Shubhashis Roy Dipta, PhD Researcher in NLP and Multimodal AI

Shubhashis Roy Dipta

PhD Researcher · UMBC

sroydip1@umbc.edu

Resume Google Scholar
Amazon Science (Alexa AI)
Seattle, WA
Applied Scientist Intern
Summer 2026
Manager: Dr. Lichao Wang
Mentors: Dr. Xiaohu Xie, Dr. Daniel Bis
Amazon Science (Alexa AI)
Seattle, WA
Applied Scientist Intern
Summer 2025
Manager: Dr. Lichao Wang
Mentors: Dr. Daniel Bis, Dr. Kun Zhou
Paper: PA3: Policy-Aware Agent Alignment[EMNLP main]
Scale AI
San Francisco, CA
Machine Learning Research Intern
Summer 2024
Manager: Dr. Adrian Lam
Mentor: Vijay Kalmath
Blog: RLHF for Text-to-SQL
See more
University of Maryland, Baltimore County
Ph.D. in Computer Science
Fall 2023 - Present
Advisor: Dr. Frank Ferraro
Grade: 4.00/4.00
Publications: See Here (From 2022)
University of Maryland, Baltimore County
M.Sc. in Computer Science
Spring 2021 - Spring 2023
Awards: Phi Kappa Phi
Grade: 4.00/4.00
Morgan State University
Research Assistant
2017 - 2019
Advisor: Dr. Iman Dehzangi
Publications: 4 Journal
UniShopr.com
Bangladesh
Founder
2017 - 2021

Upcoming Travel

  • NeurIPS 2026 in Atlanta, GA (Dec 9-13)
  • ICLR 2027 in San Francisco, CA (Apr 26-30)
  • NAACL 2027 in San Francisco, CA (Jun 1-5)
👋 I'm open to meet! Email me to schedule a chat!

Peer Review

Reviewed 41+ papers across top venues since 2023.

22+ conference papers 13+ workshop papers 6+ journal papers

I’m a final-year CS Ph.D. researcher at the University of Maryland, Baltimore County (UMBC), advised by Dr. Frank Ferraro, with research internships at Amazon Science (Alexa AI; Summer 2025 + 2026) and Scale AI (Summer 2024). I make models expose the structure behind an answer, events, questions, policies, graphs, then train and score against that structure with GRPO or OPD. My work on agentic LLMs earned a $20K Google Cloud Gemini Academic Program Award (2026).

Graduating Spring 2027 · No visa sponsorship needed · actively seeking Research Scientist roles in NLP / Multimodal AI. Please reach out if you have an opening.

Recent News (See All)

Sep 26, 2026 🎉 Three papers accepted at NeurIPS 2026 - Rock Tokens to the main conference, plus TRIAGE and AgentCollabBench to the E&D track.
Sep 6, 2026 🚀 New preprint - OracleZoom: recursive super-resolution to 256× with ground truth only at 4×. A VLM judge prefers it over Chain-of-Zoom 78% of the time. Demo.
Aug 22, 2026 🎉 Four papers accepted at EMNLP 2026 - PA3 to the main conference, plus †DAGGER, Many Dialects and Register Shifts to Findings.
Jun 16, 2026 🎉 Awarded a $20,000 Google Cloud research grant through the Gemini Academic Program to support my LLM agentic research at UMBC. Featured by UMBC and on LinkedIn.

Beyond Research

I’ve competed internationally in algorithms and robotics, ranking 8th out of 300+ teams at the 2018 ACM ICPC Asia Dhaka Regional with multiple regional and national placements, reaching the top 70 on Kaggle 🥉 in the Birdcall Identification competition, and placing 9th at the University Rover Challenge 2015 (Utah, USA) and 22nd at the European Rover Challenge 2016 (Poland). Full list of awards →

Before the PhD, I also founded UniShopr (2017-2021), a cross-border e-commerce platform serving consumers in Bangladesh.

