paper-conference

RONA: Pragmatically Diverse Image Captioning with Coherence Relations
Writing assistants like Grammarly and Microsoft Copilot typically generate diverse image captions through syntactic and semantic variation. In contrast, human-written captions rely on pragmatic cues to convey a central message. To better capture this pragmatic diversity, we introduce RONA, a prompting strategy for Multi-modal Large Language Models (MLLMs) that uses Coherence Relations as a variation axis. RONA produces more diverse and accurate captions than standard MLLM baselines across domains. Code: https://github.com/aashish2000/RONA
Aashish Anantha Ramakrishnan, Aadarsh Anantha Ramakrishnan, Dongwon Lee
Cite Code DOI Arxiv
From Intentions to Techniques: A Comprehensive Taxonomy and Challenges in Text Watermarking for Large Language Models
In this paper, we present a unified overview of different perspectives behind designing watermarking techniques through a comprehensive survey of the research literature.
Harsh Nishant Lalai, Aashish Anantha Ramakrishnan, Raj Sanjay Shah, Dongwon Lee
Cite DOI Arxiv
ANNA: Abstractive Text-to-Image Synthesis with Filtered News Captions
In this paper, we launch ANNA, an Abstractive News captioNs dAtaset extracted from online news articles in a variety of different contexts. Through our experiments, we show that techniques such as transfer learning achieve limited success in understanding abstractive captions but still fail to consistently learn the relationships between content and context features.
Aashish Anantha Ramakrishnan, Sharon X Huang, Dongwon Lee
Cite Dataset DOI Arxiv