IAI, TCG CREST Researchers Secure Multiple Acceptances at EMNLP 2026 and TMLR

September 25, 2026

The Institute for Advancing Intelligence (IAI) at TCG CREST is pleased to announce a series of significant research accomplishments with the acceptance of multiple papers at premier international venues in natural language processing and machine learning research. These contributions underscore the institute’s commitment to advancing safe, trustworthy, and impactful artificial intelligence through rigorous interdisciplinary research.

Among the achievements are three paper acceptances at EMNLP 2026, one of the world’s leading conferences in Natural Language Processing, including one Main Conference (Long) paper and two Findings (Long) papers. In addition, IAI researchers have secured an acceptance at Transactions on Machine Learning Research (TMLR), a leading journal in the field of machine learning.

The accepted papers are:

PAPER 1:

SafeTutors: Benchmarking Pedagogical Safety in AI Tutoring Systems

Authors: Rima Hazra, Bikram Ghuku, Ilona Marchenko, Yaroslava Tokarieva, Sayan Layek, Somnath Banerjee, Julia Stoyanovich and Mykola Pechenizkiy.

This work investigates pedagogical safety in AI-powered tutoring systems, introducing benchmarks that enable systematic evaluation of educational AI technologies and their impact on learners.

PAPER 2:

SafeMath: Inference-time Safety Improves Math Accuracy

Authors: Sagnik Basu, Subhrajit Mitra, Aman Juneja, Somnath Banerjee, Rima Hazra and Animesh Mukherjee.

The study explores the relationship between safety mechanisms and mathematical reasoning in AI systems, demonstrating how safety-focused interventions can enhance performance in complex problem-solving tasks.

PAPER 3:

Lost in Interpretation: The Plausibility–Faithfulness Trade-off in Cross-Lingual Explanations

Authors: Somnath Banerjee, Pranav Jha, Rima Hazra and Animesh Mukherjee.

This research examines the challenges of explainability across languages, addressing important questions regarding the reliability and faithfulness of AI-generated explanations in multilingual contexts.

PAPER 4:

Attributional Safety Failures in Large Language Models under Code-Mixed Perturbations

Authors: Somnath Banerjee, Pratyush Chatterjee, Shanu Kumar, Sayan Layek, Parag Agrawal, Rima Hazra, and Animesh Mukherjee.

Accepted at Transactions on Machine Learning Research (TMLR), this work investigates safety vulnerabilities in large language models operating in code-mixed linguistic environments, an increasingly important area of research for globally deployed AI systems.

These achievements reflect the strong research contributions and expertise of the Dr. Rima Hazra, further enriched through collaborations with researchers from leading institutions, including Eindhoven University of Technology (TU/e), New York University (NYU) Brooklyn, Singapore Institute of Technology, Indian Institute of Technology Kharagpur, and Microsoft. Together, these collaborations demonstrate the global reach, interdisciplinary breadth, and collaborative spirit of AI research at IAI, TCG CREST.

TCG CREST congratulates Dr. Rima Hazra and all collaborating authors on these outstanding accomplishments. Their work continues to strengthen the institute’s contributions to responsible AI, machine learning, and natural language processing research on the international stage.