Taesik Gong

Taesik Gong, Jongheon Jeong, Taewon Kim et al.

Test-time adaptation (TTA) is an emerging paradigm that addresses distributional shifts between training and testing phases without additional data acquisition or labeling cost; only unlabeled test data streams are used for continual model adaptation. Previous TTA schemes assume that the test samples are independent and identically distributed (i.i.d.), even though they are often temporally correlated (non-i.i.d.) in application scenarios, e.g., autonomous driving. We discover that most existing TTA methods fail dramatically under such scenarios. Motivated by this, we present a new test-time adaptation scheme that is robust against non-i.i.d. test data streams. Our novelty is mainly two-fold: (a) Instance-Aware Batch Normalization (IABN) that corrects normalization for out-of-distribution samples, and (b) Prediction-balanced Reservoir Sampling (PBRS) that simulates i.i.d. data stream from non-i.i.d. stream in a class-balanced manner. Our evaluation with various datasets, including real-world non-i.i.d. streams, demonstrates that the proposed robust TTA not only outperforms state-of-the-art TTA algorithms in the non-i.i.d. setting, but also achieves comparable performance to those algorithms under the i.i.d. assumption. Code is available at https://github.com/TaesikGong/NOTE.

Taesik Gong, Yewon Kim, Taeckyung Lee et al.

Test-time adaptation (TTA) aims to address distributional shifts between training and testing data using only unlabeled test data streams for continual model adaptation. However, most TTA methods assume benign test streams, while test samples could be unexpectedly diverse in the wild. For instance, an unseen object or noise could appear in autonomous driving. This leads to a new threat to existing TTA algorithms; we found that prior TTA algorithms suffer from those noisy test samples as they blindly adapt to incoming samples. To address this problem, we present Screening-out Test-Time Adaptation (SoTTA), a novel TTA algorithm that is robust to noisy samples. The key enabler of SoTTA is two-fold: (i) input-wise robustness via high-confidence uniform-class sampling that effectively filters out the impact of noisy samples and (ii) parameter-wise robustness via entropy-sharpness minimization that improves the robustness of model parameters against large gradients from noisy samples. Our evaluation with standard TTA benchmarks with various noisy scenarios shows that our method outperforms state-of-the-art TTA methods under the presence of noisy samples and achieves comparable accuracy to those methods without noisy samples. The source code is available at https://github.com/taeckyung/SoTTA .

Hyungjun Yoon, Biniyam Aschalew Tolera, Taesik Gong et al.

Large language models (LLMs) have demonstrated exceptional abilities across various domains. However, utilizing LLMs for ubiquitous sensing applications remains challenging as existing text-prompt methods show significant performance degradation when handling long sensor data sequences. We propose a visual prompting approach for sensor data using multimodal LLMs (MLLMs). We design a visual prompt that directs MLLMs to utilize visualized sensor data alongside the target sensory task descriptions. Additionally, we introduce a visualization generator that automates the creation of optimal visualizations tailored to a given sensory task, eliminating the need for prior task-specific knowledge. We evaluated our approach on nine sensory tasks involving four sensing modalities, achieving an average of 10% higher accuracy than text-based prompts and reducing token costs by 15.8 times. Our findings highlight the effectiveness and cost-efficiency of visual prompts with MLLMs for various sensory tasks. The source code is available at https://github.com/diamond264/ByMyEyes.

Taesik Gong, Josh Belanich, Krishna Somandepalli et al.

Speech emotion recognition (SER) models typically rely on costly human-labeled data for training, making scaling methods to large speech datasets and nuanced emotion taxonomies difficult. We present LanSER, a method that enables the use of unlabeled data by inferring weak emotion labels via pre-trained large language models through weakly-supervised learning. For inferring weak labels constrained to a taxonomy, we use a textual entailment approach that selects an emotion label with the highest entailment score for a speech transcript extracted via automatic speech recognition. Our experimental results show that models pre-trained on large datasets with this weak supervision outperform other baseline models on standard SER datasets when fine-tuned, and show improved label efficiency. Despite being pre-trained on labels derived only from text, we show that the resulting representations appear to model the prosodic content of speech.

Hyungjun Yoon, Hyeongheon Cha, Hoang C. Nguyen et al.

