Source author record

Sharath Chandra Guntuku

Sharath Chandra Guntuku appears in the imported research catalog. Authorship, coauthor and topic links are available while profile ownership is still unclaimed.

ResearcherUnclaimed source record

Computation and Language Social and Information Networks Human-Computer Interaction Cryptography and Security cs.CY Machine Learning Multimedia Networking and Internet Architecture

Catalog footprint

What is connected

7works

8topics

4close collaborators

Actions

Connect this record

Open graph Browse works

Inspect adjacent papers, topics, institutions and collaborators without losing the researcher page.

Building this map preview

BZPEER is loading the nearby papers, people, topics and institutions for this page.

preprint2026arXiv

A Concise Agent is Less Expert: Revealing Side Effects of Using Style Features on Conversational Agents

Style features such as friendly, helpful, or concise are widely used in prompts to steer the behavior of Large Language Model (LLM) conversational agents, yet their unintended side effects remain poorly understood. In this work, we present the first systematic study of cross-feature stylistic side effects. We conduct a comprehensive survey of 127 conversational agent papers from ACL Anthology and identify 12 frequently used style features. Using controlled, synthetic dialogues across task-oriented and open domain settings, we quantify how prompting for one style feature causally affects others via a pairwise LLM as a Judge evaluation framework. Our results reveal consistent and structured side effects, such as prompting for conciseness significantly reduces perceived expertise. They demonstrate that style features are deeply entangled rather than orthogonal. To support future research, we introduce CASSE (Conversational Agent Stylistic Side Effects), a dataset capturing these complex interactions. We further evaluate prompt based and activation steering based mitigation strategies and find that while they can partially restore suppressed traits, they often degrade the primary intended style. These findings challenge the assumption of faithful style control in LLMs and highlight the need for multi-objective and more principled approaches to safe, targeted stylistic steering in conversational agents.

preprint2026arXiv

Conceptors for Semantic Steering

Activation-based steering provides control of LLM behavior at inference time, but the dominant paradigm reduces each concept to a single direction whose geometry is left largely unexamined. Rather than selecting a single steering direction, we use conceptors: soft projection matrices estimated from activations pooled across both poles of a bipolar concept, which preserve the concept's full multidimensional subspace. A geometric analysis shows the bipolar subspace strictly subsumes the single-vector baseline. We further show that the conceptor quota provides a parameter-free layer-selection diagnostic, predicting concept separability with Pearson correlations up to r=0.96 across three instruction-tuned models and three semantic dimensions. Beyond selection, conceptors admit a closed-form Boolean algebra (AND, OR, NOT): we evaluate conceptor compositionality on thematically related sub-concepts. Across a systematic five-axis design-space evaluation, conceptors match or outperform additive baselines at layers where concept subspaces are multi-dimensional while producing substantially fewer degenerate outputs. Conceptor steering is a geometrically principled, compositional, and practically safer alternative to single-direction steering from a limited number of contrastive pairs.

preprint2022arXiv

A Weakly-Supervised Iterative Graph-Based Approach to Retrieve COVID-19 Misinformation Topics

The COVID-19 pandemic has been accompanied by an `infodemic' -- of accurate and inaccurate health information across social media. Detecting misinformation amidst dynamically changing information landscape is challenging; identifying relevant keywords and posts is arduous due to the large amount of human effort required to inspect the content and sources of posts. We aim to reduce the resource cost of this process by introducing a weakly-supervised iterative graph-based approach to detect keywords, topics, and themes related to misinformation, with a focus on COVID-19. Our approach can successfully detect specific topics from general misinformation-related seed words in a few seed texts. Our approach utilizes the BERT-based Word Graph Search (BWGS) algorithm that builds on context-based neural network embeddings for retrieving misinformation-related posts. We utilize Latent Dirichlet Allocation (LDA) topic modeling for obtaining misinformation-related themes from the texts returned by BWGS. Furthermore, we propose the BERT-based Multi-directional Word Graph Search (BMDWGS) algorithm that utilizes greater starting context information for misinformation extraction. In addition to a qualitative analysis of our approach, our quantitative analyses show that BWGS and BMDWGS are effective in extracting misinformation-related content compared to common baselines in low data resource settings. Extracting such content is useful for uncovering prevalent misconceptions and concerns and for facilitating precision public health messaging campaigns to improve health behaviors.

