Jaehoon Lee

Aarohi Srivastava, Abhinav Rastogi, Abhishek Rao et al. · allen-ai, amazon-science

Language models demonstrate both quantitative improvement and new qualitative capabilities with increasing scale. Despite their potentially transformative impact, these new capabilities are as yet poorly characterized. In order to inform future research, prepare for disruptive new model capabilities, and ameliorate socially harmful effects, it is vital that we understand the present and near-future capabilities and limitations of language models. To address this challenge, we introduce the Beyond the Imitation Game benchmark (BIG-bench). BIG-bench currently consists of 204 tasks, contributed by 450 authors across 132 institutions. Task topics are diverse, drawing problems from linguistics, childhood development, math, common-sense reasoning, biology, physics, social bias, software development, and beyond. BIG-bench focuses on tasks that are believed to be beyond the capabilities of current language models. We evaluate the behavior of OpenAI's GPT models, Google-internal dense transformer architectures, and Switch-style sparse transformers on BIG-bench, across model sizes spanning millions to hundreds of billions of parameters. In addition, a team of human expert raters performed all tasks in order to provide a strong baseline. Findings include: model performance and calibration both improve with scale, but are poor in absolute terms (and when compared with rater performance); performance is remarkably similar across model classes, though with benefits from sparsity; tasks that improve gradually and predictably commonly involve a large knowledge or memorization component, whereas tasks that exhibit "breakthrough" behavior at a critical scale often involve multiple steps or components, or brittle metrics; social bias typically increases with scale in settings with ambiguous context, but this can be improved with prompting.

Mitchell Wortsman, Peter J. Liu, Lechao Xiao et al. · anthropic, deepmind

Teams that have trained large Transformer-based models have reported training instabilities at large scale that did not appear when training with the same hyperparameters at smaller scales. Although the causes of such instabilities are of scientific interest, the amount of resources required to reproduce them has made investigation difficult. In this work, we seek ways to reproduce and study training stability and instability at smaller scales. First, we focus on two sources of training instability described in previous work: the growth of logits in attention layers (Dehghani et al., 2023) and divergence of the output logits from the log probabilities (Chowdhery et al., 2022). By measuring the relationship between learning rate and loss across scales, we show that these instabilities also appear in small models when training at high learning rates, and that mitigations previously employed at large scales are equally effective in this regime. This prompts us to investigate the extent to which other known optimizer and model interventions influence the sensitivity of the final loss to changes in the learning rate. To this end, we study methods such as warm-up, weight decay, and the $μ$Param (Yang et al., 2022), and combine techniques to train small models that achieve similar losses across orders of magnitude of learning rate variation. Finally, to conclude our exploration we study two cases where instabilities can be predicted before they emerge by examining the scaling behavior of model activation and gradient norms.

Katie Everett, Lechao Xiao, Mitchell Wortsman et al. · anthropic, deepmind

Robust and effective scaling of models from small to large width typically requires the precise adjustment of many algorithmic and architectural details, such as parameterization and optimizer choices. In this work, we propose a new perspective on parameterization by investigating a key assumption in prior work about the alignment between parameters and data and derive new theoretical results under weaker assumptions and a broader set of optimizers. Our extensive empirical investigation includes tens of thousands of models trained with all combinations of three optimizers, four parameterizations, several alignment assumptions, more than a dozen learning rates, and fourteen model sizes up to 26.8B parameters. We find that the best learning rate scaling prescription would often have been excluded by the assumptions in prior work. Our results show that all parameterizations, not just maximal update parameterization (muP), can achieve hyperparameter transfer; moreover, our novel per-layer learning rate prescription for standard parameterization outperforms muP. Finally, we demonstrate that an overlooked aspect of parameterization, the epsilon parameter in Adam, must be scaled correctly to avoid gradient underflow and propose Adam-atan2, a new numerically stable, scale-invariant version of Adam that eliminates the epsilon hyperparameter entirely.

Insu Han, Amir Zandieh, Jaehoon Lee et al. · deepmind

Infinite width limit has shed light on generalization and optimization aspects of deep learning by establishing connections between neural networks and kernel methods. Despite their importance, the utility of these kernel methods was limited in large-scale learning settings due to their (super-)quadratic runtime and memory complexities. Moreover, most prior works on neural kernels have focused on the ReLU activation, mainly due to its popularity but also due to the difficulty of computing such kernels for general activations. In this work, we overcome such difficulties by providing methods to work with general activations. First, we compile and expand the list of activation functions admitting exact dual activation expressions to compute neural kernels. When the exact computation is unknown, we present methods to effectively approximate them. We propose a fast sketching method that approximates any multi-layered Neural Network Gaussian Process (NNGP) kernel and Neural Tangent Kernel (NTK) matrices for a wide range of activation functions, going beyond the commonly analyzed ReLU activation. This is done by showing how to approximate the neural kernels using the truncated Hermite expansion of any desired activation functions. While most prior works require data points on the unit sphere, our methods do not suffer from such limitations and are applicable to any dataset of points in $\mathbb{R}^d$. Furthermore, we provide a subspace embedding for NNGP and NTK matrices with near input-sparsity runtime and near-optimal target dimension which applies to any \emph{homogeneous} dual activation functions with rapidly convergent Taylor expansion. Empirically, with respect to exact convolutional NTK (CNTK) computation, our method achieves $106\times$ speedup for approximate CNTK of a 5-layer Myrtle network on CIFAR-10 dataset.

Jiri Hron, Laura Culp, Gamaleldin Elsayed et al. · anthropic, deepmind

While many capabilities of language models (LMs) improve with increased training budget, the influence of scale on hallucinations is not yet fully understood. Hallucinations come in many forms, and there is no universally accepted definition. We thus focus on studying only those hallucinations where a correct answer appears verbatim in the training set. To fully control the training data content, we construct a knowledge graph (KG)-based dataset, and use it to train a set of increasingly large LMs. We find that for a fixed dataset, larger and longer-trained LMs hallucinate less. However, hallucinating on $\leq5$% of the training data requires an order of magnitude larger model, and thus an order of magnitude more compute, than Hoffmann et al. (2022) reported was optimal. Given this costliness, we study how hallucination detectors depend on scale. While we see detector size improves performance on fixed LM's outputs, we find an inverse relationship between the scale of the LM and the detectability of its hallucinations.

