10 papers
Position: It is Time to Virtualize Foundation Models with a Self-evolving Operating System Layer
Suparna Bhattacharya, Tarun Kumar, Cong Xu +7
AI applications have shifted from single, monolithic foundation models (FM) to compound agentic systems. Yet today's stacks remain fragmented: even as protocols (e.g., MCP, A2A) ea…
Neuro-symbolic PRM: Enhancing Scientific Reasoning via Structured Traces and Symbolic Verification
Yuxin Zi, Cong Xu, Suparna Bhattacharya +2
While tool-augmented Large Language Models have significantly improved multi-step reasoning in quantitative STEM tasks, a critical residual failure mode remains: intermediate reaso…
MPSelectTune: Prompt-type Selection for Fine-tuning improves Concept Unlearning in LLMs
Shubhadip Nag, Srinjoy Das, Agniva Saha +5
LLMs can be conveniently adapted to a diverse set of tasks, e.g, prediction, question-answering tasks, etc, using appropriate prompts with few-shot examples. Biased or harmful conc…
Latent Performance Profiling of Large Language Models
Tanmoy Chakraborty, Ayan Sengupta, Suparna Bhattacharya +7
Large language models (LLMs) frequently achieve impressive scores on standardized benchmarks, yet accuracy alone offers a limited view of their capabilities. Evaluating open-source…
Hard to See, Hard to Label: Generative and Symbolic Acquisition for Subtle Visual Phenomena
Renjith Prasad, Rishabh Sharma, Andrew E. Shao +8
Subtle visual anomalies such as hairline cracks, sub-millimeter voids, and low-contrast inclusions are structurally atypical yet visually ambiguous, making them both difficult to a…
Erasing CLIP Memories: Non-Destructive, Data-Free Zero-Shot class Unlearning in CLIP Models
Ashish Mishra, Tarun Kumar, Gyanaranjan Nayak +3
We introduce a novel, closed-form approach for selective unlearning in multimodal models, specifically targeting pretrained models such as CLIP. Our method leverages nullspace proj…
Selective, Controlled and Domain-Agnostic Unlearning in Pretrained CLIP: A Training- and Data-Free Approach
Ashish Mishra, Gyanaranjan Nayak, Tarun Kumar +3
Pretrained models like CLIP have demonstrated impressive zero-shot classification capabilities across diverse visual domains, spanning natural images, artistic renderings, and abst…
Forecast2Anomaly (F2A): Adapting Multivariate Time Series Foundation Models for Anomaly Prediction
Atif Hassan, Tarun Kumar, Ashish Mishra +7
Forecasting anomalies (anomaly prediction) in multivariate time series from different real-world, dynamic, and complex systems is vital for preempting critical failures, leading to…
Data Efficient Evaluation of Large Language Models and Text-to-Image Models via Adaptive Sampling
Cong Xu, Gayathri Saranathan, Mahammad Parwez Alam +5
Evaluating LLMs and text-to-image models is a computationally intensive task often overlooked. Efficient evaluation is crucial for understanding the diverse capabilities of these m…
VTruST: Controllable value function based subset selection for Data-Centric Trustworthy AI
Soumi Das, Shubhadip Nag, Shreyyash Sharma +2
Trustworthy AI is crucial to the widespread adoption of AI in high-stakes applications with fairness, robustness, and accuracy being some of the key trustworthiness metrics. In thi…