Frederick Liu

Ekin Akyürek, Tolga Bolukbasi, Frederick Liu et al. · microsoft-research, mit

Language models (LMs) have been shown to memorize a great deal of factual knowledge contained in their training data. But when an LM generates an assertion, it is often difficult to determine where it learned this information and whether it is true. In this paper, we propose the problem of fact tracing: identifying which training examples taught an LM to generate a particular factual assertion. Prior work on training data attribution (TDA) may offer effective tools for identifying such examples, known as "proponents". We present the first quantitative benchmark to evaluate this. We compare two popular families of TDA methods -- gradient-based and embedding-based -- and find that much headroom remains. For example, both methods have lower proponent-retrieval precision than an information retrieval baseline (BM25) that does not have access to the LM at all. We identify key challenges that may be necessary for further improvement such as overcoming the problem of gradient saturation, and also show how several nuanced implementation details of existing neural TDA methods can significantly improve overall fact tracing performance.

Valerii Likhosherstov, Krzysztof Choromanski, Avinava Dubey et al. · cambridge

We introduce chefs' random tables (CRTs), a new class of non-trigonometric random features (RFs) to approximate Gaussian and softmax kernels. CRTs are an alternative to standard random kitchen sink (RKS) methods, which inherently rely on the trigonometric maps. We present variants of CRTs where RFs are positive, a key requirement for applications in recent low-rank Transformers. Further variance reduction is possible by leveraging statistics which are simple to compute. One instantiation of CRTs, the optimal positive random features (OPRFs), is to our knowledge the first RF method for unbiased softmax kernel estimation with positive and bounded RFs, resulting in exponentially small tails and much lower variance than its counterparts. As we show, orthogonal random features applied in OPRFs provide additional variance reduction for any dimensionality $d$ (not only asymptotically for sufficiently large $d$, as for RKS). We test CRTs on many tasks ranging from non-parametric classification to training Transformers for text, speech and image data, obtaining new state-of-the-art results for low-rank text Transformers, while providing linear space and time complexity.

Ziqi Wang, Yuexin Wu, Frederick Liu et al.

Knowledge distillation is one of the primary methods of transferring knowledge from large to small models. However, it requires massive task-specific data, which may not be plausible in many real-world applications. Data augmentation methods such as representation interpolation, token replacement, or augmentation with models are applied to tackle this problem. However, these data augmentation methods either potentially cause shifts in decision boundaries (representation interpolation), are not expressive enough (token replacement), or introduce too much computational overhead (augmentation with models). To this end, we propose AugPro (Augmentation with Projection), an effective and efficient data augmentation method for distillation. Our method builds on top of representation interpolation augmentation methods to maintain the diversity of expressions and converts the augmented data to tokens to avoid shifting decision boundaries. It uses simple operations that come with little computational overhead. The results on multiple GLUE tasks show that our methods can improve distillation performance by a large margin at a low time cost. Codes are available at https://github.com/google-research/google-research/tree/master/augpro.

Valerii Likhosherstov, Krzysztof Choromanski, Avinava Dubey et al. · cambridge

The problem of efficient approximation of a linear operator induced by the Gaussian or softmax kernel is often addressed using random features (RFs) which yield an unbiased approximation of the operator's result. Such operators emerge in important applications ranging from kernel methods to efficient Transformers. We propose parameterized, positive, non-trigonometric RFs which approximate Gaussian and softmax-kernels. In contrast to traditional RF approximations, parameters of these new methods can be optimized to reduce the variance of the approximation, and the optimum can be expressed in closed form. We show that our methods lead to variance reduction in practice ($e^{10}$-times smaller variance and beyond) and outperform previous methods in a kernel regression task. Using our proposed mechanism, we also present FAVOR#, a method for self-attention approximation in Transformers. We show that FAVOR# outperforms other random feature methods in speech modelling and natural language processing.

Yaqing Wang, Jialin Wu, Tanmaya Dabral et al.

Large language models (LLMs) and vision language models (VLMs) demonstrate excellent performance on a wide range of tasks by scaling up parameter counts from O(10^9) to O(10^{12}) levels and further beyond. These large scales make it impossible to adapt and deploy fully specialized models given a task of interest. Parameter-efficient fine-tuning (PEFT) emerges as a promising direction to tackle the adaptation and serving challenges for such large models. We categorize PEFT techniques into two types: intrusive and non-intrusive. Intrusive PEFT techniques directly change a model's internal architecture. Though more flexible, they introduce significant complexities for training and serving. Non-intrusive PEFT techniques leave the internal architecture unchanged and only adapt model-external parameters, such as embeddings for input. In this work, we describe AdaLink as a non-intrusive PEFT technique that achieves competitive performance compared to SoTA intrusive PEFT (LoRA) and full model fine-tuning (FT) on various tasks. We evaluate using both text-only and multimodal tasks, with experiments that account for both parameter-count scaling and training regime (with and without instruction tuning).

