AI/ML Research at ECE, IISc | 2026 Conference Highlights

ECE at IISc is a home to an active community working on mathematical foundations of machine learning — including sequential inference, bandits, reinforcement learning, statistical learning, and sequential decision making under uncertainty. We are delighted to share six accepted papers appearing at ICML, COLT, ALT, and AISTATS, 2026. 

International Conference on Machine Learning (ICML) 2026:  

  • Beyond First-order Asymptotics in Sequential Mean Testing 
    Vikas Deep, Shubhada Agrawal 
  • Asymptotically Optimal Sequential Testing with Markovian Data 
    Alhad Sethi, Kavali Sofia Sagar, Shubhada Agrawal, Debabrota Basu, P.N. Karthik 

Conference on Learning Theory (COLT) 2026:  

  • Almost sure null bankruptcy of testing-by-betting strategies 
    Hongjian Wang, Shubhada Agrawal, Aaditya Ramdas 

International Conference on Algorithmic Learning Theory (ALT) 2026 

  • Eventually LIL Regret: Almost Sure $\ln\ln T$ Regret for a sub-Gaussian Mixture on Unbounded Data 
    Shubhada Agrawal, Aaditya Ramdas 

Annual Conference on Artificial Intelligence and Statistics (AISTATS) 2026 

  • Monotone and Conservative Policy Iteration Beyond the Tabular Case 
    S.R. Eshwar, Gugan Thoppe, Ananyabrata Barua, Aditya Gopalan, and Gal Dalal 
  • Bad Values but Good Behavior: Learning Highly Misspecified Bandits with Function Approximation 
    Debangshu Banerjee, Aditya Gopalan 

From mathematical foundations to vision, language, graphs, and scientific applications, ML research at ECE spans the full spectrum from theory to practice. We are pleased to share six papers appearing at ICML, AISTATS, AAAI, ECCV, and WACV 2026. 

International Conference on Machine Learning (ICML) 2026: 

  • Interpretable Discovery of One-Parameter Subgroups: A Modular Framework for Elliptical, Hyperbolic, and Parabolic Symmetries 
    Karjol P., Kashyap V., Kashyap R.V., Prathosh A.P.
  • Learning Long Range Spatio-Temporal Representations over Continuous Time Dynamic Graphs with State Space Models 
    Ayushman Raghuvanshi, Thummaluru Siddartha Reddy, Sundeep Prabhakar Chepuri, and Mahesh Chandran 

Annual Conference on Artificial Intelligence and Statistics (AISTATS) 2026: 

  • Learning Equivariant Functions via Quadratic Forms 
    Karjol P., Kashyap V.V., Kashyap R., Prathosh A.P. 

AAAI 2026 

  • Covariance scattering transforms 
    Andrea Cavallo, Ayushman Raghuvanshi, Sundeep Prabhakar Chepuri, and Elvin Isufi 

European Conference on Computer Vision (ECCV) 2026:  

  • Temporally Aware Densification for Dynamic 3D Gaussian Splatting 
    Vikram Sandu, Mayurdeep Pathak, Rajiv Soundararajan 

WACV 2026:  

  • MANTA: Physics-Informed Generalized Underwater Object Tracking 
    Srinath S., Jamadagni H., Chandrasekar A., Prathosh A.P. 

ECE researchers are also advancing preference learning, trustworthy LLMs, influence estimation, NLP, and generative models for science. We are delighted to share five papers appearing at ICLR, ACL, and EACL 2026.  

International Conference on Learning Representations (ICLR) 2026: 

  • Why DPO is a Misspecified Estimator and How to Fix It 
    Aditya Gopalan, Sayak Ray Chowdhury, and Debangshu Banerjee
  • Enhancing Trustworthiness of Fine-Tuned LLMs via Regularized Subset Selection 
    Kumar S., Sharma N., Tiwari K., Prathosh A.P. 
  • f-INE: A Hypothesis Testing Framework for Estimating Influence under Training Randomness 
    Panda S., Tarsadiya D., Sourav S., Prathosh A.P., Karimireddy S.P. 

ACL 2026  

  • BiMol-Diff: A Unified Diffusion Framework for Molecular Generation and Captioning 
    Shahane A.H., Sirohi A.K., Arora D., Kumar N., Prathosh A.P., Kumar S 

EACL 2026 

  • Chandomitra: Towards Generating Structured Sanskrit Poetry from Natural Language Inputs 
    Jagadeeshan M.B. et al., Prathosh A.P., Goyal P