4 papers
Hard Negatives Reveal What Easy Negatives Hide: Cross-Lingual Harmfulness Representations Degrade with Resource Tier Under Hard Negatives
Paras Balani, Subhrakanta Panda
Safety alignment in large language models is trained primarily in English, and recent work reports that the underlying harmfulness representation survives translation: English-trai…
Reporting Under Pressure: Separating Factual and Tonal Sycophancy in LLM Statistical Analysis
Paras Balani, Subhrakanta Panda
Large language models are increasingly asked to analyze data and report what the results mean, a task distinct from the belief- or preference-alignment settings studied in most syc…
LLMs Don't Pay for the Jump
Paras Balani, Subhrakanta Panda
Zahavy [2026] argues that Large Language Models, despite their capabilities in induction and deduction, cannot perform the abductive "Jump" that produced Einstein's equivalence pri…
Self-Referential Induction Increases Response Instability Relative to Unresolvable and Verifiable Questions in Large Language Models
Paras Balani, Subhrakanta Panda
Self-referential prompting has been shown to reliably induce large language models to produce first-person reports resembling subjective experience, but no prior work measures how…