preprint2022arXiv

Twitter Corpus of the #BlackLivesMatter Movement And Counter Protests: 2013 to 2021

Black Lives Matter (BLM) is a decentralized social movement protesting violence against Black individuals and communities, with a focus on police brutality. The movement gained significant attention following the killings of Ahmaud Arbery, Breonna Taylor, and George Floyd in 2020. The #BlackLivesMatter social media hashtag has come to represent the grassroots movement, with similar hashtags counter protesting the BLM movement, such as #AllLivesMatter, and #BlueLivesMatter. We introduce a data set of 63.9 million tweets from 13.0 million users from over 100 countries which contain one of the following keywords: BlackLivesMatter, AllLivesMatter, and BlueLivesMatter. This data set contains all currently available tweets from the beginning of the BLM movement in 2013 to 2021. We summarize the data set and show temporal trends in use of both the BlackLivesMatter keyword and keywords associated with counter movements. Additionally, for each keyword, we create and release a set of Latent Dirichlet Allocation (LDA) topics (i.e., automatically clustered groups of semantically co-occuring words) to aid researchers in identifying linguistic patterns across the three keywords.

preprint2020arXiv

Studying Politeness across Cultures Using English Twitter and Mandarin Weibo

Modeling politeness across cultures helps to improve intercultural communication by uncovering what is considered appropriate and polite. We study the linguistic features associated with politeness across US English and Mandarin Chinese. First, we annotate 5,300 Twitter posts from the US and 5,300 Sina Weibo posts from China for politeness scores. Next, we develop an English and Chinese politeness feature set, `PoliteLex'. Combining it with validated psycholinguistic dictionaries, we then study the correlations between linguistic features and perceived politeness across cultures. We find that on Mandarin Weibo, future-focusing conversations, identifying with a group affiliation, and gratitude are considered to be more polite than on English Twitter. Death-related taboo topics, lack of or poor choice of pronouns, and informal language are associated with higher impoliteness on Mandarin Weibo compared to English Twitter. Finally, we build language-based machine learning models to predict politeness with an F1 score of 0.886 on Mandarin Weibo and a 0.774 on English Twitter.

preprint2016arXiv

Personality, Culture, and System Factors - Impact on Affective Response to Multimedia

Whilst affective responses to various forms and genres of multimedia content have been well researched, precious few studies have investigated the combined impact that multimedia system parameters and human factors have on affect. Consequently, in this paper we explore the role that two primordial dimensions of human factors - personality and culture - in conjunction with system factors - frame rate, resolution, and bit rate - have on user affect and enjoyment of multimedia presentations. To this end, a two-site, cross-cultural study was undertaken, the results of which produced three predictve models. Personality and Culture traits were shown statistically to represent 5.6% of the variance in positive affect, 13.6% in negative affect and 9.3% in enjoyment. The correlation between affect and enjoyment, was significant. Predictive modeling incorporating human factors showed about 8%, 7% and 9% improvement in predicting positive affect, negative affect and enjoyment respectively when compared to models trained only on system factors. Results and analysis indicate the significant role played by human factors in influencing affect that users experience while watching multimedia.

preprint2013arXiv

Real-time Peer-to-Peer Botnet Detection Framework based on Bayesian Regularized Neural Network

Over the past decade, the Cyberspace has seen an increasing number of attacks coming from botnets using the Peer-to-Peer (P2P) architecture. Peer-to-Peer botnets use a decentralized Command & Control architecture. Moreover, a large number of such botnets already exist, and newer versions- which significantly differ from their parent bot- are also discovered practically every year. In this work, the authors propose and implement a novel hybrid framework for detecting P2P botnets in live network traffic by integrating Neural Networks with Bayesian Regularization. Bayesian Regularization helps in achieving better generalization of the dataset, thereby enabling the detection of botnet activity even of those bots which were never used in training the Neural Network. Hence such a framework is suitable for detection of newer and unseen botnets in live traffic of a network. This was verified by testing the Framework on test data unseen to the Detection module (using untrained botnet dataset), and the authors were successful in detecting this activity with an accuracy of 99.2 %.