Pre-training representations acquired via self-supervised learning could achieve high accuracy on even tasks with small training data. Unlike in vision and natural language processing domains, pre-training for IMU-based applications is challenging, as there are few public datasets with sufficient size and diversity to learn generalizable representations. To overcome this problem, we propose IMG2IMU that adapts pre-trained representation from large-scale images to diverse IMU sensing tasks. We convert the sensor data into visually interpretable spectrograms for the model to utilize the knowledge gained from vision. We further present a sensor-aware pre-training method for images that enables models to acquire particularly impactful knowledge for IMU sensing applications. This involves using contrastive learning on our augmentation set customized for the properties of sensor data. Our evaluation with four different IMU sensing tasks shows that IMG2IMU outperforms the baselines pre-trained on sensor data by an average of 9.6%p F1-score, illustrating that vision knowledge can be usefully incorporated into IMU sensing applications where only limited training data is available.

Changmin Lee, Jaemin Kim, Taesik Gong

With the rapid emergence of personal AI agents based on Large Language Models (LLMs), implementing them on-device has become essential for privacy and responsiveness. To handle the inherently personal and context-dependent nature of real-world requests, such agents must ground their generation in device-resident personal context. However, under tight memory budgets, the core bottleneck is what to store so that retrieval remains aligned with the user. We propose EPIC (Efficient Preference-aligned Index Construction), which focuses on user preferences as a compact and stable form of personal context and integrates them throughout the RAG pipeline. EPIC selectively retains preference-relevant information from raw data and aligns retrieval toward preference-aligned contexts. Across four benchmarks covering conversations, debates, explanations, and recommendations, EPIC reduces indexing memory by 2,404 times, improves preference-following accuracy by 20.17 percentage points, and achieves 33.33 times lower retrieval latency over the best-performing baseline. In our on-device experiment, EPIC maintains a memory footprint under 1 MB with 29.35 ms/query latency in streaming updates.

Taeckyung Lee, Sorn Chottananurak, Taesik Gong et al.

Test-time adaptation (TTA) has emerged as a viable solution to adapt pre-trained models to domain shifts using unlabeled test data. However, TTA faces challenges of adaptation failures due to its reliance on blind adaptation to unknown test samples in dynamic scenarios. Traditional methods for out-of-distribution performance estimation are limited by unrealistic assumptions in the TTA context, such as requiring labeled data or re-training models. To address this issue, we propose AETTA, a label-free accuracy estimation algorithm for TTA. We propose the prediction disagreement as the accuracy estimate, calculated by comparing the target model prediction with dropout inferences. We then improve the prediction disagreement to extend the applicability of AETTA under adaptation failures. Our extensive evaluation with four baselines and six TTA methods demonstrates that AETTA shows an average of 19.8%p more accurate estimation compared with the baselines. We further demonstrate the effectiveness of accuracy estimation with a model recovery case study, showcasing the practicality of our model recovery based on accuracy estimation. The source code is available at https://github.com/taeckyung/AETTA.

Taesik Gong, Fahim Kawsar, Chulhong Min

Tiny machine learning (TinyML) aims to run ML models on small devices and is increasingly favored for its enhanced privacy, reduced latency, and low cost. Recently, the advent of tiny AI accelerators has revolutionized the TinyML field by significantly enhancing hardware processing power. These accelerators, equipped with multiple parallel processors and dedicated per-processor memory instances, offer substantial performance improvements over traditional microcontroller units (MCUs). However, their limited data memory often necessitates downsampling input images, resulting in accuracy degradation. To address this challenge, we propose Data channel EXtension (DEX), a novel approach for efficient CNN execution on tiny AI accelerators. DEX incorporates additional spatial information from original images into input images through patch-wise even sampling and channel-wise stacking, effectively extending data across input channels. By leveraging underutilized processors and data memory for channel extension, DEX facilitates parallel execution without increasing inference latency. Our evaluation with four models and four datasets on tiny AI accelerators demonstrates that this simple idea improves accuracy on average by 3.5%p while keeping the inference latency the same on the AI accelerator. The source code is available at https://github.com/Nokia-Bell-Labs/data-channel-extension.

Eunsu Baek, Sunghwan Han, Taesik Gong et al.

Domain shift remains a persistent challenge in deep-learning-based computer vision, often requiring extensive model modifications or large labeled datasets to address. Inspired by human visual perception, which adjusts input quality through corrective lenses rather than over-training the brain, we propose Lens, a novel camera sensor control method that enhances model performance by capturing high-quality images from the model's perspective rather than relying on traditional human-centric sensor control. Lens is lightweight and adapts sensor parameters to specific models and scenes in real-time. At its core, Lens utilizes VisiT, a training-free, model-specific quality indicator that evaluates individual unlabeled samples at test time using confidence scores without additional adaptation costs. To validate Lens, we introduce ImageNet-ES Diverse, a new benchmark dataset capturing natural perturbations from varying sensor and lighting conditions. Extensive experiments on both ImageNet-ES and our new ImageNet-ES Diverse show that Lens significantly improves model accuracy across various baseline schemes for sensor control and model modification while maintaining low latency in image captures. Lens effectively compensates for large model size differences and integrates synergistically with model improvement techniques. Our code and dataset are available at github.com/Edw2n/Lens.git.

