13 papers
Inner Momentum for Differentially Private Muon
Bishnu Bhusal, Minh Vu, Ben Southworth +3
Differentially private training clips each per-example gradient before adding noise. This clipping is radial for each example, yet unequal clipping factors can distort the relative…
One Model, Two Physical Stories: Auditing Misalignment in Multi-Modal World Modeling
Geigh Zollicoffer, Minh Vu, Rajiv Ranasinghe +1
World models, systems that generate what happens next given current environmental conditions, are increasingly being implemented with multi-modal generation in mind. However, gener…
Sanity Checks for Long-Form Hallucination Detection
Geigh Zollicoffer, Minh Vu, Hongli Zhan +2
Hallucination detection methods for large language models increasingly operate on chain-of-thought reasoning traces, yet it remains unclear whether they evaluate the reasoning itse…
World Model Robustness via Surprise Recognition
Geigh Zollicoffer, Tanush Chopra, Mingkuan Yan +3
AI systems deployed in the real world must contend with distractions and out-of-distribution (OOD) noise that can destabilize their policies and lead to unsafe behavior. While robu…
HalluField: Detecting LLM Hallucinations via Field-Theoretic Modeling
Minh Vu, Brian K. Tran, Syed A. Shah +3
Large Language Models (LLMs) exhibit impressive reasoning and question-answering capabilities. However, they often produce inaccurate or unreliable content known as hallucinations.…
MTRE: Multi-Token Reliability Estimation for Hallucination Detection in VLMs
Geigh Zollicoffer, Minh Vu, Manish Bhattarai
Vision-language models (VLMs) now rival human performance on many multimodal tasks, yet they still hallucinate objects or generate unsafe text. Current hallucination detectors, e.g…
Topological Signatures of Adversaries in Multimodal Alignments
Minh Vu, Geigh Zollicoffer, Huy Mai +3
Multimodal Machine Learning systems, particularly those aligning text and image data like CLIP/BLIP models, have become increasingly prevalent, yet remain susceptible to adversaria…
LoRID: Low-Rank Iterative Diffusion for Adversarial Purification
Geigh Zollicoffer, Minh Vu, Ben Nebgen +3
This work presents an information-theoretic examination of diffusion-based purification methods, the state-of-the-art adversarial defenses that utilize diffusion models to remove m…
LaFA: Latent Feature Attacks on Non-negative Matrix Factorization
Minh Vu, Ben Nebgen, Erik Skau +5
As Machine Learning (ML) applications rapidly grow, concerns about adversarial attacks compromising their reliability have gained significant attention. One unsupervised ML method…
Towards Faster Matrix Diagonalization with Graph Isomorphism Networks and the AlphaZero Framework
Geigh Zollicoffer, Kshitij Bhatta, Manish Bhattarai +4
In this paper, we introduce innovative approaches for accelerating the Jacobi method for matrix diagonalization, specifically through the formulation of large matrix diagonalizatio…
Accelerating Matrix Diagonalization through Decision Transformers with Epsilon-Greedy Optimization
Kshitij Bhatta, Geigh Zollicoffer, Manish Bhattarai +4
This paper introduces a novel framework for matrix diagonalization, recasting it as a sequential decision-making problem and applying the power of Decision Transformers (DTs). Our…
Is Exploration All You Need? Effective Exploration Characteristics for Transfer in Reinforcement Learning
Jonathan C. Balloch, Rishav Bhagat, Geigh Zollicoffer +3
In deep reinforcement learning (RL) research, there has been a concerted effort to design more efficient and productive exploration methods while solving sparse-reward problems. Th…
Novelty Detection in Reinforcement Learning with World Models
Geigh Zollicoffer, Kenneth Eaton, Jonathan Balloch +4
Reinforcement learning (RL) using world models has found significant recent successes. However, when a sudden change to world mechanics or properties occurs then agent performance…