbf{E}nd-to-end \textbf{R}einforcement learning with both short-term (user sentiment) and long-term (task success) rewards. Our findings demonstrate that combining LLM capability with structured reward modelling leads to more resilient and emotionally responsive ToD systems, offering a practical path forward for next-generation conversational agents.

Stereotype Detection as a Catalyst for Enhanced Bias Detection: A Multi-Task Learning Approach

Bias and stereotypes in language models can cause harm, especially in sensitive areas like content moderation and decision-making. This paper addresses bias and stereotype detection by exploring how jointly learning these tasks enhances model performance. We introduce StereoBias, a unique dataset labeled for bias and stereotype detection across five categories: religion, gender, socio-economic status, race, profession, and others, enabling a deeper study of their relationship. Our experiments compare encoder-only models and fine-tuned decoder-only models using QLoRA. While encoder-only models perform well, decoder-only models also show competitive results. Crucially, joint training on bias and stereotype detection significantly improves bias detection compared to training them separately. Additional experiments with sentiment analysis confirm that the improvements stem from the connection between bias and stereotypes, not multi-task learning alone. These findings highlight the value of leveraging stereotype information to build fairer and more effective AI systems.

MEGA: xLSTM with Multihead Exponential Gated Fusion for Precise Aspect-based Sentiment Analysis

Aspect-based Sentiment Analysis (ABSA) is a critical Natural Language Processing (NLP) task that extracts aspects from text and determines their associated sentiments, enabling fine-grained analysis of user opinions. Existing ABSA methods struggle to balance computational efficiency with high performance: deep learning models often lack global context, transformers demand significant computational resources, and Mamba-based approaches face CUDA dependency and diminished local correlations. Recent advancements in Extended Long Short-Term Memory (xLSTM) models, particularly their efficient modeling of long-range dependencies, have significantly advanced the NLP community. However, their potential in ABSA remains untapped. To this end, we propose xLSTM with Multihead Exponential Gated Fusion (MEGA), a novel framework integrating a bi-directional mLSTM architecture with forward and partially flipped backward (PF-mLSTM) streams. The PF-mLSTM enhances localized context modeling by processing the initial sequence segment in reverse with dedicated parameters, preserving critical short-range patterns. We further introduce an mLSTM-based multihead cross exponential gated fusion mechanism (MECGAF) that dynamically combines forward mLSTM outputs as query and key with PF-mLSTM outputs as value, optimizing short-range dependency capture while maintaining global context and efficiency. Experimental results on three benchmark datasets demonstrate that MEGA outperforms state-of-the-art baselines, achieving superior accuracy and efficiency in ABSA tasks.

FINN-GL: Generalized Mixed-Precision Extensions for FPGA-Accelerated LSTMs

Recurrent neural networks (RNNs), particularly LSTMs, are effective for time-series tasks like sentiment analysis and short-term stock prediction. However, their computational complexity poses challenges for real-time deployment in resource constrained environments. While FPGAs offer a promising platform for energy-efficient AI acceleration, existing tools mainly target feed-forward networks, and LSTM acceleration typically requires full custom implementation. In this paper, we address this gap by leveraging the open-source and extensible FINN framework to enable the generalized deployment of LSTMs on FPGAs. Specifically, we leverage the Scan operator from the Open Neural Network Exchange (ONNX) specification to model the recurrent nature of LSTM computations, enabling support for mixed quantisation within them and functional verification of LSTM-based models. Furthermore, we introduce custom transformations within the FINN compiler to map the quantised ONNX computation graph to hardware blocks from the HLS kernel library of the FINN compiler and Vitis HLS. We validate the proposed tool-flow by training a quantised ConvLSTM model for a mid-price stock prediction task using the widely used dataset and generating a corresponding hardware IP of the model using our flow, targeting the XCZU7EV device. We show that the generated quantised ConvLSTM accelerator through our flow achieves a balance between performance (latency) and resource consumption, while matching (or bettering) inference accuracy of state-of-the-art models with reduced precision. We believe that the generalisable nature of the proposed flow will pave the way for resource-efficient RNN accelerator designs on FPGAs.

Commander-GPT: Dividing and Routing for Multimodal Sarcasm Detection

Multimodal sarcasm understanding is a high-order cognitive task. Although large language models (LLMs) have shown impressive performance on many downstream NLP tasks, growing evidence suggests that they struggle with sarcasm understanding. In this paper, we propose Commander-GPT, a modular decision routing framework inspired by military command theory. Rather than relying on a single LLM's capability, Commander-GPT orchestrates a team of specialized LLM agents where each agent will be selectively assigned to a focused sub-task such as context modeling, sentiment analysis, etc. Their outputs are then routed back to the commander, which integrates the information and performs the final sarcasm judgment. To coordinate these agents, we introduce three types of centralized commanders: (1) a trained lightweight encoder-based commander (e.g., multi-modal BERT); (2) four small autoregressive language models, serving as moderately capable commanders (e.g., DeepSeek-VL); (3) two large LLM-based commander (Gemini Pro and GPT-4o) that performs task routing, output aggregation, and sarcasm decision-making in a zero-shot fashion. We evaluate Commander-GPT on the MMSD and MMSD 2.0 benchmarks, comparing five prompting strategies. Experimental results show that our framework achieves 4.4% and 11.7% improvement in F1 score over state-of-the-art (SoTA) baselines on average, demonstrating its effectiveness.

MFTCXplain: A Multilingual Benchmark Dataset for Evaluating the Moral Reasoning of LLMs through Hate Speech Multi-hop Explanation

Ensuring the moral reasoning capabilities of Large Language Models (LLMs) is a growing concern as these systems are used in socially sensitive tasks. Nevertheless, current evaluation benchmarks present two major shortcomings: a lack of annotations that justify moral classifications, which limits transparency and interpretability; and a predominant focus on English, which constrains the assessment of moral reasoning across diverse cultural settings. In this paper, we introduce MFTCXplain, a multilingual benchmark dataset for evaluating the moral reasoning of LLMs via hate speech multi-hop explanation using Moral Foundation Theory (MFT). The dataset comprises 3,000 tweets across Portuguese, Italian, Persian, and English, annotated with binary hate speech labels, moral categories, and text span-level rationales. Empirical results highlight a misalignment between LLM outputs and human annotations in moral reasoning tasks. While LLMs perform well in hate speech detection (F1 up to 0.836), their ability to predict moral sentiments is notably weak (F1 < 0.35). Furthermore, rationale alignment remains limited mainly in underrepresented languages. These findings show the limited capacity of current LLMs to internalize and reflect human moral reasoning.

Quantifying Fairness in LLMs Beyond Tokens: A Semantic and Statistical Perspective

Large Language Models (LLMs) often generate responses with inherent biases, undermining their reliability in real-world applications. Existing evaluation methods often overlook biases in long-form responses and the intrinsic variability of LLM outputs. To address these challenges, we propose FiSCo(Fine-grained Semantic Computation), a novel statistical framework to evaluate group-level fairness in LLMs by detecting subtle semantic differences in long-form responses across demographic groups. Unlike prior work focusing on sentiment or token-level comparisons, FiSCo goes beyond surface-level analysis by operating at the claim level, leveraging entailment checks to assess the consistency of meaning across responses. We decompose model outputs into semantically distinct claims and apply statistical hypothesis testing to compare inter- and intra-group similarities, enabling robust detection of subtle biases. We formalize a new group counterfactual fairness definition and validate FiSCo on both synthetic and human-annotated datasets spanning gender, race, and age. Experiments show that FiSco more reliably identifies nuanced biases while reducing the impact of stochastic LLM variability, outperforming various evaluation metrics.