Koustuv Saha

The paper introduces LLUMI, an open-source framework that improves LLM writing assistance for mental health support using community feedback, demonstrating comparable performance to proprietary models while enhancing privacy.

This study demonstrates that an LLM's assigned support role (e.g., Inform, Coach, Relate) significantly alters its safety profile and the types of risks it presents when assisting users in complex caregiving situations.

The paper introduces the Tacit Understanding Index (TUX) to measure non-explicit alignment between humans and LLMs, finding that this alignment is significantly structured by individual person-level traits rather than simple random similarity.

The paper demonstrates that increasing the toxicity of prompts significantly degrades the factual reliability of LLMs, a degradation linked to the selective amplification of perturbation-sensitive nodes within the model's internal circuits.