Detecting AI-generated text (AIGT) remains challenging because existing approaches rely on token-level statistical signals or independent stylometric features, causing them to overfit to specific generators and fail under distribution shift. We identify a structural signal at the sentence-pair level: LLMs produce inter-sentence transition variance that deviates from human writing through inflated variance driven by recurring similarity bursts at paragraph boundaries and templated transitions. We formalize this as Relational Over-Regularization (ROR) and validate it across four benchmarks (p < 0.001). To exploit this signal, we propose the Cross-Source Stylometric Fingerprint Graph (CSFG), a graph-based framework that encodes positional, sequential, semantic, and transition deviation signals as learnable GNN edge features. We emphasize that the core contribution is this relational problem formulation itself, not the GNN architecture - CSFG is one concrete instantiation for operationalizing ROR. The per-edge signed deviation 𝛿_ij operationalizes ROR without hand-crafted thresholds and acts as a false-positive calibrator. CSFG achieves 97.14% accuracy under binary detection, outperforming the strongest graph-based baseline by 11.14 pp, with a false-positive rate of 1.57% and robust generalization to unseen LLMs in the inflated-variance regime; detection degrades for generators whose transition variance falls at or below the human baseline.

Large language models are increasingly used in emotionally consequential interactions. Response-level evaluation, however, struggles to diagnose how empathy fails across turns. We introduce MATE (Multi-turn Assessment of calibraTed Empathy), a framework that evaluates empathy as a multi-turn, perception-centered process. Across two controlled studies (N=82), baseline responses exhibit disclosure-insensitive miscalibration: generic validation and formulaic tone read as hollow across turns. A prompt-level self-critique condition makes these patterns less prominent but coincides with a different failure, agreement-skewed miscalibration, where affirmation becomes insufficiently contingent on disclosure context under high persona alignment (62.5% of fit participants in Study 2). Response-level analysis is consistent with a tone–agreement dissociation, as the condition shifts tone toward naturalness while agreement density also increases (+98% under fit vs. +25% under unfit). These findings reframe empathic behavior as relational calibration—consistency among disclosure depth, persona, and response behavior—rather than maximization. Response behavior itself spans tone, agreement, and contextual specificity, sub-dimensions that can drift independently.

Multi-agent systems utilizing large language models often assign authoritative roles to improve performance, yet the impact of authority bias on agent interactions remains underexplored. We present the first systematic analysis of role-based authority bias in free-form multi-agent evaluation using ChatEval. Applying French and Raven’s power-based theory, we classify authoritative roles into legitimate, referent, and expert types and analyze their influence across 12-turn conversations. Experiments with GPT-4o and DeepSeek R1 reveal that Expert and Referent power roles exert stronger influence than Legitimate power roles. Crucially, authority bias emerges not through active conformity by general agents, but through authoritative roles consistently maintaining their positions while general agents demonstrate flexibility. Furthermore, authority influence requires clear position statements, as neutral responses fail to generate bias. These findings provide key insights for designing multi-agent frameworks with asymmetric interaction patterns.

Whisper exposes speech through a fixed 1500-token encoder interface, now a default representation for ASR decoders and Whisper-based speech language models (SpeechLMs), yet its redundancy remains largely unexamined. We propose stride-k subsampling, a deterministic indexing operation that retains every k-th token after the convolutional stem or encoder transformer. Across five Whisper scales, k=2 preserves baseline WER at both positions, with CKA attributing this stability to acoustic overlap at the stem and attention-induced redistribution at the encoder output. Applying stride-2 at both positions cuts audio tokens by 75% and total GFLOPs by 52–58%, with small WER costs on most ASR benchmarks and larger costs on harder ones. The same configuration extends to three Whisper-based SpeechLMs, yielding modest accuracy drops on stronger baselines and larger drops on weaker ones, while reducing end-to-end latency by 19.6–27.4%. Requiring no training or auxiliary computation, stride-k subsampling suggests that the standard Whisper audio-token interface is over-provisioned for many downstream uses.

AI text detectors are increasingly employed in academic settings, but it remains unclear whether their outputs reflect AI authorship itself or broader linguistic features associated with polished academic English. Previous studies have reported high false-positive rates (FPRs) for non-native English writing, but population-level comparisons confound authorship with differences in topic, domain, and writing style. Professional editing provides a useful setting for examining this issue because it changes the linguistic form of manuscripts while preserving authorship and content. We examined 135,389 document pairs from a professional English editing service (2018–2025), comprising non-native manuscripts and their native-edited versions, to assess how editing affects detector responses controlling for content and authorship. For the 13 AI text detectors, FPRs for human-written texts varied widely, from 0.0% to 100.0%. Responses varied across detectors: the same edits increased AI scores in some detectors but decreased them in others. Notably, score changes correlated with the extent of editing. The findings identify professional editing style as a key confounding variable in AI detector outputs, rather than establishing a full separation of text origin from linguistic style, raising concerns about fairness and reliability in academic settings.

As large language models (LLMs) generate more human-like texts, concerns about the side effects of AI-generated texts (AIGT) have grown. So, researchers have developed methods for detecting AIGT. However, two challenges remain. First, the performance of detecting black-box LLMs is low because existing models focus on probabilistic features. Second, most AIGT detectors have been tested on a single-candidate setting, which assumes that we know the origin of an AIGT and which may deviate from the real-world scenario. To resolve these challenges, we propose DART, which consists of four steps: rephrasing, semantic parsing, scoring, and multiclass classification. We conducted three experiments to test the performance of DART. The experimental result shows that DART can discriminate multiple black-box LLMs without probabilistic features and the origin of AIGT.

