Project EKA
An India-centric foundation LLM programme with IndiaAI / MeitY, including the Eka-Eval evaluation suite.
Project page →Department of CSE · IIT Gandhinagar
Decoding and crafting language with AI
LINGO is a research laboratory at IIT Gandhinagar. Work is organised around four themes: multilingual and multimodal intelligence, foundation model architectures, reasoning and agentic intelligence, and evaluation and trustworthy AI. We build datasets, models, and systems for both Indian and global languages.
Research projects
Selected systems and models
Chitrabhasha
Multimodal datasets and foundation models for Indian languages (ANRF, 2026–2029).
Project page →Code-mixed NLP
Datasets and tools for Hindi–English code-mixing, including COMI-LINGUA and Commentator.
Project page →Project Unity
Ganga Hindi language models and UnityAI-Guard, a multilingual toxicity-detection system for Indian languages.
Project page →Research themes
Areas of work
Multilingual & Multimodal Intelligence
Language diversity, code-mixing, and multimodal understanding.
Theme overview →
Foundation Model Architectures & Efficient Intelligence
New architectures, efficient computation, and model adaptation.
Theme overview →
Reasoning, World Models & Agentic Intelligence
Reasoning, procedural execution, world models, and grounded intelligence.
Theme overview →
Evaluation, Reliability & Trustworthy AI
Evaluation, alignment, robustness, security, and interpretability.
Theme overview →0Publications
0Project funding
0Awards
0Datasets
0Ph.D. students
0M.Tech. students
0Project staff
0Active projects
Selected publications
Recent papers
- IndicTalk: A Large-Scale Persona-Based Multilingual Conversational Corpus for Indic Languages. Gawande, Singh. arXiv:2607.23242.
- COMI-LINGUA: Expert Annotated Large-Scale Dataset for Multitask NLP in Hindi-English Code-Mixing. Sheth, Beniwal, Singh. Findings of EMNLP 2025. Code-mixed NLP
- UNITYAI-GUARD: Pioneering Toxicity Detection Across Low-Resource Indian Languages. Beniwal et al. EMNLP 2025 System Demonstrations. Project Unity
- Eka-Eval: A Comprehensive Evaluation Framework for Large Language Models in Indian Languages. Sinha, Sheth, Upperwal, Singh. arXiv:2507.01853. Project EKA
- Commentator: A Code-mixed Multilingual Text Annotation Framework. Sheth, Nisar, Prajapati, Beniwal, Singh. EMNLP 2024 (System Demonstrations). Code-mixed NLP
- Cross-lingual Editing in Multilingual Language Models. Beniwal, Nandagopan, Singh. Findings of EACL 2024.
The laboratory
LINGO Labs, IIT Gandhinagar
Founded in 2019 and led by Prof. Mayank Singh, Jibaben Patel Chair Professor in Artificial Intelligence and Associate Professor in Computer Science and Engineering. The group also contributes to national programmes including IndiaAI, ANRF, DST-FIST, and the MoE Centre of Excellence on AI for Sustainable Cities.
Research is organised around multilingual and multimodal intelligence, foundation model architectures, reasoning and agentic intelligence, and evaluation and trustworthy AI, with past programmes in scholarly and educational NLP.
Contact