Featured Publications

A few papers that best represent my work. See all publications

  1. OracleZoom: On-Policy Self-Distillation Inspired Reference-Constrained Recursive Image Super Resolution
    Under review
    OracleZoom: On-Policy Self-Distillation Inspired Reference-Constrained Recursive Image Super Resolution
    Shubhashis Roy Dipta*, Sourajit Saha*, Shaswati Saha, and 1 more author
    Under review
    * Equal contribution
    arXiv Code 🤗 Hugging Face Website Demo
  2. DecomposeRL: Learning to Ask Useful, Informative, and Diverse Questions for Semi-Supervised, Traceable Claim Verification
    Under review
    DecomposeRL: Learning to Ask Useful, Informative, and Diverse Questions for Semi-Supervised, Traceable Claim Verification
    Shubhashis Roy Dipta, Ankur Padia, and Francis Ferraro
    Under review
    arXiv Code 🤗 Hugging Face Website
  3. Cornerstones or Stumbling Blocks? Deciphering the Rock Tokens in On-Policy Distillation
    NeurIPS
    Cornerstones or Stumbling Blocks? Deciphering the Rock Tokens in On-Policy Distillation
    Yuxuan Jiang*, Runchao Li*, Shubhashis Roy Dipta*, and 2 more authors
    NeurIPS 2026
    * Equal contribution
    arXiv
  4. TRIAGE: Evaluating Prospective Metacognitive Control in LLMs under Resource Constraints
    NeurIPS
    TRIAGE: Evaluating Prospective Metacognitive Control in LLMs under Resource Constraints
    Zabir Al Nazi, and Shubhashis Roy Dipta
    NeurIPS 2026
    arXiv
  5. AgentCollabBench: Diagnosing When Good Agents Make Bad Collaborators
    NeurIPS
    AgentCollabBench: Diagnosing When Good Agents Make Bad Collaborators
    Aritra Mazumder, Shubhashis Roy Dipta, Nusrat Jahan Lia, and 10 more authors
    NeurIPS 2026
    arXiv Code Website
  6. PA3: Policy-Aware Agent Alignment through Chain-of-Thought
    EMNLP
    PA3: Policy-Aware Agent Alignment through Chain-of-Thought
    Shubhashis Roy Dipta, Daniel Bis, Kun Zhou, and 4 more authors
    EMNLP 2026
    Work done during internship at Amazon Alexa AI
    arXiv Video
  7. †DAGGER: Distractor-Aware Graph Generation for Executable Reasoning in Math Problems
    EMNLP
    †DAGGER: Distractor-Aware Graph Generation for Executable Reasoning in Math Problems
    Zabir Al Nazi, Shubhashis Roy Dipta, and Sudipta Kar
    EMNLP 2026
    arXiv Code Website
  8. GanitLLM: Difficulty-Aware Bengali Mathematical Reasoning through Curriculum-GRPO
    ACL
    GanitLLM: Difficulty-Aware Bengali Mathematical Reasoning through Curriculum-GRPO
    Shubhashis Roy Dipta, Khairul Mahbub, and Nadia Najjar
    ACL 2026
    arXiv Code 🤗 Hugging Face Website Video Poster
  9. VC-Inspector: Advancing Reference-free Evaluation of Video Captions with Factual Analysis
    ACL
    VC-Inspector: Advancing Reference-free Evaluation of Video Captions with Factual Analysis
    Shubhashis Roy Dipta, Tz-Ying Wu, and Subarna Tripathi
    ACL 2026
    arXiv Code 🤗 Hugging Face Website Video Slides Poster
  10. Multimodal Unlearning Across Vision, Language, Video, and Audio: Survey of Methods, Datasets, and Benchmarks
    ACL
    Multimodal Unlearning Across Vision, Language, Video, and Audio: Survey of Methods, Datasets, and Benchmarks
    Nobin Sarwar, Shubhashis Roy Dipta, Zheyuan Liu, and 1 more author
    ACL 2026
    arXiv PDF Code Website
  11. Q2E: Query-to-Event Decomposition for Zero-Shot Multilingual Text-to-Video Retrieval
    AACL
    Q2E: Query-to-Event Decomposition for Zero-Shot Multilingual Text-to-Video Retrieval
    Shubhashis Roy Dipta, and Francis Ferraro
    AACL 2025
    arXiv Code 🤗 Hugging Face Website Video Slides Poster
  12. Learning How to Use Tools, Not Just When: Pattern-Aware Tool-Integrated Reasoning
    MathAI @NeurIPS
    Learning How to Use Tools, Not Just When: Pattern-Aware Tool-Integrated Reasoning
    Ningning Xu, Yuxuan Jiang, and Shubhashis Roy Dipta
    MathAI @NeurIPS 2025
    arXiv

Patents

  • AI-Based System and Method for Code Localization Using Dependency Graph Retrieval for Bug Identification in Software Repositories
    US provisional application filed, March 2026
  • System and Method for Anchor-Based Memory Indexing and Policy-Driven Retrieval Across Heterogeneous Sources
    US provisional application filed, March 2026

Research Mentorship