Large Language Models (LLMs) have been recently adopted in interactive systems requiring communication. As the false belief in a model can harm the usability of such systems, LLMs should not have cognitive biases that humans have. Psychologists especially focus on the False Consensus Effect (FCE), a cognitive bias where individuals overestimate the extent to which others share their beliefs or behaviors, because FCE can distract smooth communication by posing false beliefs. However, previous studies have less examined FCE in LLMs thoroughly, which needs more consideration of confounding biases, general situations, and prompt changes. Therefore, in this paper, we conduct two studies to examine the FCE phenomenon in LLMs. In Study 1, we investigate whether LLMs have FCE. In Study 2, we explore how various prompting styles affect the demonstration of FCE. As a result of these studies, we identified that popular LLMs have FCE. Also, the result specifies the conditions when FCE becomes more or less prevalent compared to normal usage.

We introduce VoiceBBQ, a spoken extension of the BBQ (Bias Benchmark for Question answering) - a dataset that measures social bias by presenting ambiguous or disambiguated contexts followed by questions that may elicit stereotypical responses. Due to the nature of speech modality, social bias in Spoken Language Models (SLMs) can emerge from two distinct sources: 1) content aspect and 2) acoustic aspect. The dataset converts every BBQ context into controlled voice conditions, enabling per-axis accuracy, bias, and consistency scores that remain comparable to the original text benchmark. Using VoiceBBQ, we evaluate two SLMs—LLaMA-Omni and Qwen2-Audio—and observe architectural contrasts: LLaMA-Omni retains strong acoustic sensitivity, amplifying gender and accent bias, whereas Qwen2-Audio substantially dampens these cues while preserving content fidelity. VoiceBBQ thus provides a compact, drop-in testbed for jointly diagnosing content and acoustic bias across spoken language models.

In this study, we investigate soundness of current Abstract Meaning Representation (AMR) similarity metrics in terms of equivalence and inequivalence. Specifically, AMR guidelines provide several equivalence and inequivalence conditions to reflect the meaning aspect of the semantics. Thus, it is important to examine an AMR metric’s soundness, i.e., whether the metric correctly reflects the guidelines. However, the existing metrics have less investigated their soundness. In this work, we propose a new experimental method using simulated data and a series of statistical tests to verify the metric’s soundness. Our experimental result revealed that all existing metrics such as Smatch, SemBLEU, S2match, Smatch++, WWLK-theta, WWLK-k3e2n, and SEMA did not fully meet the AMR guidelines in terms of equivalence and inequivalence aspects. Also, to alleviate this soundness problem, we suggest a revised metric called Smatch#, which adopts simple graph standardization technique that can improve the soundness of an existing metric.

In this paper, we propose a neural model EPT-X (Expression-Pointer Transformer with Explanations), which utilizes natural language explanations to solve an algebraic word problem. To enhance the explainability of the encoding process of a neural model, EPT-X adopts the concepts of plausibility and faithfulness which are drawn from math word problem solving strategies by humans. A plausible explanation is one that includes contextual information for the numbers and variables that appear in a given math word problem. A faithful explanation is one that accurately represents the reasoning process behind the model’s solution equation. The EPT-X model yields an average baseline performance of 69.59% on our PEN dataset and produces explanations with quality that is comparable to human output. The contribution of this work is two-fold. (1) EPT-X model: An explainable neural model that sets a baseline for algebraic word problem solving task, in terms of model’s correctness, plausibility, and faithfulness. (2) New dataset: We release a novel dataset PEN (Problems with Explanations for Numbers), which expands the existing datasets by attaching explanations to each number/variable.

Solving algebraic word problems has recently emerged as an important natural language processing task. To solve algebraic word problems, recent studies suggested neural models that generate solution equations by using ‘Op (operator/operand)’ tokens as a unit of input/output. However, such a neural model suffered two issues: expression fragmentation and operand-context separation. To address each of these two issues, we propose a pure neural model, Expression-Pointer Transformer (EPT), which uses (1) ‘Expression’ token and (2) operand-context pointers when generating solution equations. The performance of the EPT model is tested on three datasets: ALG514, DRAW-1K, and MAWPS. Compared to the state-of-the-art (SoTA) models, the EPT model achieved a comparable performance accuracy in each of the three datasets; 81.3% on ALG514, 59.5% on DRAW-1K, and 84.5% on MAWPS. The contribution of this paper is two-fold; (1) We propose a pure neural model, EPT, which can address the expression fragmentation and the operand-context separation. (2) The fully automatic EPT model, which does not use hand-crafted features, yields comparable performance to existing models using hand-crafted features, and achieves better performance than existing pure neural models by at most 40%.

Math word problem solving is an emerging research topic in Natural Language Processing. Recently, to address the math word problem-solving task, researchers have applied the encoder-decoder architecture, which is mainly used in machine translation tasks. The state-of-the-art neural models use hand-crafted features and are based on generation methods. In this paper, we propose the GEO (Generation of Equations by utilizing Operators) model that does not use hand-crafted features and addresses two issues that are present in existing neural models: 1. missing domain-specific knowledge features and 2. losing encoder-level knowledge. To address missing domain-specific feature issue, we designed two auxiliary tasks: operation group difference prediction and implicit pair prediction. To address losing encoder-level knowledge issue, we added an Operation Feature Feed Forward (OP3F) layer. Experimental results showed that the GEO model outperformed existing state-of-the-art models on two datasets, 85.1% in MAWPS, and 62.5% in DRAW-1K, and reached comparable performance of 82.1% in ALG514 dataset.