PeopleofSuperintelligence

Researcher

Aryan Seth

Aryan Seth is a computer scientist whose research spans fairness in facial expression recognition, privacy, and language-model evaluation. He studied for a BE in Computer Science at BITS Pilani from 2021 to 2025 and coauthored work on the InFER++ dataset, FAIR-FER, and latent alignment for facial-expression fairness. In 2025, he coauthored DebateBench and AffectSRNet; in 2026, he coauthored a study of how large candidate sets affect language-model decisions. As of October 2026, he is a DFKI research assistant working on information-sharing in federated bandits; the group includes him among its alumni, following research roles at Iowa State and IITB and an internship at Boston Consulting Group.

Now at
DFKI

Timeline

  1. 2021–2025

    Computer science at BITS Pilani

    From 2021 to 2025, he studies for a BE in Computer Science at BITS Pilani's Pilani Campus.

  2. 2024

    InFER++ and Indian facial data

    In 2024, he coauthors InFER++, a real-world Indian facial-expression dataset paper, with Syed Sameen Ahmad Rizvi, Jagat Sesh Challa, and Pratik Narang. The work appears in the IEEE Open Journal of the Computer Society.

  3. March 2024

    FAIR-FER at AAAI

    In March 2024, he coauthors the AAAI student abstract "FAIR-FER: A Latent Alignment Approach for Mitigating Bias in Facial Expression Recognition" with Syed Sameen Ahmad Rizvi and Pratik Narang; all three are affiliated with Birla Institute of Technology & Science, Pilani. It uses representation learning in latent spaces to reduce bias in facial-expression recognition and improve a model's fairness and accuracy.

  4. October 2024

    Latent alignment for fairness

    In October 2024, he coauthors "Balancing the Scales: Enhancing Fairness in Facial Expression Recognition with Latent Alignment" with Syed Sameen Ahmad Rizvi and Pratik Narang. The paper uses representation learning in latent spaces to mitigate bias in facial-expression recognition systems and improve model fairness and accuracy.

  5. February 2025

    DebateBench's long-context test

    In February 2025, he coauthors DebateBench, a dataset of 256 speeches from 32 British Parliamentary debates for evaluating long-context reasoning in language models. Preliminary tests with GPT o1, GPT-4o, and Claude Haiku show that the models struggle on the benchmark.

  6. February 2025

    AffectSRNet for facial images

    In February 2025, he coauthors AffectSRNet, an emotion-aware super-resolution framework that reconstructs high-quality facial images from low-resolution inputs while maintaining expression intensity and fidelity.

  7. September 2026

    Large option sets and LLMs

    In September 2026, he coauthors "Overwhelmed by Choice: Studying LLM Decision Making at Scale." The study finds that model accuracy falls as candidate sets grow; hierarchical partitioning and permutation-based inference improve accuracy by roughly 20 percentage points at 160 candidates on HotpotQA and MIMIC.

  8. As of October 2026

    Research assistant at DFKI

    As of October 2026, he is a research assistant in DFKI's Data Science and its Applications group, where he works on information-sharing in federated bandits; DFKI includes him among its alumni. His interests include fairness and privacy. Before DFKI, he interns as a management consultant with Boston Consulting Group and serves as a research assistant in Iowa State University's Dependable Computation Lab and IITB's Center for Machine Intelligence.

Career

  1. NowDFKIResearch assistant
  2. Boston Consulting GroupManagement consultant intern
  3. Iowa State UniversityResearch assistant
  4. IIT BombayResearch assistant

Education

  1. 2025Birla Institute of Technology and Science, PilaniBE in Computer Science

From the biography.

  1. Nonprofit roles are 2.3 times as common in AI safetyIncludes DFKI. We estimate where people focused on safety, alignment and evaluation work. Nonprofit roles appear for 25.1% of safety people with known affiliations, compared with 11.0% in other areas.
  2. The top 10 organizations employed 22% of people shaping AI in 2025: 51% of company researchers and 21% of university researchersIncludes DFKI. We estimate the share of people shaping AI who worked at the top 10 organizations each year from 1955 to 2025, comparing concentration across companies, universities, and institutes and labs.
  3. The United States employs an estimated 48% of people shaping AIIncludes DFKI. We estimate where the people shaping AI work today, how universities, companies and research institutes split across countries, and how country shares have changed since the 1960s.
  4. More people shaping AI studied at Massachusetts Institute of Technology than at any other institutionIncludes DFKI. We estimate where the people shaping AI studied and where their careers led. Massachusetts Institute of Technology leads with 467 former students, including 291 recorded doctorates.

Sources and credits

Photo: We have no published photo of Aryan Seth yet. Send us a free-licensed photo.

  1. Aryan Seth — Data Science and its Applications dsa.dfki.de. Checked 22 October 2026.
  2. Aryan Seth aryan-seth.github.io. Checked 22 October 2026.
  3. FAIR-FER: A Latent Alignment Approach for Mitigating Bias in Facial Expression Recognition (Student Abstract) | Proceedings of the AAAI Conference on Artificial Intelligence ojs.aaai.org. Checked 22 October 2026.
  4. [2410.19444] Balancing the Scales: Enhancing Fairness in Facial Expression Recognition with Latent Alignment arxiv.org. Checked 22 October 2026.
  5. [2502.06279] DebateBench: A Challenging Long Context Reasoning Benchmark For Large Language Models arxiv.org. Checked 22 October 2026.
  6. [2502.09932] AffectSRNet : Facial Emotion-Aware Super-Resolution Network arxiv.org. Checked 22 October 2026.
  7. [2609.32809] Overwhelmed by Choice: Studying LLM Decision Making at Scale arxiv.org. Checked 22 October 2026.
  8. Team | Data Science and its Applications dsa.dfki.de. Checked 22 October 2026.