arXiv subjects

Smitha Muthya Sudheendra

Publications and source records attributed to Smitha Muthya Sudheendra.

6 recordsLinked to original sources

arXiv2026

When Decodability Is Not Enough: Logical Validity Representations, Behavioral Dissociation, and Causal Tests in Language Models

Large language models can look capable of logical reasoning, but correct or incorrect answers alone tell us little about what the model represents internally. We study logical verification in five open-weight transformer models using matched valid--invalid premise--claim pairs that vary across inference families, semantic domains, templates, and difficulty levels. Despite near-chance behavioral performance, logical validity is often almost perfectly decodable from hidden states and remains strongly decodable under held-out templates, domains, and inference families. Validity also remains highly decodable on behaviorally incorrect examples in the conditions where correctness-conditioned evaluation is well defined. At the same time, exhaustive leave-one-out tests reveal clear limits to this generalization, and interventions along probe-derived validity directions have only weak, nonspecific effects compared with random controls. Our results suggest that representing validity, expressing it in behavior, and using it causally are distinct. Validity related information can be strongly decodable from a model's hidden states without being reliably expressed in its output.

cs.CL↗

arXiv2026

PragAlign: Feedback-Guided Pragmatic Alignment for Controlled Synthetic Dialogue Generation

Synthetic dialogue generation can support research in privacy-restricted service settings, but generated conversations must preserve communicative intent, affective meaning, and natural dialogue flow. We introduce PragAlign, a feedback-guided framework for controlled synthetic dialogue generation conditioned on service context, target intent, and target emotion, with auxiliary trait-style controls. PragAlign uses a generate--evaluate--revise loop in which an LLM-based evaluator scores intent alignment, emotion alignment, coherence, fluency, and aggregate quality, then provides criterion-specific feedback for up to three refinement rounds. On 800 matched dialogue specifications, PragAlign achieves 99.50\% evaluator-defined acceptance, compared with 72.25\% for one-shot generation and 95.88\% for repeated generation without structured feedback. This indicates that repeated attempts account for much of the gain over one-shot generation, while structured feedback primarily improves last-mile multi-constraint satisfaction rather than broad average quality. Refinement gains are concentrated in emotion alignment, which is also the dominant failure mode in ablations. A separate human evaluation of 1,200 generated dialogues shows that intent expression and dialogue flow are highly recognizable to annotators, while emotion appropriateness is less stable and more subjective. These results support PragAlign as a quality-control framework for improving evaluator-defined communicative constraint satisfaction, while showing that affective realization and independent human-perceived quality remain open challenges.

cs.CL↗

arXiv2026

ReasonScaffold: A Scaffolded Reasoning-based Annotation Protocol for Human-AI Co-Annotation

Human annotation is central to NLP evaluation, yet subjective tasks often exhibit substantial variability across annotators. While large language models (LLMs) can provide structured reasoning to support annotation, their influence on human annotation behavior remains underexplored. We introduce \textbf{ReasonScaffold}, a scaffolded reasoning annotation protocol that exposes LLM-generated explanations while withholding predicted labels. We study how reasoning affects human annotation behavior in a controlled setting, rather than evaluating annotation accuracy. Using a two-pass protocol inspired by Delphi-style revision, annotators first label instances independently and then revise their decisions after viewing model-generated reasoning. We evaluate the approach on sentiment classification and opinion detection tasks, analyzing changes in inter-annotator agreement and revision behavior. To quantify these effects, we introduce the Annotator Effort Proxy (AEP), a metric capturing the proportion of labels revised after exposure to reasoning. Our results show that exposure to reasoning is associated with increased agreement, along with minimal revision, suggesting that reasoning helps resolve ambiguous cases without inducing widespread changes. These findings provide insight into how reasoning explanations shape annotation consistency and highlight reasoning-based scaffolds as a practical mechanism for human--AI co-annotation workflows.

cs.CL↗

arXiv2026

CMV-Fuse: Cross Modal-View Fusion of AMR, Syntax, and Knowledge Representations for Aspect Based Sentiment Analysis

Natural language understanding inherently depends on integrating multiple complementary perspectives spanning from surface syntax to deep semantics and world knowledge. However, current Aspect-Based Sentiment Analysis (ABSA) systems typically exploit isolated linguistic views, thereby overlooking the intricate interplay between structural representations that humans naturally leverage. We propose CMV-Fuse, a Cross-Modal View fusion framework that emulates human language processing by systematically combining multiple linguistic perspectives. Our approach systematically orchestrates four linguistic perspectives: Abstract Meaning Representations, constituency parsing, dependency syntax, and semantic attention, enhanced with external knowledge integration. Through hierarchical gated attention fusion across local syntactic, intermediate semantic, and global knowledge levels, CMV-Fuse captures both fine-grained structural patterns and broad contextual understanding. A novel structure aware multi-view contrastive learning mechanism ensures consistency across complementary representations while maintaining computational efficiency. Extensive experiments demonstrate substantial improvements over strong baselines on standard benchmarks, with analysis revealing how each linguistic view contributes to more robust sentiment analysis.

cs.CL↗

arXiv2025

Beyond Citations: A Cross-Domain Metric for Dataset Impact and Shareability

The scientific community increasingly relies on open data sharing, yet existing metrics inadequately capture the true impact of datasets as research outputs. Traditional measures, such as the h-index, focus on publications and citations but fail to account for dataset accessibility, reuse, and cross-disciplinary influence. We propose the X-index, a novel author-level metric that quantifies the value of data contributions through a two-step process: (i) computing a dataset-level value score (V-score) that integrates breadth of reuse, FAIRness, citation impact, and transitive reuse depth, and (ii) aggregating V-scores into an author-level X-index. Using datasets from computational social science, medicine, and crisis communication, we validate our approach against expert ratings, achieving a strong correlation. Our results demonstrate that the X-index provides a transparent, scalable, and low-cost framework for assessing data-sharing practices and incentivizing open science. The X-index encourages sustainable data-sharing practices and gives institutions, funders, and platforms a tangible way to acknowledge the lasting influence of research datasets.

cs.CY↗

arXiv2024

Impact Of Emotions on Information Seeking And Sharing Behaviors During Pandemic

We propose a novel approach to assess the public's coping behavior during the COVID-19 outbreak by examining the emotions. Specifically, we explore (1) changes in the public's emotions with the COVID-19 crisis progression and (2) the impacts of the public's emotions on their information-seeking, information-sharing behaviors, and compliance with stay-at-home policies. We base the study on the appraisal tendency framework, detect the public's emotions by fine-tuning a pre-trained RoBERTa model, and cross-analyze third-party behavioral data. We demonstrate the feasibility and reliability of our proposed approach in providing a large-scale examination of the publi's emotions and coping behaviors in a real-world crisis: COVID-19. The approach complements prior crisis communication research, mainly based on self-reported, small-scale experiments and survey data. Our results show that anger and fear are more prominent than other emotions experienced by the public at the pandemic's outbreak stage. Results also show that the extent of low certainty and passive emotions (e.g., sadness, fear) was related to increased information-seeking and information-sharing behaviors. Additionally, high-certainty (e.g., anger) and low-certainty (e.g., sadness, fear) emotions during the outbreak correlated to the public's compliance with stay-at-home orders.

cs.SI↗