Published in Preprint, 2024
This paper introduces RADIANT, a novel approach for entity-context alignment in retrieval-augmented generation.
Recommended citation: Yaswanth, N., et al. (2024). "RADIANT - Retrieval AugmenteD entIty-context AligNmenT - Introducing RAG-ability."
Published in Under Review at TMLR 2025 (Accepted at 2 NeurIPS 2024 workshops), 2024
This paper introduces a hierarchical approach to improve language model evaluation and reasoning capabilities.
Recommended citation: Yaswanth, N., et al. (2024). "Review-Feedback-Reason (ReFeR): Improving Evaluation and Reasoning through Hierarchy of Models." arXiv preprint arXiv:2407.12877. https://arxiv.org/abs/2407.12877
Published in EMNLP 2024, 2024
This paper presents an approach for automatically evaluating video generation models using simulated human feedback.
Recommended citation: Yaswanth, N., et al. (2024). "VideoScore: Building Automatic Metrics to Simulate Fine-grained Human Feedback for Video Generation." Proceedings of EMNLP 2024. https://arxiv.org/abs/2406.15252v2
Published in NeurIPS 2024 Workshop, 2024
This paper presents a benchmark for evaluating multimodal language models on image implication understanding tasks.
Recommended citation: Yaswanth, N., et al. (2024). "II-Bench: An Image Implication Understanding Benchmark for Multimodal LLMs." arXiv preprint arXiv:2406.05862. https://arxiv.org/abs/2406.05862
Published in Under Review, 2024
This paper addresses the challenge of balancing contradictory objectives in text-to-image alignment.
Recommended citation: Yaswanth, N., et al. (2024). "YinYang-Align: Benchmarking Contradictory Objectives and Proposing Multi-Objective Optimization based DPO for Text-to-Image Alignment." arXiv preprint arXiv:2502.03512. https://arxiv.org/abs/2502.03512
Published in arXiv Preprint, 2024
This paper introduces an improved approach to direct preference optimization for language models.
Recommended citation: Yaswanth, N., et al. (2024). "DPO Kernels: A Semantically-Aware, Kernel-Enhanced, and Divergence-Rich Paradigm." arXiv preprint arXiv:2501.03271. https://arxiv.org/abs/2501.03271
Published in EMNLP 2023 (Findings), 2023
This paper introduces a novel task focused on identifying correct parsing and semantic relations between components of compounds in Sanskrit.
Recommended citation: Yaswanth, N., Guntupalli, S. K., & Goyal, P. (2023). "DepNeCTI: Dependency-based Nested Compound Type Identification for Sanskrit." Findings of the Association for Computational Linguistics: EMNLP 2023. https://aclanthology.org/2023.findings-emnlp.914/
I’ve been involved in several research projects related to Large Language Models, multimodal understanding, and evaluation metrics. My research interests include improving LLM reasoning capabilities, enhancing evaluation frameworks, and developing more efficient pre-training methodologies.
For a complete list of my publications, please visit my Google Scholar profile.
Several papers are currently under review at leading conferences and journals. Please refer to my CV for the most recent updates on my research work.