C. Daniel Freeman, Laura Culp, Aaron Parisi et al. · anthropic, deepmind

We introduce and study the problem of adversarial arithmetic, which provides a simple yet challenging testbed for language model alignment. This problem is comprised of arithmetic questions posed in natural language, with an arbitrary adversarial string inserted before the question is complete. Even in the simple setting of 1-digit addition problems, it is easy to find adversarial prompts that make all tested models (including PaLM2, GPT4, Claude2) misbehave, and even to steer models to a particular wrong answer. We additionally provide a simple algorithm for finding successful attacks by querying those same models, which we name "prompt inversion rejection sampling" (PIRS). We finally show that models can be partially hardened against these attacks via reinforcement learning and via agentic constitutional loops. However, we were not able to make a language model fully robust against adversarial arithmetic attacks.

Ben Adlam, Jaehoon Lee, Shreyas Padhy et al. · deepmind

Neural kernels have drastically increased performance on diverse and nonstandard data modalities but require significantly more compute, which previously limited their application to smaller datasets. In this work, we address this by massively parallelizing their computation across many GPUs. We combine this with a distributed, preconditioned conjugate gradients algorithm to enable kernel regression at a large scale (i.e. up to five million examples). Using this approach, we study scaling laws of several neural kernels across many orders of magnitude for the CIFAR-5m dataset. Using data augmentation to expand the original CIFAR-10 training dataset by a factor of 20, we obtain a test accuracy of 91.2\% (SotA for a pure kernel method). Moreover, we explore neural kernels on other data modalities, obtaining results on protein and small molecule prediction tasks that are competitive with SotA methods.

Seunghan Lee, Jaehoon Lee, Jun Seo et al.

Recent advances in time series foundation models (TSFMs) demonstrate strong expressive capacity through large-scale pretraining across diverse time series domains. Zero-shot time series forecasting with TSFMs, however, exhibits limited generalization to unseen datasets, which retrieval-augmented forecasting addresses by leveraging an external knowledge base. Existing approaches rely on a fixed number of retrieved samples that may introduce irrelevant information. To this end, we propose Cross-RAG, a zero-shot retrieval-augmented forecasting framework that selectively attends to query-relevant retrieved samples. Cross-RAG models input-level relevance between the query and retrieved samples via query-retrieval cross-attention, while jointly incorporating information from the query and retrieved samples. Extensive experiments demonstrate that Cross-RAG consistently improves zero-shot forecasting performance across various TSFMs and RAG methods, and additional analyses confirm its effectiveness across diverse retrieval scenarios. Code is available at https://github.com/seunghan96/cross-rag/.

Sheo Yon Jhin, Jaehoon Lee, Minju Jo et al.

Deep learning inspired by differential equations is a recent research trend and has marked the state of the art performance for many machine learning tasks. Among them, time-series modeling with neural controlled differential equations (NCDEs) is considered as a breakthrough. In many cases, NCDE-based models not only provide better accuracy than recurrent neural networks (RNNs) but also make it possible to process irregular time-series. In this work, we enhance NCDEs by redesigning their core part, i.e., generating a continuous path from a discrete time-series input. NCDEs typically use interpolation algorithms to convert discrete time-series samples to continuous paths. However, we propose to i) generate another latent continuous path using an encoder-decoder architecture, which corresponds to the interpolation process of NCDEs, i.e., our neural network-based interpolation vs. the existing explicit interpolation, and ii) exploit the generative characteristic of the decoder, i.e., extrapolation beyond the time domain of original data if needed. Therefore, our NCDE design can use both the interpolated and the extrapolated information for downstream machine learning tasks. In our experiments with 5 real-world datasets and 12 baselines, our extrapolation and interpolation-based NCDEs outperform existing baselines by non-trivial margins.

Suyoung Bae, CheolWon Na, Jaehoon Lee et al.

As Large Language Models (LLMs) have become capable of generating long and descriptive code summaries, accurate and reliable evaluation of factual consistency has become a critical challenge. However, previous evaluation methods are primarily designed for short summaries of isolated code snippets. Consequently, they struggle to provide fine-grained evaluation of multi-sentence functionalities and fail to accurately assess dependency context commonly found in real-world code summaries. To address this, we propose ReFEree, a reference-free and fine-grained method for evaluating factual consistency in real-world code summaries. We define factual inconsistency criteria specific to code summaries and evaluate them at the segment level using these criteria along with dependency information. These segment-level results are then aggregated into a fine-grained score. We construct a code summarization benchmark with human-annotated factual consistency labels. The evaluation results demonstrate that ReFEree achieves the highest correlation with human judgment among 13 baselines, improving 15-18% over the previous state-of-the-art. Our code and data are available at https://github.com/bsy99615/ReFEree.git.

Sheo Yon Jhin, Jaehoon Lee, Noseong Park

Anomaly detection is an important field that aims to identify unexpected patterns or data points, and it is closely related to many real-world problems, particularly to applications in finance, manufacturing, cyber security, and so on. While anomaly detection has been studied extensively in various fields, detecting future anomalies before they occur remains an unexplored territory. In this paper, we present a novel type of anomaly detection, called Precursor-of-Anomaly (PoA) detection. Unlike conventional anomaly detection, which focuses on determining whether a given time series observation is an anomaly or not, PoA detection aims to detect future anomalies before they happen. To solve both problems at the same time, we present a neural controlled differential equation-based neural network and its multi-task learning algorithm. We conduct experiments using 17 baselines and 3 datasets, including regular and irregular time series, and demonstrate that our presented method outperforms the baselines in almost all cases. Our ablation studies also indicate that the multitasking training method significantly enhances the overall performance for both anomaly and PoA detection.

Charlie Snell, Jaehoon Lee, Kelvin Xu et al.

Enabling LLMs to improve their outputs by using more test-time computation is a critical step towards building generally self-improving agents that can operate on open-ended natural language. In this paper, we study the scaling of inference-time computation in LLMs, with a focus on answering the question: if an LLM is allowed to use a fixed but non-trivial amount of inference-time compute, how much can it improve its performance on a challenging prompt? Answering this question has implications not only on the achievable performance of LLMs, but also on the future of LLM pretraining and how one should tradeoff inference-time and pre-training compute. Despite its importance, little research attempted to understand the scaling behaviors of various test-time inference methods. Moreover, current work largely provides negative results for a number of these strategies. In this work, we analyze two primary mechanisms to scale test-time computation: (1) searching against dense, process-based verifier reward models; and (2) updating the model's distribution over a response adaptively, given the prompt at test time. We find that in both cases, the effectiveness of different approaches to scaling test-time compute critically varies depending on the difficulty of the prompt. This observation motivates applying a "compute-optimal" scaling strategy, which acts to most effectively allocate test-time compute adaptively per prompt. Using this compute-optimal strategy, we can improve the efficiency of test-time compute scaling by more than 4x compared to a best-of-N baseline. Additionally, in a FLOPs-matched evaluation, we find that on problems where a smaller base model attains somewhat non-trivial success rates, test-time compute can be used to outperform a 14x larger model.

