Hateful Person or Hateful Model? Investigating the Role of Personas in Hate Speech Detection by Large Language Models

Hate speech detection is a socially sensitive and inherently subjective task, with judgments often varying based on personal traits. While prior work has examined how socio-demographic factors influence annotation, the impact of personality traits on Large Language Models (LLMs) remains largely unexplored. In this paper, we present the first comprehensive study on the role of persona prompts in hate speech classification, focusing on MBTI-based traits. A human annotation survey confirms that MBTI dimensions significantly affect labeling behavior. Extending this to LLMs, we prompt four open-source models with MBTI personas and evaluate their outputs across three hate speech datasets. Our analysis uncovers substantial persona-driven variation, including inconsistencies with ground truth, inter-persona disagreement, and logit-level biases. These findings highlight the need to carefully define persona prompts in LLM-based annotation workflows, with implications for fairness and alignment with human values.

Look Within or Look Beyond? A Theoretical Comparison Between Parameter-Efficient and Full Fine-Tuning

Parameter-Efficient Fine-Tuning (PEFT) methods achieve performance comparable to Full Fine-Tuning (FFT) while requiring significantly fewer computing resources, making it the go-to choice for researchers. We find that although PEFT can achieve competitive results on some benchmarks, its performance falls short of FFT in complex tasks, such as reasoning and instruction-based fine-tuning. In this paper, we compare the characteristics of PEFT and FFT in terms of representational capacity and robustness based on optimization theory. We theoretically demonstrate that PEFT is a strict subset of FFT. By providing theoretical upper bounds for PEFT, we show that the limited parameter space constrains the model's representational ability, making it more susceptible to perturbations. Experiments on 15 datasets encompassing classification, generation, reasoning, instruction fine-tuning tasks and 11 adversarial test sets validate our theories. We hope that these results spark further research beyond the realms of well established PEFT. The source code is in the anonymous Github repository\footnote{https://github.com/misonsky/PEFTEval}.

Mechanistic Understanding and Mitigation of Language Confusion in English-Centric Large Language Models

Language confusion -- where large language models (LLMs) generate unintended languages against the user's need -- remains a critical challenge, especially for English-centric models. We present the first mechanistic interpretability (MI) study of language confusion, combining behavioral benchmarking with neuron-level analysis. Using the Language Confusion Benchmark (LCB), we show that confusion points (CPs) -- specific positions where language switches occur -- are central to this phenomenon. Through layer-wise analysis with TunedLens and targeted neuron attribution, we reveal that transition failures in the final layers drive confusion. We further demonstrate that editing a small set of critical neurons, identified via comparative analysis with multilingual-tuned models, substantially mitigates confusion without harming general competence or fluency. Our approach matches multilingual alignment in confusion reduction for most languages and yields cleaner, higher-quality outputs. These findings provide new insights into the internal dynamics of LLMs and highlight neuron-level interventions as a promising direction for robust, interpretable multilingual language modeling.

Tracing Multilingual Factual Knowledge Acquisition in Pretraining

Large Language Models (LLMs) are capable of recalling multilingual factual knowledge present in their pretraining data. However, most studies evaluate only the final model, leaving the development of factual recall and crosslingual consistency throughout pretraining largely unexplored. In this work, we trace how factual recall and crosslingual consistency evolve during pretraining, focusing on OLMo-7B as a case study. We find that both accuracy and consistency improve over time for most languages. We show that this improvement is primarily driven by the fact frequency in the pretraining corpus: more frequent facts are more likely to be recalled correctly, regardless of language. Yet, some low-frequency facts in non-English languages can still be correctly recalled. Our analysis reveals that these instances largely benefit from crosslingual transfer of their English counterparts -- an effect that emerges predominantly in the early stages of pretraining. We pinpoint two distinct pathways through which multilingual factual knowledge acquisition occurs: (1) frequency-driven learning, which is dominant and language-agnostic, and (2) crosslingual transfer, which is limited in scale and typically constrained to relation types involving named entities. We release our code and data to facilitate further research at https://github.com/cisnlp/multilingual-fact-tracing.

Lost in Multilinguality: Dissecting Cross-lingual Factual Inconsistency in Transformer Language Models

Multilingual language models (MLMs) store factual knowledge across languages but often struggle to provide consistent responses to semantically equivalent prompts in different languages. While previous studies point out this cross-lingual inconsistency issue, the underlying causes remain unexplored. In this work, we use mechanistic interpretability methods to investigate cross-lingual inconsistencies in MLMs. We find that MLMs encode knowledge in a language-independent concept space through most layers, and only transition to language-specific spaces in the final layers. Failures during the language transition often result in incorrect predictions in the target language, even when the answers are correct in other languages. To mitigate this inconsistency issue, we propose a linear shortcut method that bypasses computations in the final layers, enhancing both prediction accuracy and cross-lingual consistency. Our findings shed light on the internal mechanisms of MLMs and provide a lightweight, effective strategy for producing more consistent factual outputs.