Joonha Park, Jiseung Jeong, Taesik Gong

Vision-Language-Action (VLA) policies are typically evaluated as if the user had finished typing or speaking before the robot begins acting. In real deployment, however, users take several seconds to enter a request, leaving the policy idle for a substantial fraction of the interaction. We introduce Premover, a lightweight module that converts this idle window into useful precomputation. Premover keeps the VLA backbone frozen and attaches two small projection heads, one for image patches, one for language tokens, that map an intermediate layer of the backbone into a shared space. The resulting focus map is supervised by simulator-rendered target-object segmentation masks and applied as a per-patch reweighting of the next step's image tokens. A single scalar readiness threshold, trained jointly from streaming prefixes, decides when the policy should begin acting. On the LIBERO benchmark suite, Premover reduces mean wall-clock time from 34.0 to 29.4 seconds, a 13.6% reduction, while matching the full-prompt baseline's success rate (95.1% vs. 95.0%); naive premoving, by contrast, collapses to 66.4%.

Taeckyung Lee, Sorn Chottananurak, Junsu Kim et al.

Deep learning models perform poorly when domain shifts exist between training and test data. Test-time adaptation (TTA) is a paradigm to mitigate this issue by adapting pre-trained models using only unlabeled test samples. However, existing TTA methods can fail under severe domain shifts, while recent active TTA approaches requiring full-class labels are impractical due to high labeling costs. To address this issue, we introduce a new setting of TTA with binary feedback. This setting uses a few binary feedback inputs from annotators to indicate whether model predictions are correct, thereby significantly reducing the labeling burden of annotators. Under the setting, we propose BiTTA, a novel dual-path optimization framework that leverages reinforcement learning to balance binary feedback-guided adaptation on uncertain samples with agreement-based self-adaptation on confident predictions. Experiments show BiTTA achieves 13.3%p accuracy improvements over state-of-the-art baselines, demonstrating its effectiveness in handling severe distribution shifts with minimal labeling effort. The source code is available at https://github.com/taeckyung/BiTTA.

Hyeongheon Cha, Dong Min Kim, Hye Won Chung et al.

Test-Time Adaptation (TTA) adjusts models using unlabeled test data to handle dynamic distribution shifts. However, existing methods rely on frequent adaptation and high computational cost, making them unsuitable for resource-constrained edge environments. To address this, we propose SNAP, a sparse TTA framework that reduces adaptation frequency and data usage while preserving accuracy. SNAP maintains competitive accuracy even when adapting based on only 1% of the incoming data stream, demonstrating its robustness under infrequent updates. Our method introduces two key components: (i) Class and Domain Representative Memory (CnDRM), which identifies and stores a small set of samples that are representative of both class and domain characteristics to support efficient adaptation with limited data; and (ii) Inference-only Batch-aware Memory Normalization (IoBMN), which dynamically adjusts normalization statistics at inference time by leveraging these representative samples, enabling efficient alignment to shifting target domains. Integrated with five state-of-the-art TTA algorithms, SNAP reduces latency by up to 93.12%, while keeping the accuracy drop below 3.3%, even across adaptation rates ranging from 1% to 50%. This demonstrates its strong potential for practical use on edge devices serving latency-sensitive applications. The source code is available at https://github.com/chahh9808/SNAP.

Taesik Gong, Si Young Jang, Utku Günay Acer et al.

The advent of tiny artificial intelligence (AI) accelerators enables AI to run at the extreme edge, offering reduced latency, lower power cost, and improved privacy. When integrated into wearable devices, these accelerators open exciting opportunities, allowing various AI apps to run directly on the body. We present Synergy that provides AI apps with best-effort performance via system-driven holistic collaboration over AI accelerator-equipped wearables. To achieve this, Synergy provides device-agnostic programming interfaces to AI apps, giving the system visibility and controllability over the app's resource use. Then, Synergy maximizes the inference throughput of concurrent AI models by creating various execution plans for each app considering AI accelerator availability and intelligently selecting the best set of execution plans. Synergy further improves throughput by leveraging parallelization opportunities over multiple computation units. Our evaluations with 7 baselines and 8 models demonstrate that, on average, Synergy achieves a 23.0 times improvement in throughput, while reducing latency by 73.9% and power consumption by 15.8%, compared to the baselines.

Eunsu Baek, Keondo Park, Jeonggil Ko et al.