Seunghan Lee, Jaehoon Lee, Jun Seo et al.

TabPFN has recently gained attention as a foundation model for tabular datasets, achieving strong performance by leveraging in-context learning on synthetic data. However, we find that TabPFN is vulnerable to label shift, often overfitting to the majority class in the training dataset. To address this limitation, we propose DistPFN, the first test-time posterior adjustment method designed for tabular foundation models. DistPFN rescales predicted class probabilities by downweighting the influence of the training prior (i.e., the class distribution of the context) and emphasizing the contribution of the model's predicted posterior, without architectural modification or additional training. We further introduce DistPFN-T, which incorporates temperature scaling to adaptively control the adjustment strength based on the discrepancy between prior and posterior. We evaluate our methods on over 250 OpenML datasets, demonstrating substantial improvements for various TabPFN-based models in classification tasks under label shift, while maintaining strong performance in standard settings without label shift. Code is available at this repository: https://github.com/seunghan96/DistPFN.

Haksoo Lim, Sewon Park, Minjung Kim et al.

The time-series anomaly detection is one of the most fundamental tasks for time-series. Unlike the time-series forecasting and classification, the time-series anomaly detection typically requires unsupervised (or self-supervised) training since collecting and labeling anomalous observations are difficult. In addition, most existing methods resort to limited forms of anomaly measurements and therefore, it is not clear whether they are optimal in all circumstances. To this end, we present a multivariate time-series anomaly detector based on score-based generative models, called MadSGM, which considers the broadest ever set of anomaly measurement factors: i) reconstruction-based, ii) density-based, and iii) gradient-based anomaly measurements. We also design a conditional score network and its denoising score matching loss for the time-series anomaly detection. Experiments on five real-world benchmark datasets illustrate that MadSGM achieves the most robust and accurate predictions.

Mitchell Wortsman, Jaehoon Lee, Justin Gilmer et al.

Previous research observed accuracy degradation when replacing the attention softmax with a point-wise activation such as ReLU. In the context of vision transformers, we find that this degradation is mitigated when dividing by sequence length. Our experiments training small to large vision transformers on ImageNet-21k indicate that ReLU-attention can approach or match the performance of softmax-attention in terms of scaling behavior as a function of compute.

Jaehoon Lee, Chan Kim, Gyumin Lee et al.

Forecasting future outcomes from recent time series data is not easy, especially when the future data are different from the past (i.e. time series are under temporal drifts). Existing approaches show limited performances under data drifts, and we identify the main reason: It takes time for a model to collect sufficient training data and adjust its parameters for complicated temporal patterns whenever the underlying dynamics change. To address this issue, we study a new approach; instead of adjusting model parameters (by continuously re-training a model on new data), we build a hypernetwork that generates other target models' parameters expected to perform well on the future data. Therefore, we can adjust the model parameters beforehand (if the hypernetwork is correct). We conduct extensive experiments with 6 target models, 6 baselines, and 4 datasets, and show that our HyperGPA outperforms other baselines.

Yongsung Kim, Wooseok Song, Jaihyun Lew et al.

Visual Geometry Grounded Transformer (VGGT) has advanced 3D vision, yet its global attention layers suffer from quadratic computational costs that hinder scalability. Several sparsification-based acceleration techniques have been proposed to alleviate this issue, but they often suffer from substantial accuracy degradation. We hypothesize that the accuracy degradation stems from the heterogeneity in head-wise sparsification sensitivity, as the existing methods apply a uniform sparsity pattern across all heads. Motivated by this hypothesis, we present a two-stage sparsification pipeline that effectively quantifies and exploits headwise sparsification sensitivity. In the first stage, we measure head-wise sparsification sensitivity using a novel metric, the Head Sensitivity Score (HeSS), which approximates the Hessian with respect to two distinct error terms on a small calibration set. In the inference stage, we perform HeSS-Guided Sparsification, leveraging the pre-computed HeSS to reallocate the total attention budget-assigning denser attention to sensitive heads and sparser attention to more robust ones. We demonstrate that HeSS effectively captures head-wise sparsification sensitivity and empirically confirm that attention heads in the global attention layers exhibit heterogeneous sensitivity characteristics. Extensive experiments further show that our method effectively mitigates performance degradation under high sparsity, demonstrating strong robustness across varying sparsification levels. Code is available at https://github.com/libary753/HeSS.

Jaehoon Lee, Sohyun Kim, Wanggeun Park et al.

Existing benchmarks for visual document retrieval (VDR) largely overlook non-English languages and the structural complexity of official publications. To address this gap, we introduce SDS KoPub VDR, the first large-scale, public benchmark for retrieving and understanding Korean public documents. The benchmark is built upon 361 real-world documents, including 256 files under the KOGL Type 1 license and 105 from official legal portals, capturing complex visual elements like tables, charts, and multi-column layouts. To establish a reliable evaluation set, we constructed 600 query-page-answer triples. These were initially generated using multimodal models (e.g., GPT-4o) and subsequently underwent human verification to ensure factual accuracy and contextual relevance. The queries span six major public domains and are categorized by the reasoning modality required: text-based, visual-based, and cross-modal. We evaluate SDS KoPub VDR on two complementary tasks: (1) text-only retrieval and (2) multimodal retrieval, which leverages visual features alongside text. This dual-task evaluation reveals substantial performance gaps, particularly in multimodal scenarios requiring cross-modal reasoning, even for state-of-the-art models. As a foundational resource, SDS KoPub VDR enables rigorous and fine-grained evaluation and provides a roadmap for advancing multimodal AI in real-world document intelligence. The dataset is available at https://huggingface.co/datasets/SamsungSDS-Research/SDS-KoPub-VDR-Benchmark.

Seunghan Lee, Jun Seo, Jaehoon Lee et al.

Temporal knowledge graphs (TKGs) represent time-stamped relational facts and support a wide range of reasoning tasks over evolving events. However, existing methods produce entity representations that are static at the entity level, in that each representation is a function of learned parameters only and retains no trace of the interactions in which the entity has participated. In this paper, we depart from this static view and propose that each entity be modeled as an adaptive process whose representation is refined every time the entity participates in a fact. To this end, we propose AdaTKG, which maintains a per-entity memory that is updated with every observed interaction, with the memory accumulating online and predictions improving as more interactions arrive. Specifically, we instantiate the memory update as a learnable exponential moving average governed by a single shared scalar instead of using learnable parameters for each entity, enabling AdaTKG to handle entities unseen during training. Extensive experiments confirm consistent gains over TKG baselines, demonstrating the effectiveness of adaptive memory. Code is publicly available at: https://github.com/seunghan96/AdaTKG.

