Papers by Koustuv Saha
RubRIX: Rubric-Driven Risk Mitigation in Caregiver-AI Interactions (2026.findings-acl)
Copied to clipboard
Drishti Goel, Jeongah Lee, Qiuyue Zhong, Violeta J. Rodriguez, Daniel S. Brown, Ravi Karkar, Dong Whi Yoo, Koustuv Saha
| Challenge: | Existing frameworks focused on general risks may not adequately capture nuanced risks of LLMs in caregiving contexts. |
| Approach: | They propose a theory-driven, clinician-validated framework for evaluating risks in LLMs . RubRIX operationalizes five empirically-derived risk dimensions: Inattention, Bias Stigma, Information Inaccuracy, Uncritical Affirmation, and Epistemic Arrogance. |
| Outcome: | The proposed framework reduces risk components by 45-98% after one iteration across models. |
SLM-Mod: Small Language Models Surpass LLMs at Content Moderation (2025.naacl-long)
Copied to clipboard
| Challenge: | Large language models (LLMs) are expensive to query in real-time and do not allow for a community-specific approach to content moderation. |
| Approach: | They propose to use small language models for community-specific content moderation tasks by fine-tuning and evaluating their performance against larger open- and closed-sourced models. |
| Outcome: | The proposed models outperform zero-shot LLMs in content moderation tasks with 11.5% higher accuracy and 25.7% higher recall across all communities. |
MoMoE: Mixture of Moderation Experts Framework for AI-Assisted Online Governance (2025.emnlp-main)
Copied to clipboard
| Challenge: | Existing approaches for content moderation require a separate model for every community and are opaque in their decision-making. |
| Approach: | They propose a modular framework that adds post-hoc explanations to enable scalable content moderation. |
| Outcome: | The proposed framework yields scalable, transparent moderation without fine-tuning across domains. |