Current AI advances largely rely on scaling neural models and expanding training datasets to achieve generalization and robustness. Despite notable successes, this paradigm incurs significant environmental, economic, and ethical costs, limiting sustainability and equitable access. Inspired by biological sensory systems, where adaptation occurs dynamically at the input (e.g., adjusting pupil size, refocusing vision)--we advocate for adaptive sensing as a necessary and foundational shift. Adaptive sensing proactively modulates sensor parameters (e.g., exposure, sensitivity, multimodal configurations) at the input level, significantly mitigating covariate shifts and improving efficiency. Empirical evidence from recent studies demonstrates that adaptive sensing enables small models (e.g., EfficientNet-B0) to surpass substantially larger models (e.g., OpenCLIP-H) trained with significantly more data and compute. We (i) outline a roadmap for broadly integrating adaptive sensing into real-world applications spanning humanoid, healthcare, autonomous systems, agriculture, and environmental monitoring, (ii) critically assess technical and ethical integration challenges, and (iii) propose targeted research directions, such as standardized benchmarks, real-time adaptive algorithms, multimodal integration, and privacy-preserving methods. Collectively, these efforts aim to transition the AI community toward sustainable, robust, and equitable artificial intelligence systems.

Yoojin Kwon, Hongjun Suh, Wooseok Lee et al.

Modern on-device neural network applications must operate under resource constraints while adapting to unpredictable domain shifts. However, this combined challenge-model compression and domain adaptation-remains largely unaddressed, as prior work has tackled each issue in isolation: compressed networks prioritize efficiency within a fixed domain, whereas large, capable models focus on handling domain shifts. In this work, we propose CoDA, a frequency composition-based framework that unifies compression and domain adaptation. During training, CoDA employs quantization-aware training (QAT) with low-frequency components, enabling a compressed model to selectively learn robust, generalizable features. At test time, it refines the compact model in a source-free manner (i.e., test-time adaptation, TTA), leveraging the full-frequency information from incoming data to adapt to target domains while treating high-frequency components as domain-specific cues. LFC are aligned with the trained distribution, while HFC unique to the target distribution are solely utilized for batch normalization. CoDA can be integrated synergistically into existing QAT and TTA methods. CoDA is evaluated on widely used domain-shift benchmarks, including CIFAR10-C and ImageNet-C, across various model architectures. With significant compression, it achieves accuracy improvements of 7.96%p on CIFAR10-C and 5.37%p on ImageNet-C over the full-precision TTA baseline.

Hyungjun Yoon, Jaehyun Kwak, Biniyam Aschalew Tolera et al.

Self-supervised learning has emerged as a method for utilizing massive unlabeled data for pre-training models, providing an effective feature extractor for various mobile sensing applications. However, when deployed to end-users, these models encounter significant domain shifts attributed to user diversity. We investigate the performance degradation that occurs when self-supervised models are fine-tuned in heterogeneous domains. To address the issue, we propose SelfReplay, a few-shot domain adaptation framework for personalizing self-supervised models. SelfReplay proposes self-supervised meta-learning for initial model pre-training, followed by a user-side model adaptation by replaying the self-supervision with user-specific data. This allows models to adjust their pre-trained representations to the user with only a few samples. Evaluation with four benchmarks demonstrates that SelfReplay outperforms existing baselines by an average F1-score of 8.8%p. Our on-device computational overhead analysis on a commodity off-the-shelf (COTS) smartphone shows that SelfReplay completes adaptation within an unobtrusive latency (in three minutes) with only a 9.54% memory consumption, demonstrating the computational efficiency of the proposed method.

Taesik Gong, Yewon Kim, Adiba Orzikulova et al.

Many applications utilize sensors in mobile devices and machine learning to provide novel services. However, various factors such as different users, devices, and environments impact the performance of such applications, thus making the domain shift (i.e., distributional shift between the training domain and the target domain) a critical issue in mobile sensing. Despite attempts in domain adaptation to solve this challenging problem, their performance is unreliable due to the complex interplay among diverse factors. In principle, the performance uncertainty can be identified and redeemed by performance validation with ground-truth labels. However, it is infeasible for every user to collect high-quality, sufficient labeled data. To address the issue, we present DAPPER (Domain AdaPtation Performance EstimatoR) that estimates the adaptation performance in a target domain with only unlabeled target data. Our key idea is to approximate the model performance based on the mutual information between the model inputs and corresponding outputs. Our evaluation with four real-world sensing datasets compared against six baselines shows that on average, DAPPER outperforms the state-of-the-art baseline by 39.8% in estimation accuracy. Moreover, our on-device experiment shows that DAPPER achieves up to 396X less computation overhead compared with the baselines.