Seunghan Lee, Jun Seo, Jaehoon Lee et al.

Recent advances in multimodal learning have motivated the integration of auxiliary modalities such as text or vision into time series (TS) forecasting. However, most existing methods provide limited gains, often improving performance only in specific datasets or relying on architecture-specific designs that limit generalization. In this paper, we show that multimodal models with naive fusion strategies (e.g., simple addition or concatenation) often underperform unimodal TS models, which we attribute to the uncontrolled integration of auxiliary modalities which may introduce irrelevant information. Motivated by this observation, we explore various constrained fusion methods designed to control such integration and find that they consistently outperform naive fusion methods. Furthermore, we propose Controlled Fusion Adapter (CFA), a simple plug-in method that enables controlled cross-modal interactions without modifying the TS backbone, integrating only relevant textual information aligned with TS dynamics. CFA employs low-rank adapters to filter irrelevant textual information before fusing it into temporal representations. We conduct over 20K experiments across various datasets and TS/text models, demonstrating the effectiveness of the constrained fusion methods including CFA. Code is publicly available at: https://github.com/seunghan96/cfa/.

Seunghan Lee, Jun Seo, Jaehoon Lee et al.

Time series (TS) reasoning models (TSRMs) have shown promising capabilities in general domains, yet they consistently fail on financial domain, which exhibit unique characteristics. We propose a general 2x2 capability taxonomy for TSRMs by crossing 1) single-entity vs. multi-entity analysis with 2) assessment of the current state vs. prediction of future behavior. We instantiate this taxonomy in the financial domain -- where the distinction between deterministic assessment and stochastic prediction is particularly critical -- as ten financial reasoning tasks, forming the FinTSR-Bench benchmark based on S&P stocks. To this end, we propose FinSTaR (Financial Time Series Thinking and Reasoning), trained on FinTSR-Bench with distinct chain-of-thought (CoT) strategies tailored to each category. For assessment, which is deterministic (i.e., computable from observable data), we employ Compute-in-CoT, a programmatic CoT that enables models to derive answers directly from raw prices. For prediction, which is inherently stochastic (i.e., subject to unobservable factors), we adopt Scenario-Aware CoT, which generates diverse scenarios before making a judgment, mirroring how financial analysts reason under uncertainty. The proposed method achieves 78.9% average accuracy on FinTSR-Bench, substantially outperforming LLM and TSRM baselines. Furthermore, we show that the four capability categories are complementary and mutually reinforcing through joint training, and that Scenario-Aware CoT consistently improves prediction accuracy over standard CoT. Code is publicly available at: https://github.com/seunghan96/FinSTaR.

Haksoo Lim, Jaehoon Lee, Sewon Park et al.

Score-based generative models (SGMs) have demonstrated unparalleled sampling quality and diversity in numerous fields, such as image generation, voice synthesis, and tabular data synthesis, etc. Inspired by those outstanding results, we apply SGMs to synthesize time-series by learning its conditional score function. To this end, we present a conditional score network for time-series synthesis, deriving a denoising score matching loss tailored for our purposes. In particular, our presented denoising score matching loss is the conditional denoising score matching loss for time-series synthesis. In addition, our framework is such flexible that both regular and irregular time-series can be synthesized with minimal changes to our model design. Finally, we obtain exceptional synthesis performance on various time-series datasets, achieving state-of-the-art sampling diversity and quality.

Jaehoon Lee, Mingi Jung, Soohyuk Jang et al.

Large Vision-Language Models (VLMs) achieve strong multimodal understanding capabilities by leveraging high-resolution visual inputs, but the resulting large number of visual tokens creates a major computational bottleneck. Recent work mitigates this issue through visual token compression, typically compressing tokens based on saliency, diversity, or a fixed combination of both. We observe that the distribution of semantic prominence varies substantially across samples, leading to different optimal trade-offs between local saliency preservation and global coverage. This observation suggests that applying a static compression strategy across all samples can be suboptimal. Motivated by this insight, we propose PromPrune, a sample-adaptive visual token selection framework composed of semantic prominence-aware budget allocation and a two-stage selection pipeline. Our method adaptively balances local saliency preservation and global coverage according to the semantic prominence distribution of each sample. By allocating token budgets between locally salient regions and globally diverse regions, our method maintains strong performance even under high compression ratios. On LLaVA-NeXT-7B, our approach reduces FLOPs by 88% and prefill latency by 22% while preserving 97.5% of the original accuracy.

Gemini Team, Petko Georgiev, Ving Ian Lei et al. · deepmind, mila

In this report, we introduce the Gemini 1.5 family of models, representing the next generation of highly compute-efficient multimodal models capable of recalling and reasoning over fine-grained information from millions of tokens of context, including multiple long documents and hours of video and audio. The family includes two new models: (1) an updated Gemini 1.5 Pro, which exceeds the February version on the great majority of capabilities and benchmarks; (2) Gemini 1.5 Flash, a more lightweight variant designed for efficiency with minimal regression in quality. Gemini 1.5 models achieve near-perfect recall on long-context retrieval tasks across modalities, improve the state-of-the-art in long-document QA, long-video QA and long-context ASR, and match or surpass Gemini 1.0 Ultra's state-of-the-art performance across a broad set of benchmarks. Studying the limits of Gemini 1.5's long-context ability, we find continued improvement in next-token prediction and near-perfect retrieval (>99%) up to at least 10M tokens, a generational leap over existing models such as Claude 3.0 (200k) and GPT-4 Turbo (128k). Finally, we highlight real-world use cases, such as Gemini 1.5 collaborating with professionals on completing their tasks achieving 26 to 75% time savings across 10 different job categories, as well as surprising new capabilities of large language models at the frontier; when given a grammar manual for Kalamang, a language with fewer than 200 speakers worldwide, the model learns to translate English to Kalamang at a similar level to a person who learned from the same content.

Jaehoon Lee, Hankook Lee, Sungik Choi et al.

When solving forecasting problems including multiple time-series features, existing approaches often fall into two extreme categories, depending on whether to utilize inter-feature information: univariate and complete-multivariate models. Unlike univariate cases which ignore the information, complete-multivariate models compute relationships among a complete set of features. However, despite the potential advantage of leveraging the additional information, complete-multivariate models sometimes underperform univariate ones. Therefore, our research aims to explore a middle ground between these two by introducing what we term Partial-Multivariate models where a neural network captures only partial relationships, that is, dependencies within subsets of all features. To this end, we propose PMformer, a Transformer-based partial-multivariate model, with its training algorithm. We demonstrate that PMformer outperforms various univariate and complete-multivariate models, providing a theoretical rationale and empirical analysis for its superiority. Additionally, by proposing an inference technique for PMformer, the forecasting accuracy is further enhanced. Finally, we highlight other advantages of PMformer: efficiency and robustness under missing features.