XCOMPS: A Multilingual Benchmark of Conceptual Minimal Pairs

We introduce XCOMPS in this work, a multilingual conceptual minimal pair dataset covering 17 languages. Using this dataset, we evaluate LLMs' multilingual conceptual understanding through metalinguistic prompting, direct probability measurement, and neurolinguistic probing. By comparing base, instruction-tuned, and knowledge-distilled models, we find that: 1) LLMs exhibit weaker conceptual understanding for low-resource languages, and accuracy varies across languages despite being tested on the same concept sets. 2) LLMs excel at distinguishing concept-property pairs that are visibly different but exhibit a marked performance drop when negative pairs share subtle semantic similarities. 3) Instruction tuning improves performance in concept understanding but does not enhance internal competence; knowledge distillation can enhance internal competence in conceptual understanding for low-resource languages with limited gains in explicit task performance. 4) More morphologically complex languages yield lower concept understanding scores and require deeper layers for conceptual reasoning.

Language Model Re-rankers are Steered by Lexical Similarities

Language model (LM) re-rankers are used to refine retrieval results for retrieval-augmented generation (RAG). They are more expensive than lexical matching methods like BM25 but assumed to better process semantic information. To understand whether LM re-rankers always live up to this assumption, we evaluate 6 different LM re-rankers on the NQ, LitQA2 and DRUID datasets. Our results show that LM re-rankers struggle to outperform a simple BM25 re-ranker on DRUID. Leveraging a novel separation metric based on BM25 scores, we explain and identify re-ranker errors stemming from lexical dissimilarities. We also investigate different methods to improve LM re-ranker performance and find these methods mainly useful for NQ. Taken together, our work identifies and explains weaknesses of LM re-rankers and points to the need for more adversarial and realistic datasets for their evaluation.

Large Language Models as Neurolinguistic Subjects: Identifying Internal Representations for Form and Meaning

This study investigates the linguistic understanding of Large Language Models (LLMs) regarding signifier (form) and signified (meaning) by distinguishing two LLM evaluation paradigms: psycholinguistic and neurolinguistic. Traditional psycholinguistic evaluations often reflect statistical biases that may misrepresent LLMs' true linguistic capabilities. We introduce a neurolinguistic approach, utilizing a novel method that combines minimal pair and diagnostic probing to analyze activation patterns across model layers. This method allows for a detailed examination of how LLMs represent form and meaning, and whether these representations are consistent across languages. Our contributions are three-fold: (1) We compare neurolinguistic and psycholinguistic methods, revealing distinct patterns in LLM assessment; (2) We demonstrate that LLMs exhibit higher competence in form compared to meaning, with the latter largely correlated to the former; (3) We present new conceptual minimal pair datasets for Chinese (COMPS-ZH) and German (COMPS-DE), complementing existing English datasets.

BMIKE-53: Investigating Cross-Lingual Knowledge Editing with In-Context Learning

Large language models (LLMs) possess extensive parametric knowledge, but this knowledge is difficult to update with new information because retraining is very expensive and infeasible for closed-source models. Knowledge editing (KE) has emerged as a viable solution for updating the knowledge of LLMs without compromising their overall performance. On-the-fly KE methods, inspired by in-context learning (ICL), have shown great promise and allow LLMs to be treated as black boxes. In the past, KE was primarily employed in English contexts, whereas the potential for cross-lingual KE in current English-centric LLMs has not been fully explored. To foster more research in this direction, we introduce the BMIKE-53 benchmark for evaluating cross-lingual KE on 53 diverse languages across three KE task types. We also propose a gradient-free KE method called Multilingual In-context Knowledge Editing (MIKE) and evaluate it on BMIKE-53. Our evaluation focuses on cross-lingual knowledge transfer in terms of reliability, generality, locality, and portability, offering valuable insights and a framework for future research in cross-lingual KE. Our code and data are publicly accessible via the anonymous repository at https://anonymous.4open.science/r/MIKE.

A Unified Data Augmentation Framework for Low-Resource Multi-Domain Dialogue Generation

Current state-of-the-art dialogue systems heavily rely on extensive training datasets. However, challenges arise in domains where domain-specific training datasets are insufficient or entirely absent. To tackle this challenge, we propose a novel data \textbf{A}ugmentation framework for \textbf{M}ulti-\textbf{D}omain \textbf{D}ialogue \textbf{G}eneration, referred to as \textbf{AMD$^2$G}. The AMD$^2$G framework consists of a data augmentation process and a two-stage training approach: domain-agnostic training and domain adaptation training. We posit that domain corpora are a blend of domain-agnostic and domain-specific features, with certain representation patterns shared among diverse domains. Domain-agnostic training aims to enable models to learn these common expressive patterns. To construct domain-agnostic dialogue corpora, we employ a \textit{\textbf{de-domaining}} data processing technique used to remove domain-specific features. By mitigating the effects of domain-specific features, the model trained on the de-domained corpora can effectively learn common expression patterns in different domains. Subsequently, we adapt the learned domain-agnostic features to the target domain through domain adaptation training. We conduct experiments on Chinese dialogue datasets from five different domains and show that AMD$^2$G achieves superior performance compared to both direct training on the target domain corpus and collective training on all five domain corpora. Our work underscores AMD$^2$G as a viable alternative solution for low-resource multi-domain dialogue generation. Code and data associated with our work are available on GitHub repository$^{\text 1}$.