arXiv subjects

Arnuv Tandon

Publications and source records attributed to Arnuv Tandon.

7 recordsLinked to original sources

arXiv2025

End-to-End Test-Time Training for Long Context

We formulate long-context language modeling as a problem in continual learning rather than architecture design. Under this formulation, we only use a standard architecture -- a Transformer with sliding-window attention. However, our model continues learning at test time via next-token prediction on the given context, compressing the context it reads into its weights. In addition, we improve the model's initialization for learning at test time via meta-learning at training time. Overall, our method, a form of Test-Time Training (TTT), is End-to-End (E2E) both at test time (via next-token prediction) and training time (via meta-learning), in contrast to previous forms. We conduct extensive experiments with a focus on scaling properties. In particular, for 3B models trained with 164B tokens, our method (TTT-E2E) scales with context length in the same way as Transformer with full attention, while others, such as Mamba 2 and Gated DeltaNet, do not. However, similar to RNNs, TTT-E2E has constant inference latency regardless of context length, making it 2.7 times faster than full attention for 128K context. Our code is publicly available.

cs.LG↗

arXiv2025

Towards Real-Time Inference of Thin Liquid Film Thickness Profiles from Interference Patterns Using Vision Transformers

Thin film interferometry is a powerful technique for non-invasively measuring liquid film thickness with applications in ophthalmology, but its clinical translation is hindered by the challenges in reconstructing thickness profiles from interference patterns - an ill-posed inverse problem complicated by phase periodicity, imaging noise and ambient artifacts. Traditional reconstruction methods are either computationally intensive, sensitive to noise, or require manual expert analysis, which is impractical for real-time diagnostics. To address this challenge, here we present a vision transformer-based approach for real-time inference of thin liquid film thickness profiles directly from isolated interferograms. Trained on a hybrid dataset combining physiologically-relevant synthetic and experimental tear film data, our model leverages long-range spatial correlations to resolve phase ambiguities and reconstruct temporally coherent thickness profiles in a single forward pass from dynamic interferograms acquired in vivo and ex vivo. The network demonstrates state-of-the-art performance on noisy, rapidly-evolving films with motion artifacts, overcoming limitations of conventional phase-unwrapping and iterative fitting methods. Our data-driven approach enables automated, consistent thickness reconstruction at real-time speeds on consumer hardware, opening new possibilities for continuous monitoring of pre-lens ocular tear films and non-invasive diagnosis of conditions such as the dry eye disease.

cs.CV↗

arXiv2025

LoRTA: Low Rank Tensor Adaptation of Large Language Models

Low Rank Adaptation (LoRA) is a popular Parameter Efficient Fine Tuning (PEFT) method that effectively adapts large pre-trained models for downstream tasks. LoRA parameterizes model updates using low-rank matrices at each layer, significantly reducing the number of trainable parameters and, consequently, resource requirements during fine-tuning. However, the lower bound on the number of trainable parameters remains high due to the use of the low-rank matrix model. Recent works have addressed this limitation by proposing low rank tensor parameterizations for model updates. However, they only exploit redundancy across layers, or tensorize individual matrices using ad-hoc schemes that introduce additional hyperparameters. In this work, we propose a higher-order Candecomp/Parafac (CP) decomposition, enabling a more compact and flexible representation compared to existing matrix and tensor based PEFT methods. Our experiments on Natural Language Understanding, Instruction Tuning, Preference Optimization and Protein Folding benchmarks demonstrate that our method can achieve a reduction in the number of parameters while maintaining comparable performance.

cs.CL↗

arXiv2025

Test-Time Training on Video Streams

Prior work has established Test-Time Training (TTT) as a general framework to further improve a trained model at test time. Before making a prediction on each test instance, the model is first trained on the same instance using a self-supervised task such as reconstruction. We extend TTT to the streaming setting, where multiple test instances - video frames in our case - arrive in temporal order. Our extension is online TTT: The current model is initialized from the previous model, then trained on the current frame and a small window of frames immediately before. Online TTT significantly outperforms the fixed-model baseline for four tasks, on three real-world datasets. The improvements are more than 2.2x and 1.5x for instance and panoptic segmentation. Surprisingly, online TTT also outperforms its offline variant that accesses strictly more information, training on all frames from the entire test video regardless of temporal order. This finding challenges those in prior work using synthetic videos. We formalize a notion of locality as the advantage of online over offline TTT, and analyze its role with ablations and a theory based on bias-variance trade-off.

cs.CV↗

arXiv2023

Deceptive Alignment Monitoring

As the capabilities of large machine learning models continue to grow, and as the autonomy afforded to such models continues to expand, the spectre of a new adversary looms: the models themselves. The threat that a model might behave in a seemingly reasonable manner, while secretly and subtly modifying its behavior for ulterior reasons is often referred to as deceptive alignment in the AI Safety & Alignment communities. Consequently, we call this new direction Deceptive Alignment Monitoring. In this work, we identify emerging directions in diverse machine learning subfields that we believe will become increasingly important and intertwined in the near future for deceptive alignment monitoring, and we argue that advances in these fields present both long-term challenges and new research opportunities. We conclude by advocating for greater involvement by the adversarial machine learning community in these emerging directions.

cs.LG↗

arXiv2023

FACADE: A Framework for Adversarial Circuit Anomaly Detection and Evaluation

We present FACADE, a novel probabilistic and geometric framework designed for unsupervised mechanistic anomaly detection in deep neural networks. Its primary goal is advancing the understanding and mitigation of adversarial attacks. FACADE aims to generate probabilistic distributions over circuits, which provide critical insights to their contribution to changes in the manifold properties of pseudo-classes, or high-dimensional modes in activation space, yielding a powerful tool for uncovering and combating adversarial attacks. Our approach seeks to improve model robustness, enhance scalable model oversight, and demonstrates promising applications in real-world deployment settings.

cs.LG↗

arXiv2023

Interfacial Stresses on Droplet Interface Bilayers Using Two Photon Fluorescence Lifetime Imaging Microscopy

Response of lipid bilayers to external mechanical stimuli is an active area of research with implications for fundamental and synthetic cell biology. However, there is a lack of tools for systematically imposing mechanical strains and non-invasively mapping out interfacial (membrane) stress distributions on lipid bilayers. In this article, we report a miniature platform to manipulate model cell membranes in the form of droplet interface bilayers (DIBs), and non-invasively measure spatio-temporally resolved interfacial stresses using two photon fluorescence lifetime imaging of an interfacially active molecular flipper (Flipper-TR). We established the effectiveness of the developed framework by investigating interfacial stresses accompanying three key processes associated with DIBs: thin film drainage between lipid monolayer coated droplets, bilayer formation, and bilayer separation. Interestingly, the measurements also revealed fundamental aspects of DIBs including the existence of a radially decaying interfacial stress distribution post bilayer formation, and the simultaneous build up and decay of stress respectively at the bilayer corner and center during bilayer separation. Finally, utilizing interfacial rheology measurements and MD simulations, we also reveal that the tested molecular flipper is sensitive to membrane fluidity that changes with interfacial stress - expanding the scientific understanding of how molecular motors sense stress.

cond-mat.soft↗