Jaehoon Lee, Jinsung Jeon, Sheo yon Jhin et al.

The problem of processing very long time-series data (e.g., a length of more than 10,000) is a long-standing research problem in machine learning. Recently, one breakthrough, called neural rough differential equations (NRDEs), has been proposed and has shown that it is able to process such data. Their main concept is to use the log-signature transform, which is known to be more efficient than the Fourier transform for irregular long time-series, to convert a very long time-series sample into a relatively shorter series of feature vectors. However, the log-signature transform causes non-trivial spatial overheads. To this end, we present the method of LOweR-Dimensional embedding of log-signature (LORD), where we define an NRDE-based autoencoder to implant the higher-depth log-signature knowledge into the lower-depth log-signature. We show that the encoder successfully combines the higher-depth and the lower-depth log-signature knowledge, which greatly stabilizes the training process and increases the model accuracy. In our experiments with benchmark datasets, the improvement ratio by our method is up to 75\% in terms of various classification and forecasting evaluation metrics.

Suyoung Bae, Jaehoon Lee, Changkyu Choi et al.

Automated code documentation is essential for modern software development, providing the contextual grounding that both human developers and coding agents rely on to navigate large codebases. Existing repository-level approaches process components independently, causing redundant retrieval and conflicting descriptions across documents while producing outputs that lack hierarchical structure. Therefore, we propose MemDocAgent, a long-horizon agentic framework that generates documentation within a single, integrated context spanning the entire repository. It combines two components: (i) Dependency-Aware Traversal Guiding that predetermines a traversal order respecting dependency and granularity hierarchies; (ii) Memory-Guided Agentic Interaction, in which the agent interacts with RepoMemory, a shared memory accumulating prior work traces through read, write, and verify operations. Through an in-depth multi-criteria evaluation, MemDocAgent achieves the best performance over both open and closed-source baselines and demonstrates practical applicability in real software development workflows.

Avi Singh, John D. Co-Reyes, Rishabh Agarwal et al. · anthropic, deepmind

Fine-tuning language models~(LMs) on human-generated data remains a prevalent practice. However, the performance of such models is often limited by the quantity and diversity of high-quality human data. In this paper, we explore whether we can go beyond human data on tasks where we have access to scalar feedback, for example, on math problems where one can verify correctness. To do so, we investigate a simple self-training method based on expectation-maximization, which we call ReST$^{EM}$, where we (1) generate samples from the model and filter them using binary feedback, (2) fine-tune the model on these samples, and (3) repeat this process a few times. Testing on advanced MATH reasoning and APPS coding benchmarks using PaLM-2 models, we find that ReST$^{EM}$ scales favorably with model size and significantly surpasses fine-tuning only on human data. Overall, our findings suggest self-training with feedback can substantially reduce dependence on human-generated data.

Jaehoon Lee, Seungwoo Lee, Younghwi Kim et al.

Time-series forecasting is fundamental in industrial domains like manufacturing and smart factories. As systems evolve toward automation, models must operate on edge devices (e.g., PLCs, microcontrollers) with strict constraints on latency and memory, limiting parameters to a few thousand. Conventional deep architectures are often impractical here. We propose the Fourier-Efficient Adaptive Temporal Hierarchy Forecaster (FEATHer) for accurate long-term forecasting under severe limits. FEATHer introduces: (i) ultra-lightweight multiscale decomposition into frequency pathways; (ii) a shared Dense Temporal Kernel using projection-depthwise convolution-projection without recurrence or attention; (iii) frequency-aware branch gating that adaptively fuses representations based on spectral characteristics; and (iv) a Sparse Period Kernel reconstructing outputs via period-wise downsampling to capture seasonality. FEATHer maintains a compact architecture (as few as 400 parameters) while outperforming baselines. Across eight benchmarks, it achieves the best ranking, recording 60 first-place results with an average rank of 2.05. These results demonstrate that reliable long-range forecasting is achievable on constrained edge hardware, offering a practical direction for industrial real-time inference.

Jascha Sohl-Dickstein, Roman Novak, Samuel S. Schoenholz et al.

There are currently two parameterizations used to derive fixed kernels corresponding to infinite width neural networks, the NTK (Neural Tangent Kernel) parameterization and the naive standard parameterization. However, the extrapolation of both of these parameterizations to infinite width is problematic. The standard parameterization leads to a divergent neural tangent kernel while the NTK parameterization fails to capture crucial aspects of finite width networks such as: the dependence of training dynamics on relative layer widths, the relative training dynamics of weights and biases, and overall learning rate scale. Here we propose an improved extrapolation of the standard parameterization that preserves all of these properties as width is taken to infinity and yields a well-defined neural tangent kernel. We show experimentally that the resulting kernels typically achieve similar accuracy to those resulting from an NTK parameterization, but with better correspondence to the parameterization of typical finite width networks. Additionally, with careful tuning of width parameters, the improved standard parameterization kernels can outperform those stemming from an NTK parameterization. We release code implementing this improved standard parameterization as part of the Neural Tangents library at https://github.com/google/neural-tangents.

Roman Novak, Lechao Xiao, Jiri Hron et al.

Neural Tangents is a library designed to enable research into infinite-width neural networks. It provides a high-level API for specifying complex and hierarchical neural network architectures. These networks can then be trained and evaluated either at finite-width as usual or in their infinite-width limit. Infinite-width networks can be trained analytically using exact Bayesian inference or using gradient descent via the Neural Tangent Kernel. Additionally, Neural Tangents provides tools to study gradient descent training dynamics of wide but finite networks in either function space or weight space. The entire library runs out-of-the-box on CPU, GPU, or TPU. All computations can be automatically distributed over multiple accelerators with near-linear scaling in the number of devices. Neural Tangents is available at www.github.com/google/neural-tangents. We also provide an accompanying interactive Colab notebook.

Jaehoon Lee, Suhwan Park, Tae Yoon Lim et al.

