Somak Aditya
Scholar

Somak Aditya

Google Scholar ID: 2shiHpwAAAAJ
Assistant Professor, IIT Kharagpur
Knowledge RepresentationCommonsense ReasoningNatural Language ProcessingNatural Language UnderstandingVisual Reasoning
Citations & Impact
All-time
Citations
733
 
H-index
13
 
i10-index
15
 
Publications
20
 
Co-authors
26
list available
Resume (English only)
Academic Achievements
  • Two papers accepted at EMNLP 2025: main conference paper 'EduVidQA' and Findings paper 'NLKI', a lightweight framework for commonsense VQA.
  • Paper accepted at ACL 2025 Findings on evaluating LLMs’ math and coding skills via ontology-guided interventions.
  • Long paper accepted at NAACL 2025 main track on word sensitivity estimation for adversarial text generation.
  • Awarded USD 8000 Azure credits from Microsoft AFMR for the 'Education Video Copilot' project.
  • Served as Demo Track Co-Chair for CODS 2025.
  • Invited speaker at IIIT-H CVIT Summer School, Microsoft Research India Academic Summit, IISc, and National Research Council Canada.
Background
  • Assistant Professor (Grade I) in the Department of Computer Science and Engineering at IIT Kharagpur.
  • Research focuses on enhancing, evaluating, and explaining complex reasoning capabilities of AI systems.
  • Works at the intersection of classical and statistical AI, commonly known as neuro-symbolic methods.
  • Aims to enable deployable AI systems to reason with human-like commonsense knowledge.
  • Research spans reasoning in Transformer-based language models, multi-hop reasoning, visual question answering, and image understanding.