8 papers
Where To Look? : Causal Tracing of Vision Encoders in VLM
Naren Kumar S, Tirth Bhatt, Mayank Singh
Vision-language models can describe an image with remarkable accuracy, yet a more fundamental question remains unanswered: what visual information actually drives their answers? In…
Task-Conditional Flow Matching for Balanced Multilingual Text Embedding Adaptation
Tirth Bhatt, Naren Kumar S, Mayank Singh
Multilingual text embedding models are commonly adapted using a single training objective across diverse tasks, despite different tasks requiring fundamentally different optimizati…
Sycophancy as a Multilingual Alignment Failure: How Safety Degrades Across Languages, Topics, and Models
Arya Shah, Himanshu Beniwal, Mayank Singh +1
Safety-aligned large language models often exhibit sycophancy, which is the tendency to affirm users' opinions regardless of factual accuracy. Although well-studied in English, its…
Where Does Toxicity Live? Mechanistic Localization and Targeted Suppression in Language Models
Himanshu Beniwal, Mayank Singh
Large language models frequently generate toxic, hateful, or harmful content, yet existing mitigation methods rely on costly retraining or output-level filtering with no mechanisti…
One Instruction Does Not Fit All: How Well Do Embeddings Align Personas and Instructions in Low-Resource Indian Languages?
Arya Shah, Himanshu beniwal, Mayank Singh
Aligning multilingual assistants with culturally grounded user preferences is essential for serving India's linguistically diverse population of over one billion speakers across mu…
Char-mander Use mBackdoor! A Study of Cross-lingual Backdoor Attacks in Multilingual LLMs
Himanshu Beniwal, Sailesh Panda, Birudugadda Srivibhav +1
We explore \textbf{C}ross-lingual \textbf{B}ackdoor \textbf{AT}tacks (X-BAT) in multilingual Large Language Models (mLLMs), revealing how backdoors inserted in one language can aut…