The financial domain involves a variety of important time-series problems. Recently, time-series analysis methods that jointly leverage textual and numerical information have gained increasing attention. Accordingly, numerous efforts have been made to construct text-paired time-series datasets in the financial domain. However, financial markets are characterized by complex interdependencies, in which a company's stock price is influenced not only by company-specific events but also by events in other companies and broader macroeconomic factors. Existing approaches that pair text with financial time-series data based on simple keyword matching often fail to capture such complex relationships. To address this limitation, we propose a semantic-based and multi-level pairing framework. Specifically, we extract company-specific context for the target company from SEC filings and apply an embedding-based matching mechanism to retrieve semantically relevant news articles based on this context. Furthermore, we classify news articles into four levels (macro-level, sector-level, related company-level, and target-company level) using large language models (LLMs), enabling multi-level pairing of news articles with the target company. Applying this framework to publicly-available news datasets, we construct \textbf{FinTexTS}, a new large-scale text-paired stock price dataset. Experimental results on \textbf{FinTexTS} demonstrate the effectiveness of our semantic-based and multi-level pairing strategy in stock price forecasting. In addition to publicly-available news underlying \textbf{FinTexTS}, we show that applying our method to proprietary yet carefully curated news sources leads to higher-quality paired data and improved stock price forecasting performance.

Brian Lester, Jaehoon Lee, Alex Alemi et al. · anthropic

In this paper, we explore the idea of training large language models (LLMs) over highly compressed text. While standard subword tokenizers compress text by a small factor, neural text compressors can achieve much higher rates of compression. If it were possible to train LLMs directly over neurally compressed text, this would confer advantages in training and serving efficiency, as well as easier handling of long text spans. The main obstacle to this goal is that strong compression tends to produce opaque outputs that are not well-suited for learning. In particular, we find that text naïvely compressed via Arithmetic Coding is not readily learnable by LLMs. To overcome this, we propose Equal-Info Windows, a novel compression technique whereby text is segmented into blocks that each compress to the same bit length. Using this method, we demonstrate effective learning over neurally compressed text that improves with scale, and outperforms byte-level baselines by a wide margin on perplexity and inference speed benchmarks. While our method delivers worse perplexity than subword tokenizers for models trained with the same parameter count, it has the benefit of shorter sequence lengths. Shorter sequence lengths require fewer autoregressive generation steps, and reduce latency. Finally, we provide extensive analysis of the properties that contribute to learnability, and offer concrete suggestions for how to further improve the performance of high-compression tokenizers.

Jaihyun Lew, Soohyuk Jang, Jaehoon Lee et al.

Transformers, a groundbreaking architecture proposed for Natural Language Processing (NLP), have also achieved remarkable success in Computer Vision. A cornerstone of their success lies in the attention mechanism, which models relationships among tokens. While the tokenization process in NLP inherently ensures that a single token does not contain multiple semantics, the tokenization of Vision Transformer (ViT) utilizes tokens from uniformly partitioned square image patches, which may result in an arbitrary mixing of visual concepts in a token. In this work, we propose to substitute the grid-based tokenization in ViT with superpixel tokenization, which employs superpixels to generate a token that encapsulates a sole visual concept. Unfortunately, the diverse shapes, sizes, and locations of superpixels make integrating superpixels into ViT tokenization rather challenging. Our tokenization pipeline, comprised of pre-aggregate extraction and superpixel-aware aggregation, overcomes the challenges that arise in superpixel tokenization. Extensive experiments demonstrate that our approach, which exhibits strong compatibility with existing frameworks, enhances the accuracy and robustness of ViT on various downstream tasks.

Bodam Kim, Hiskias Dingeto, Taeyoun Kwon et al.

As large language models become increasingly integrated into daily life, audio has emerged as a key interface for human-AI interaction. However, this convenience also introduces new vulnerabilities, making audio a potential attack surface for adversaries. Our research introduces WhisperInject, a two-stage adversarial audio attack framework that can manipulate state-of-the-art audio language models to generate harmful content. Our method uses imperceptible perturbations in audio inputs that remain benign to human listeners. The first stage uses a novel reward-based optimization method, Reinforcement Learning with Projected Gradient Descent (RL-PGD), to guide the target model to circumvent its own safety protocols and generate harmful native responses. This native harmful response then serves as the target for Stage 2, Payload Injection, where we use Projected Gradient Descent (PGD) to optimize subtle perturbations that are embedded into benign audio carriers, such as weather queries or greeting messages. Validated under the rigorous StrongREJECT, LlamaGuard, as well as Human Evaluation safety evaluation framework, our experiments demonstrate a success rate exceeding 86% across Qwen2.5-Omni-3B, Qwen2.5-Omni-7B, and Phi-4-Multimodal. Our work demonstrates a new class of practical, audio-native threats, moving beyond theoretical exploits to reveal a feasible and covert method for manipulating AI behavior.

Hoyoung Lee, Wonbin Ahn, Suhwan Park et al.

Thematic investing, which aims to construct portfolios aligned with structural trends, remains a challenging endeavor due to overlapping sector boundaries and evolving market dynamics. A promising direction is to build semantic representations of investment themes from textual data. However, despite their power, general-purpose LLM embedding models are not well-suited to capture the nuanced characteristics of financial assets, since the semantic representation of investment assets may differ fundamentally from that of general financial text. To address this, we introduce THEME, a framework that fine-tunes embeddings using hierarchical contrastive learning. THEME aligns themes and their constituent stocks using their hierarchical relationship, and subsequently refines these embeddings by incorporating stock returns. This process yields representations effective for retrieving thematically aligned assets with strong return potential. Empirical results demonstrate that THEME excels in two key areas. For thematic asset retrieval, it significantly outperforms leading large language models. Furthermore, its constructed portfolios demonstrate compelling performance. By jointly modeling thematic relationships from text and market dynamics from returns, THEME generates stock embeddings specifically tailored for a wide range of practical investment applications.

Dayoon Ko, Jinyoung Kim, Sohyeon Kim et al.

Dense retrievers encode texts into embeddings to efficiently retrieve relevant documents from large databases in response to user queries. However, real-world corpora continually evolve, leading to a shift from the original training distribution of the retriever. Without timely updates or retraining, indexing newly emerging documents can degrade retrieval performance for future queries. Thus, identifying when a dense retriever requires an update is critical for maintaining robust retrieval systems. In this paper, we propose a novel task of predicting whether a corpus is out-of-distribution (OOD) relative to a dense retriever before indexing. Addressing this task allows us to proactively manage retriever updates, preventing potential retrieval failures. We introduce GradNormIR, an unsupervised approach that leverages gradient norms to detect OOD corpora effectively. Experiments on the BEIR benchmark demonstrate that GradNormIR enables timely updates of dense retrievers in evolving document collections, significantly enhancing retrieval robustness and efficiency.

Sungik Choi, Sungwoo Park, Jaehoon Lee et al.

