Search arXiv⌕ Search

arXiv subjects

Sanjana Cheerla

Publications and source records attributed to Sanjana Cheerla.

2 recordsLinked to original sources

AVCORP: Identifying and Resisting Deception via Communication Theory and Theory of Mind

Today's LLM-based agents are vulnerable to deception by users and other agents. Defending against deception requires Theory of Mind (ToM) reasoning to infer the beliefs and goals of others as well as modeling tactics based on communication theory. To enable training models to detect deception, we adopt the well-known deception game, Avalon. We present a corpus, AVCORP, of 5,000 annotated utterances from 250 games of Avalon. AVCORP includes (1) 37 tactics mapped to a 4x4 behavior matrix, and (2) reasoning traces capturing abductive hypotheses, suspicion levels, and nested ToM beliefs. We train Qwen3-8B on AVCORP and evaluate it on 25 held-out games and 20 games played by humans. On held-out games, our best trained detector outperforms its untrained baseline and frontier models (GPT-5.6 Sol, Gemini 3.1 Pro, DeepSeek-V4-Pro), matching Claude Opus 5.5. On human games, every trained detector exceeds the untrained model in F1 score across all rounds.

cs.AI↗

Weapons of Online Harassment: Menacing and Profiling Users via Social Apps

Viewing social apps as sociotechnical systems makes clear that they are not mere pieces of technology but mediate human interaction and may unintentionally enable harmful behaviors like online harassment. As more users interact through social apps, instances of harassment increase. We observed that app reviews often describe harassment. Accordingly, we built a dataset of over 3 million reviews and 1,800 apps. We discovered that two forms of harassment are prevalent, Menacing and Profiling. We built a computational model for identifying reviews indicating harassment, achieving high recalls of 90% for Menacing and 85% for Profiling. We analyzed the data further to better understand the terrain of harassment. Surprisingly, abusers most often have female identities. Also, what distinguishes negative from neutral reviews is the greater prevalence of anger, disgust, and fear. Applying our model, we identified 1,395 apps enabling harassment and notified developers of the top 48 with the highest user-reported harassment.

cs.CY↗