BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Computer Science and Engineering - ECPv6.13.0//NONSGML v1.0//EN
CALSCALE:GREGORIAN
METHOD:PUBLISH
X-ORIGINAL-URL:https://homecse.iitd.ac.in
X-WR-CALDESC:Events for Computer Science and Engineering
REFRESH-INTERVAL;VALUE=DURATION:PT1H
X-Robots-Tag:noindex
X-PUBLISHED-TTL:PT1H
BEGIN:VTIMEZONE
TZID:Asia/Kolkata
BEGIN:STANDARD
TZOFFSETFROM:+0530
TZOFFSETTO:+0530
TZNAME:IST
DTSTART:20260101T000000
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
DTSTART;TZID=Asia/Kolkata:20260506T120000
DTEND;TZID=Asia/Kolkata:20260506T130000
DTSTAMP:20261010T121650
CREATED:20260503T085759Z
LAST-MODIFIED:20260503T085759Z
UID:2517-1778068800-1778072400@homecse.iitd.ac.in
SUMMARY:Learning Hierarchical Control via Feasible Subgoal Prediction by Utsav Singh
DESCRIPTION:Venue: SIT001 \nAbstract: Solving long-horizon tasks remains a central challenge in robotics because agents must explore efficiently\, assign credit over long time scales\, and act under sparse supervision. To address this\, hierarchical reinforcement learning (HRL) offers a promising alternative to flat reinforcement learning (RL) by enabling a high-level policy to propose subgoals and a low-level policy to execute them. However\, in practice\, HRL suffers from a fundamental issue: the higher level can propose subgoals that are infeasible for the lower level to achieve\, leading to training instability and sub-optimal performance. In this talk\, I will present my research around a central idea: hierarchy is effective only when its high-level decisions are grounded in the capabilities of the lower-level policies. I will discuss methods for training high-level policies to predict feasible subgoals by leveraging expert demonstrations\, preference-based feedback\, and visually grounded reward synthesis. Across challenging navigation and manipulation tasks\, these approaches improve training stability\, mitigate non-stationarity\, and enable agents to solve complex sparse-reward tasks in both simulation and real-world robotic settings. More broadly\, this work argues that building intelligent robotic agents that can solve long-horizon tasks is not just a matter of better planning\, but of ensuring that high-level decisions remain aligned with what lower-level policies can actually achieve. \n  \nBio: Utsav Singh received his Ph.D. from the Department of Computer Science and Engineering at the Indian Institute of Technology (IIT) Kanpur\, advised by Dr. Vinay P. Namboodiri and Dr. Sunil E. Simon. He will be joining the University of Central Florida as a Postdoctoral Researcher in June 2026\, working with Dr. Mubarak Shah and Dr. Amrit Singh Bedi. Prior to his Ph.D.\, he received his M.Tech. from IIT Kanpur. His research interests span hierarchical reinforcement learning (HRL)\, embodied intelligence\, and bilevel optimization\, with a current focus on enabling language-guided robotic control and improving reasoning in large language models (LLMs) via bilevel actor-critic frameworks. His research has been published at leading venues including ICML\, NeurIPS\, ICLR\, and AAAI.
URL:https://homecse.iitd.ac.in/event/learning-hierarchical-control-via-feasible-subgoal-prediction-by-utsav-singh/
LOCATION:SIT 001\, Amar Nath and Shashi Khosla School of Information Technology\, IIT Delhi\, Hauz Khas\, New Delhi 110016\, India\, Delhi\, Delhi\, 110016\, India
END:VEVENT
BEGIN:VEVENT
DTSTART;TZID=Asia/Kolkata:20260520T160000
DTEND;TZID=Asia/Kolkata:20260520T170000
DTSTAMP:20261010T121650
CREATED:20260513T060719Z
LAST-MODIFIED:20260513T060719Z
UID:2520-1779292800-1779296400@homecse.iitd.ac.in
SUMMARY:Algorithmic Behaviours in In-Context Learning by Dr. Aditya Gangrade
DESCRIPTION:Venue: Bharti-501/ MS Teams \nAbstract: In-Context Learning (ICL) is a remarkable phenomenon whereby transformer-based LLMs can use data contained within their prompts to adapt their responses\, without changing their weights. This suggests that such models encode learning mechanisms. The recent literature has used statistical learning problems as a test-bed to investigate ICL\, and established that ICL can be realised for a wide range of function classes. However\, the mechanisms these models use to learn are poorly characterised. \nI will describe work on extracting and analysing learning algorithms embedded in the weights of transformers trained to perform ICL in two settings: linear-activation transformers for linear regression\, and softmax-activation transformers for linear classification. Through the former\, I will illustrate a high-level ‘simplify and validate’ strategy that allows extraction\, and through the latter\, I will describe a symmetry-driven strategy for evoking structure in these weights. In these settings\, we recover concrete iterative procedures that use existing ideas (Newton-Schulz; mean-shift methods) in new ways that are distinct from gradient descent. Further\, we show that transformers trained on variations of these problems implement modified versions of the same dynamics. This suggests that such models recover certain `stable’ algorithmic motifs\, and adapt them in response to problem structure. \nBased on work done jointly with Patrick Lutz\, Themistoklis Haris\, Arjun Chandra\, Hadi Daneshmand\, and Venkatesh Saligrama. \nBio:  Aditya Gangrade is a research scientist at the ECE department at Boston University. He obtained his Ph. D. in Systems Engineering from Boston University\, and previously held postdoctoral positions at Carnegie Mellon University and the University of Michigan. His research interests span theoretical and methodological aspects of machine learning\, with recent focus on safety in sequential decision making\, and on in-context learning phenomena.
URL:https://homecse.iitd.ac.in/event/algorithmic-behaviours-in-in-context-learning-by-dr-aditya-gangrade/
LOCATION:Bharti 501\, IIT Campus\, Hauz Khas\, New Delhi
END:VEVENT
BEGIN:VEVENT
DTSTART;TZID=Asia/Kolkata:20260528T110000
DTEND;TZID=Asia/Kolkata:20260528T120000
DTSTAMP:20261010T121650
CREATED:20260515T060527Z
LAST-MODIFIED:20260515T060527Z
UID:2523-1779966000-1779969600@homecse.iitd.ac.in
SUMMARY:Faster Queries\, Smarter Execution: Factorization Meets Vectorization in Modern Data Systems
DESCRIPTION:Abstract:\n\nThis talk explores a novel approach to speeding up complex database queries by combining two powerful ideas: factorization and vectorized execution. Instead of processing large intermediate results (which can be slow and memory-intensive)\, the method represents data in a compact\, factorized form that avoids redundancy. It also applies vectorized processing techniques to process data efficiently at the hardware level.\n\nSpeaker Bio:\nSunny Yasser is a PhD student in Computer Engineering at Polytechnique Montréal\, in the Data and AI Systems (DAIS) Lab under Prof. Amine Mhedhbi. He is also associated with MILA\, the premier AI research lab started by Prof. Yoshua Bengio. His research centers on high-performance query processing\, with an emphasis on compression-aware execution on modern hardware.  His broader work explores low-level system optimizations for scalable analytical and AI-driven data workloads. He has published papers at VLDB and SIGMOD previously.
URL:https://homecse.iitd.ac.in/event/faster-queries-smarter-execution-factorization-meets-vectorization-in-modern-data-systems/
LOCATION:Bharti 501\, IIT Campus\, Hauz Khas\, New Delhi
CATEGORIES:Seminars
END:VEVENT
END:VCALENDAR