Dramatic advances in the quality of the latent diffusion models (LDMs) also led to the malicious use of AI-generated images. While current AI-generated image detection methods assume the availability of real/AI-generated images for training, this is practically limited given the vast expressibility of LDMs. This motivates the training-free detection setup where no related data are available in advance. The existing LDM-generated image detection method assumes that images generated by LDM are easier to reconstruct using an autoencoder than real images. However, we observe that this reconstruction distance is overfitted to background information, leading the current method to underperform in detecting images with simple backgrounds. To address this, we propose a novel method called HFI. Specifically, by viewing the autoencoder of LDM as a downsampling-upsampling kernel, HFI measures the extent of aliasing, a distortion of high-frequency information that appears in the reconstructed image. HFI is training-free, efficient, and consistently outperforms other training-free methods in detecting challenging images generated by various generative models. We also show that HFI can successfully detect the images generated from the specified LDM as a means of implicit watermarking. HFI outperforms the best baseline method while achieving magnitudes of

Jun Seo, Hyeokjun Choe, Seohui Bae et al.

Time series forecasting is a critical task for artificial intelligence with numerous real-world applications. Traditional approaches primarily rely on historical time series data to predict the future values. However, in practical scenarios, this is often insufficient for accurate predictions due to the limited information available. To address this challenge, multimodal time series forecasting methods which incorporate additional data modalities, mainly text data, alongside time series data have been explored. In this work, we introduce the Adaptive Information Routing (AIR) framework, a novel approach for multimodal time series forecasting. Unlike existing methods that treat text data on par with time series data as interchangeable auxiliary features for forecasting, AIR leverages text information to dynamically guide the time series model by controlling how and to what extent multivariate time series information should be combined. We also present a text-refinement pipeline that employs a large language model to convert raw text data into a form suitable for multimodal forecasting, and we introduce a benchmark that facilitates multimodal forecasting experiments based on this pipeline. Experiment results with the real world market data such as crude oil price and exchange rates demonstrate that AIR effectively modulates the behavior of the time series model using textual inputs, significantly enhancing forecasting accuracy in various time series forecasting tasks.

Seongbo Jang, Minjin Jeon, Jaehoon Lee et al.

While research on dialogue response generation has primarily focused on generating coherent responses conditioning on textual context, the critical question of when to respond grounded on the temporal context remains underexplored. To bridge this gap, we propose a novel task called timely dialogue response generation and introduce the TimelyChat benchmark, which evaluates the capabilities of language models to predict appropriate time intervals and generate time-conditioned responses. Additionally, we construct a large-scale training dataset by leveraging unlabeled event knowledge from a temporal commonsense knowledge graph and employing a large language model (LLM) to synthesize 55K event-driven dialogues. We then train Timer, a dialogue agent designed to proactively predict time intervals and generate timely responses that align with those intervals. Experimental results show that Timer outperforms prompting-based LLMs and other fine-tuned baselines in both turn-level and dialogue-level evaluations. We publicly release our data, model, and code.

Jaehoon Lee, Jihyeon Hyeong, Jinsung Jeon et al.

Tabular data synthesis has received wide attention in the literature. This is because available data is often limited, incomplete, or cannot be obtained easily, and data privacy is becoming increasingly important. In this work, we present a generalized GAN framework for tabular synthesis, which combines the adversarial training of GANs and the negative log-density regularization of invertible neural networks. The proposed framework can be used for two distinctive objectives. First, we can further improve the synthesis quality, by decreasing the negative log-density of real records in the process of adversarial training. On the other hand, by increasing the negative log-density of real records, realistic fake records can be synthesized in a way that they are not too much close to real records and reduce the chance of potential information leakage. We conduct experiments with real-world datasets for classification, regression, and privacy attacks. In general, the proposed method demonstrates the best synthesis quality (in terms of task-oriented evaluation metrics, e.g., F1) when decreasing the negative log-density during the adversarial training. If increasing the negative log-density, our experimental results show that the distance between real and fake records increases, enhancing robustness against privacy attacks.

Timothy Nguyen, Roman Novak, Lechao Xiao et al.

The effectiveness of machine learning algorithms arises from being able to extract useful features from large amounts of data. As model and dataset sizes increase, dataset distillation methods that compress large datasets into significantly smaller yet highly performant ones will become valuable in terms of training efficiency and useful feature extraction. To that end, we apply a novel distributed kernel based meta-learning framework to achieve state-of-the-art results for dataset distillation using infinitely wide convolutional neural networks. For instance, using only 10 datapoints (0.02% of original dataset), we obtain over 65% test accuracy on CIFAR-10 image classification task, a dramatic improvement over the previous best test accuracy of 40%. Our state-of-the-art results extend across many other settings for MNIST, Fashion-MNIST, CIFAR-10, CIFAR-100, and SVHN. Furthermore, we perform some preliminary analyses of our distilled datasets to shed light on how they differ from naturally occurring data.

Jayoung Kim, Jinsung Jeon, Jaehoon Lee et al.

Synthesizing tabular data is attracting much attention these days for various purposes. With sophisticate synthetic data, for instance, one can augment its training data. For the past couple of years, tabular data synthesis techniques have been greatly improved. Recent work made progress to address many problems in synthesizing tabular data, such as the imbalanced distribution and multimodality problems. However, the data utility of state-of-the-art methods is not satisfactory yet. In this work, we significantly improve the utility by designing our generator and discriminator based on neural ordinary differential equations (NODEs). After showing that NODEs have theoretically preferred characteristics for generating tabular data, we introduce our designs. The NODE-based discriminator performs a hidden vector evolution trajectory-based classification rather than classifying with a hidden vector at the last layer only. Our generator also adopts an ODE layer at the very beginning of its architecture to transform its initial input vector (i.e., the concatenation of a noisy vector and a condition vector in our case) onto another latent vector space suitable for the generation process. We conduct experiments with 13 datasets, including but not limited to insurance fraud detection, online news article prediction, and so on, and our presented method outperforms other state-of-the-art tabular data synthesis methods in many cases of our classification, regression, and clustering experiments.

Yasaman Bahri, Ethan Dyer, Jared Kaplan et al.

The population loss of trained deep neural networks often follows precise power-law scaling relations with either the size of the training dataset or the number of parameters in the network. We propose a theory that explains the origins of and connects these scaling laws. We identify variance-limited and resolution-limited scaling behavior for both dataset and model size, for a total of four scaling regimes. The variance-limited scaling follows simply from the existence of a well-behaved infinite data or infinite width limit, while the resolution-limited regime can be explained by positing that models are effectively resolving a smooth data manifold. In the large width limit, this can be equivalently obtained from the spectrum of certain kernels, and we present evidence that large width and large dataset resolution-limited scaling exponents are related by a duality. We exhibit all four scaling regimes in the controlled setting of large random feature and pretrained models and test the predictions empirically on a range of standard architectures and datasets. We also observe several empirical relationships between datasets and scaling exponents under modifications of task and architecture aspect ratio. Our work provides a taxonomy for classifying different scaling regimes, underscores that there can be different mechanisms driving improvements in loss, and lends insight into the microscopic origins of and relationships between scaling exponents.

