arXiv subjects

Vamshi Krishna Bonagiri

Publications and source records attributed to Vamshi Krishna Bonagiri.

9 recordsLinked to original sources

arXiv2026

Cognitive Digital Twins: Ethical Risks and Governance for AI Systems That Model the Mind

As AI systems become increasingly persistent and personalized, they make possible a class of technologies that we call cognitive digital twins (CDTs): dynamic computational representations of a specific person's cognition, updated from behavioral, contextual, or physiological data in order to model, predict, or simulate that person's cognition, or to act as that person's communicative or decision-making proxy. CDTs combine cognitive inference with longitudinal representation, simulation, and proxy action in ways that existing governance strategies for personal assistants, autonomous agents, recommender systems, and automated decision systems only partially address. This paper makes four contributions. First, we define CDTs and distinguish them from adjacent systems. Second, we introduce a 5A governance framework organized around authority, autonomy, access and control, accountability, and availability. Third, we identify CDT-specific risks, from misrepresentation and epistemic authority shifts to shadow twins, simulated participation, proxy action, and proxy-power asymmetries. Fourth, we analyze governance gaps and propose requirements for high-risk CDTs that strengthen consent, purpose limitation, validity, traceability, contestation, independent review, and model retirement. Existing frameworks primarily regulate data processing, automated decisions, or autonomous actions; CDTs also require governance at the level of cognitive representation itself, before any final decision or external action occurs. We argue that CDTs require governance not only because they can act for people, but because they can become infrastructures through which cognition is represented, simulated, classified, and operationalized.

cs.AI↗

arXiv2026

Check Yourself Before You Wreck Yourself: Selectively Quitting Improves LLM Agent Safety

As Large Language Model (LLM) agents increasingly operate in complex environments with real-world consequences, their safety becomes critical. While uncertainty quantification is well-studied for single-turn tasks, multi-turn agentic scenarios with real-world tool access present unique challenges where uncertainties and ambiguities compound, leading to severe or catastrophic risks beyond traditional text generation failures. We propose using "quitting" as a simple yet effective behavioral mechanism for LLM agents to recognize and withdraw from situations where they lack confidence. Leveraging the ToolEmu framework, we conduct a systematic evaluation of quitting behavior across 12 state-of-the-art LLMs. Our results demonstrate a highly favorable safety-helpfulness trade-off: agents prompted to quit with explicit instructions improve safety by an average of +0.39 on a 0-3 scale across all models (+0.64 for proprietary models), while maintaining a negligible average decrease of -0.03 in helpfulness. Our analysis demonstrates that simply adding explicit quit instructions proves to be a highly effective safety mechanism that can immediately be deployed in existing agent systems, and establishes quitting as an effective first-line defense mechanism for autonomous agents in high-stakes applications.

cs.CL↗

arXiv2026

Intrinsic Guardrails: How Semantic Geometry of Personality Interacts with Emergent Misalignment in LLMs

Fine-tuning Large Language Models (LLMs) on benign narrow data can sometimes induce broad harmful behaviors, a vulnerability termed emergent misalignment (EM). While prior work links these failures to specific directions in the activation space, their relationship to the model's broader persona remains unexplored. We map the latent personality space of LLMs through established psychometric profiles like the Big Five, Dark Triad, and LLM-specific behaviors (e.g. evil, sycophancy), and show that the semantic geometry is highly stable across aligned models and their corrupted fine-tunes. Through causal interventions, we find that directions isolating social valence, such as the 'Evil' persona vector, and a Semantic Valence Vector (SVV) that we introduce, function as intrinsic guardrails: ablating them drives the misalignment rates above $40$%, while amplifying them suppresses the failure mode to less than $3$%. Leveraging the structural stability of the personality space, we show that vectors extracted $\textit{a priori}$ from an instruct-tuned model transfer zero-shot to successfully regulate EM in corrupted fine-tunes. Overall, our findings suggest that harmful fine-tuning does not overwrite a model's internal representation of personality, allowing conserved representations to serve as robust, cross-distribution guardrails.

cs.CL↗

arXiv2026

Flying Pigs, FaR and Beyond: Evaluating LLM Reasoning in Counterfactual Worlds

A fundamental challenge in reasoning is navigating hypothetical, counterfactual worlds where logic may conflict with ingrained knowledge. We investigate this frontier for Large Language Models (LLMs) by asking: Can LLMs reason logically when the context contradicts their parametric knowledge? To facilitate a systematic analysis, we first introduce CounterLogic, a benchmark specifically designed to disentangle logical validity from knowledge alignment. Evaluation of 11 LLMs across six diverse reasoning datasets reveals a consistent failure: model accuracy plummets by an average of 14% in counterfactual scenarios compared to knowledge-aligned ones. We hypothesize that this gap stems not from a flaw in logical processing, but from an inability to manage the cognitive conflict between context and knowledge. Inspired by human metacognition, we propose a simple yet powerful intervention: Flag & Reason (FaR), where models are first prompted to flag potential knowledge conflicts before they reason. This metacognitive step is highly effective, narrowing the performance gap to just 7% and increasing overall accuracy by 4%. Our findings diagnose and study a critical limitation in modern LLMs' reasoning and demonstrate how metacognitive awareness can make them more robust and reliable thinkers.

cs.CL↗

arXiv2025

COBIAS: Assessing the Contextual Reliability of Bias Benchmarks for Language Models

