Riju Marwah

Trustworthy & Reliable LLMs  ·  Token-level Dynamics  ·  Interpretability  ·  Alignment

Tübingen, Germany / / marwah.riju@gmail.com / riju.marwah@student.uni-tuebingen.de / CV ↗ / LinkedIn / Google Scholar / GitHub
Riju Marwah

I am a Research Intern at the AI Institute, University of South Carolina, advised by Prof. Amit Sheth and Vishal Pallagani. I will be joining the M.Sc. Machine Learning programme at the University of Tübingen in fall 2026.

My research focuses on making large language models trustworthy, robust, and reliable through token-level dynamics. I study how technical signals and linguistic cues interact during long-context generation and design retrain-free interventions to mitigate degradation — sitting at the intersection of robustness, interpretability, and alignment.

I am always open to discussing research or collaborations — feel free to reach out.

News

  • Apr 2026 Paper accepted at ICML 2026 — Cognitive Fatigue in Autoregressive Transformers: Formalization and Measurement.
  • Jan 2026 Presented Chatsparent at AAAI 2026 in Singapore; served as a Student Scholar Volunteer.
  • Oct 2025 Paper accepted at the AAAI 2026 Demonstration Track — Chatsparent: An Interactive System for Detecting and Mitigating Cognitive Fatigue in LLMs (28% acceptance rate).

Publications

  1. 1.
    Cognitive Fatigue in Autoregressive Transformers: Formalization and Measurement ICML 2026
    Riju Marwah*, Ritvik Garimella, Vishal Pallagani, Atishay Jain, Michael Stewart, Amit Sheth
    Formalizes cognitive fatigue as a runtime state variable grounded in three token-level signals. Introduces the Fatigue Index with five explicit axioms; validated across nine models (1B–13B) on HotpotQA, TriviaQA, and SQuAD under long-context, positional, and precision stress conditions.
    Project Page ↗ PDF ↗
  2. 2.
    Chatsparent: An Interactive System for Detecting and Mitigating Cognitive Fatigue in LLMs AAAI 2026 Demo
    Riju Marwah*, Vishal Pallagani, Ritvik Garimella, Amit Sheth
    An interactive system that detects and mitigates cognitive fatigue in LLMs via token-level signals and retrain-free interventions, improving reliability and transparency during dialogue.
    Project Page ↗ PDF ↗
  3. 3.
    MicroDetect-Net (MDN): Leveraging Deep Learning to Detect Microplastics in Clam Blood Springer / ICICC 2025
    8th International Conference on Innovative Computing and Communication, Springer Nature
    Combines fluorescence microscopy (Nile Red staining) and deep learning to scan blood samples for microplastics, a step toward human blood analysis.
    PDF ↗

Experience

  1. Research Intern AI Institute, University of South Carolina Apr 2025 – Present
    • Advised by Prof. Amit Sheth (NCR Chair & Director, AIISC) and Vishal Pallagani.
    • Designed a framework to detect and mitigate cognitive fatigue in LLMs using token-level signals and real-time interventions.
    • First-authored publications at AAAI 2026 (Demo) and ICML 2026; researching long-context reliability, entropy collapse, and attention decay.
  2. Research Collaborator University of Illinois Urbana-Champaign Nov 2025 – Mar 2026
    • Studied politeness framing and reward leakage in LLMs across structured tasks and instruction-following settings.
    • Performed mechanistic interpretability analysis including early-token probing and activation patching.
  3. Generative AI Intern EY (Ernst & Young) Jan 2025 – Mar 2025
    • Built a low-code platform for agentic AI workflows using modular DAGs, vector DBs, and LLM toolchains (OpenAI, FAISS).
    • Implemented Celery–Redis task execution with production-grade scalability and semantic agent routing.
  4. Software Developer Intern National Thermal Power Corporation Jul 2024 – Sep 2024
    • Developed ASP.NET Core applications using MVC and Entity Framework for enterprise automation.
    • Implemented secure authentication using JWT, Identity Framework, and Google reCAPTCHA.

Projects

↗
Research · LLM Interpretability
Cognitive Fatigue in LLMs
Formalizing cognitive fatigue as a measurable latent state in autoregressive transformers. Fatigue Index, token-level signals, retrain-free mitigations.
↗
Industry · Agentic AI
Intelligent Workflow Orchestration
A modular, DAG-based low-code platform for composing AI-augmented pipelines using LLM agents, vector stores, and real-world data connectors. Built at EY.
↗
Community · Guide
Masters in Germany: A Complete Guide
A detailed guide for students navigating German MSc admissions, covering universities, applications, funding, and life in Germany.

Honors & Service

  • AAAI Student Volunteer Scholarship – USD 1,100 travel grant, AAAI 2026
  • Student Scholar Volunteer – AAAI 2026, Singapore
  • McKinsey Forward Program – selected for global initiative on problem-solving, business, and leadership
  • Delegate – Harvard Project for Asian & International Relations (HPAIR), 2025
  • Media Coverage – Times of India, Hindustan Times, Dwarka Parichay, Brainfeed

Education

  • M.Sc. Machine Learning Incoming
    University of Tübingen
    One of Europe's most prestigious ML programmes, affiliated with the Max Planck Institute for Intelligent Systems & the Tübingen AI Center
    Sep 2026 –
  • B.Tech. Computer Science Engineering
    Guru Gobind Singh Indraprastha University
    GPA: 8.44 / 10.0
    Nov 2022 – Jun 2026
Riju Marwah