Daniel S. Park, Jaehoon Lee, Daiyi Peng et al.

The predictions of wide Bayesian neural networks are described by a Gaussian process, known as the Neural Network Gaussian Process (NNGP). Analytic forms for NNGP kernels are known for many models, but computing the exact kernel for convolutional architectures is prohibitively expensive. One can obtain effective approximations of these kernels through Monte-Carlo estimation using finite networks at initialization. Monte-Carlo NNGP inference is orders-of-magnitude cheaper in FLOPs compared to gradient descent training when the dataset size is small. Since NNGP inference provides a cheap measure of performance of a network architecture, we investigate its potential as a signal for neural architecture search (NAS). We compute the NNGP performance of approximately 423k networks in the NAS-bench 101 dataset on CIFAR-10 and compare its utility against conventional performance measures obtained by shortened gradient-based training. We carry out a similar analysis on 10k randomly sampled networks in the mobile neural architecture search (MNAS) space for ImageNet. We discover comparative advantages of NNGP-based metrics, and discuss potential applications. In particular, we propose that NNGP performance is an inexpensive signal independent of metrics obtained from training that can either be used for reducing big search spaces, or improving training-based performance measures.

Timothy Nguyen, Zhourong Chen, Jaehoon Lee

One of the most fundamental aspects of any machine learning algorithm is the training data used by the algorithm. We introduce the novel concept of $ε$-approximation of datasets, obtaining datasets which are much smaller than or are significant corruptions of the original training data while maintaining similar model performance. We introduce a meta-learning algorithm called Kernel Inducing Points (KIP) for obtaining such remarkable datasets, inspired by the recent developments in the correspondence between infinitely-wide neural networks and kernel ridge-regression (KRR). For KRR tasks, we demonstrate that KIP can compress datasets by one or two orders of magnitude, significantly improving previous dataset distillation and subset selection methods while obtaining state of the art results for MNIST and CIFAR-10 classification. Furthermore, our KIP-learned datasets are transferable to the training of finite-width neural networks even beyond the lazy-training regime, which leads to state of the art results for neural network dataset distillation with potential applications to privacy-preservation.

Ben Adlam, Jaehoon Lee, Lechao Xiao et al.

Modern deep learning models have achieved great success in predictive accuracy for many data modalities. However, their application to many real-world tasks is restricted by poor uncertainty estimates, such as overconfidence on out-of-distribution (OOD) data and ungraceful failing under distributional shift. Previous benchmarks have found that ensembles of neural networks (NNs) are typically the best calibrated models on OOD data. Inspired by this, we leverage recent theoretical advances that characterize the function-space prior of an ensemble of infinitely-wide NNs as a Gaussian process, termed the neural network Gaussian process (NNGP). We use the NNGP with a softmax link function to build a probabilistic model for multi-class classification and marginalize over the latent Gaussian outputs to sample from the posterior. This gives us a better understanding of the implicit prior NNs place on function space and allows a direct comparison of the calibration of the NNGP and its finite-width analogue. We also examine the calibration of previous approaches to classification with the NNGP, which treat classification problems as regression to the one-hot labels. In this case the Bayesian posterior is exact, and we compare several heuristics to generate a categorical distribution over classes. We find these methods are well calibrated under distributional shift. Finally, we consider an infinite-width final layer in conjunction with a pre-trained embedding. This replicates the important practical use case of transfer learning and allows scaling to significantly larger datasets. As well as achieving competitive predictive accuracy, this approach is better calibrated than its finite width analogue.

Jaehoon Lee, Samuel S. Schoenholz, Jeffrey Pennington et al.

We perform a careful, thorough, and large scale empirical study of the correspondence between wide neural networks and kernel methods. By doing so, we resolve a variety of open questions related to the study of infinitely wide neural networks. Our experimental results include: kernel methods outperform fully-connected finite-width networks, but underperform convolutional finite width networks; neural network Gaussian process (NNGP) kernels frequently outperform neural tangent (NT) kernels; centered and ensembled finite networks have reduced posterior variance and behave more similarly to infinite networks; weight decay and the use of a large learning rate break the correspondence between finite and infinite networks; the NTK parameterization outperforms the standard parameterization for finite width networks; diagonal regularization of kernels acts similarly to early stopping; floating point precision limits kernel performance beyond a critical dataset size; regularized ZCA whitening improves accuracy; finite network performance depends non-monotonically on width in ways not captured by double descent phenomena; equivariance of CNNs is only beneficial for narrow networks far from the kernel regime. Our experiments additionally motivate an improved layer-wise scaling for weight decay which improves generalization in finite-width networks. Finally, we develop improved best practices for using NNGP and NT kernels for prediction, including a novel ensembling technique. Using these best practices we achieve state-of-the-art results on CIFAR-10 classification for kernels corresponding to each architecture class we consider.

Byeongjo Kim, Chanran Kim, Jaehoon Lee et al.

In general, sufficient data is essential for the better performance and generalization of deep-learning models. However, lots of limitations(cost, resources, etc.) of data collection leads to lack of enough data in most of the areas. In addition, various domains of each data sources and licenses also lead to difficulties in collection of sufficient data. This situation makes us hard to utilize not only the pre-trained model, but also the external knowledge. Therefore, it is important to leverage small dataset effectively for achieving the better performance. We applied some techniques in three aspects: data, loss function, and prediction to enable training from scratch with less data. With these methods, we obtain high accuracy by leveraging ImageNet data which consist of only 50 images per class. Furthermore, our model is ranked 4th in Visual Inductive Printers for Data-Effective Computer Vision Challenge.

Dario Fuoli, Zhiwu Huang, Martin Danelljan et al.

This paper reviews the NTIRE 2020 challenge on video quality mapping (VQM), which addresses the issues of quality mapping from source video domain to target video domain. The challenge includes both a supervised track (track 1) and a weakly-supervised track (track 2) for two benchmark datasets. In particular, track 1 offers a new Internet video benchmark, requiring algorithms to learn the map from more compressed videos to less compressed videos in a supervised training manner. In track 2, algorithms are required to learn the quality mapping from one device to another when their quality varies substantially and weakly-aligned video pairs are available. For track 1, in total 7 teams competed in the final test phase, demonstrating novel and effective solutions to the problem. For track 2, some existing methods are evaluated, showing promising solutions to the weakly-supervised video quality mapping problem.