Large Language Models (LLMs) often inherit biases from the web data they are trained on, which contains stereotypes and prejudices. Current methods for evaluating and mitigating these biases rely on bias-benchmark datasets. These benchmarks measure bias by observing an LLM's behavior on biased statements. However, these statements lack contextual considerations of the situations they try to present. To address this, we introduce a contextual reliability framework, which evaluates model robustness to biased statements by considering the various contexts in which they may appear. We develop the Context-Oriented Bias Indicator and Assessment Score (COBIAS) to measure a biased statement's reliability in detecting bias, based on the variance in model behavior across different contexts. To evaluate the metric, we augmented 2,291 stereotyped statements from two existing benchmark datasets by adding contextual information. We show that COBIAS aligns with human judgment on the contextual reliability of biased statements (Spearman's $ρ= 0.65, p = 3.4 * 10^{-60}$) and can be used to create reliable benchmarks, which would assist bias mitigation works.

cs.CL↗

arXiv2025

From Human Judgements to Predictive Models: Unravelling Acceptability in Code-Mixed Sentences

Current computational approaches for analysing or generating code-mixed sentences do not explicitly model ``naturalness'' or ``acceptability'' of code-mixed sentences, but rely on training corpora to reflect distribution of acceptable code-mixed sentences. Modelling human judgement for the acceptability of code-mixed text can help in distinguishing natural code-mixed text and enable quality-controlled generation of code-mixed text. To this end, we construct Cline - a dataset containing human acceptability judgements for English-Hindi~(en-hi) code-mixed text. Cline is the largest of its kind with 16,642 sentences, consisting of samples sourced from two sources: synthetically generated code-mixed text and samples collected from online social media. Our analysis establishes that popular code-mixing metrics such as CMI, Number of Switch Points, Burstines, which are used to filter/curate/compare code-mixed corpora have low correlation with human acceptability judgements, underlining the necessity of our dataset. Experiments using Cline demonstrate that simple Multilayer Perceptron (MLP) models when trained solely using code-mixing metrics as features are outperformed by fine-tuned pre-trained Multilingual Large Language Models (MLLMs). Specifically, among Encoder models XLM-Roberta and Bernice outperform IndicBERT across different configurations. Among Encoder-Decoder models, mBART performs better than mT5, however Encoder-Decoder models are not able to outperform Encoder-only models. Decoder-only models perform the best when compared to all other MLLMS, with Llama 3.2 - 3B models outperforming similarly sized Qwen, Phi models. Comparison with zero and fewshot capabilitites of ChatGPT show that MLLMs fine-tuned on larger data outperform ChatGPT, providing scope for improvement in code-mixed tasks. Zero-shot transfer from En-Hi to En-Te acceptability judgments are better than random baselines.

cs.CL↗

arXiv2024

SaGE: Evaluating Moral Consistency in Large Language Models

Despite recent advancements showcasing the impressive capabilities of Large Language Models (LLMs) in conversational systems, we show that even state-of-the-art LLMs are morally inconsistent in their generations, questioning their reliability (and trustworthiness in general). Prior works in LLM evaluation focus on developing ground-truth data to measure accuracy on specific tasks. However, for moral scenarios that often lack universally agreed-upon answers, consistency in model responses becomes crucial for their reliability. To address this issue, we propose an information-theoretic measure called Semantic Graph Entropy (SaGE), grounded in the concept of "Rules of Thumb" (RoTs) to measure a model's moral consistency. RoTs are abstract principles learned by a model and can help explain their decision-making strategies effectively. To this extent, we construct the Moral Consistency Corpus (MCC), containing 50K moral questions, responses to them by LLMs, and the RoTs that these models followed. Furthermore, to illustrate the generalizability of SaGE, we use it to investigate LLM consistency on two popular datasets -- TruthfulQA and HellaSwag. Our results reveal that task-accuracy and consistency are independent problems, and there is a dire need to investigate these issues further.

cs.CL↗

arXiv2024

Measuring Moral Inconsistencies in Large Language Models

A Large Language Model (LLM) is considered consistent if semantically equivalent prompts produce semantically equivalent responses. Despite recent advancements showcasing the impressive capabilities of LLMs in conversational systems, we show that even state-of-the-art LLMs are highly inconsistent in their generations, questioning their reliability. Prior research has tried to measure this with task-specific accuracy. However, this approach is unsuitable for moral scenarios, such as the trolley problem, with no "correct" answer. To address this issue, we propose a novel information-theoretic measure called Semantic Graph Entropy (SGE) to measure the consistency of an LLM in moral scenarios. We leverage "Rules of Thumb" (RoTs) to explain a model's decision-making strategies and further enhance our metric. Compared to existing consistency metrics, SGE correlates better with human judgments across five LLMs. In the future, we aim to investigate the root causes of LLM inconsistencies and propose improvements.

cs.CL↗

arXiv2022

Are Deepfakes Concerning? Analyzing Conversations of Deepfakes on Reddit and Exploring Societal Implications

Deepfakes are synthetic content generated using advanced deep learning and AI technologies. The advancement of technology has created opportunities for anyone to create and share deepfakes much easier. This may lead to societal concerns based on how communities engage with it. However, there is limited research available to understand how communities perceive deepfakes. We examined deepfake conversations on Reddit from 2018 to 2021 -- including major topics and their temporal changes as well as implications of these conversations. Using a mixed-method approach -- topic modeling and qualitative coding, we found 6,638 posts and 86,425 comments discussing concerns of the believable nature of deepfakes and how platforms moderate them. We also found Reddit conversations to be pro-deepfake and building a community that supports creating and sharing deepfake artifacts and building a marketplace regardless of the consequences. Possible implications derived from qualitative codes indicate that deepfake conversations raise societal concerns. We propose that there are implications for Human Computer Interaction (HCI) to mitigate the harm created from deepfakes.

cs.HC↗