Sid Mittal, Vineet Gupta, Frederick Liu et al.

Foundation models, i.e. large neural networks pre-trained on large text corpora, have revolutionized NLP. They can be instructed directly (e.g. (arXiv:2005.14165)) - this is called hard prompting - and they can be tuned using very little data (e.g. (arXiv:2104.08691)) - this technique is called soft prompting. We seek to leverage their capabilities to detect policy violations. Our contributions are: We identify a hard prompt that adapts chain-of-thought prompting to policy violation tasks. This prompt produces policy violation classifications, along with extractive explanations that justify the classification. We compose the hard-prompts with soft prompt tuning to produce a classifier that attains high accuracy with very little supervision; the same classifier also produces explanations. Though the supervision only acts on the classifications, we find that the modified explanations remain consistent with the (tuned) model's response. Along the way, we identify several unintuitive aspects of foundation models. For instance, adding an example from a specific class can actually reduce predictions of that class, and separately, the effects of tokenization on scoring etc. Based on our technical results, we identify a simple workflow for product teams to quickly develop effective policy violation detectors.

Chi Zhang, Lijuan Liu, Xiaoxue Zang et al.

Convolutional Neural Networks (CNN) have dominated the field of detection ever since the success of AlexNet in ImageNet classification [12]. With the sweeping reform of Transformers [27] in natural language processing, Carion et al. [2] introduce the Transformer-based detection method, i.e., DETR. However, due to the quadratic complexity in the self-attention mechanism in the Transformer, DETR is never able to incorporate multi-scale features as performed in existing CNN-based detectors, leading to inferior results in small object detection. To mitigate this issue and further improve performance of DETR, in this work, we investigate different methods to incorporate multi-scale features and find that a Bi-directional Feature Pyramid (BiFPN) works best with DETR in further raising the detection precision. With this discovery, we propose DETR++, a new architecture that improves detection results by 1.9% AP on MS COCO 2017, 11.5% AP on RICO icon detection, and 9.1% AP on RICO layout extraction over existing baselines.

Maximilian Mozes, Tolga Bolukbasi, Ann Yuan et al.

Pretrained large language models (LLMs) are able to solve a wide variety of tasks through transfer learning. Various explainability methods have been developed to investigate their decision making process. TracIn (Pruthi et al., 2020) is one such gradient-based method which explains model inferences based on the influence of training examples. In this paper, we explore the use of TracIn to improve model performance in the parameter-efficient tuning (PET) setting. We develop conversational safety classifiers via the prompt-tuning PET method and show how the unique characteristics of the PET regime enable TracIn to identify the cause for certain misclassifications by LLMs. We develop a new methodology for using gradient-based explainability techniques to improve model performance, G-BAIR: gradient-based automated iterative recovery. We show that G-BAIR can recover LLM performance on benchmarks after manually corrupting training labels. This suggests that influence methods like TracIn can be used to automatically perform data cleaning, and introduces the potential for interactive debugging and relabeling for PET-based transfer learning methods.

Terry Koo, Frederick Liu, Luheng He

Language models (LMs) are often expected to generate strings in some formal language; for example, structured data, API calls, or code snippets. Although LMs can be tuned to improve their adherence to formal syntax, this does not guarantee conformance, especially with smaller LMs suitable for large-scale deployment. In addition, tuning requires significant resources, making it impractical for uncommon or task-specific formats. To prevent downstream parsing errors we would ideally constrain the LM to only produce valid output, but this is severely complicated by tokenization, which is typically both ambiguous and misaligned with the formal grammar. We solve these issues through the application of automata theory, deriving an efficient closed-form solution for the regular languages, a broad class of formal languages with many practical applications, including API calls or schema-guided JSON and YAML. We also discuss pragmatic extensions for coping with the issue of high branching factor, and extend our techniques to deterministic context-free languages, which similarly admit an efficient closed-form solution. Previous work on this topic (Willard and Louf, 2023) layers bespoke solutions onto automata, leading to problems with speed, correctness, and extensibility. Instead, we reformulate the entire task in terms of automata so we can leverage well-studied and well-optimized algorithms. Our system compiles constraints ~7,000x faster, is provably correct, and can be extended in a modular fashion.

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.

Gheorghe Comanici, Eric Bieber, Mike Schaekermann et al. · amazon-science, baidu

In this report, we introduce the Gemini 2.X model family: Gemini 2.5 Pro and Gemini 2.5 Flash, as well as our earlier Gemini 2.0 Flash and Flash-Lite models. Gemini 2.5 Pro is our most capable model yet, achieving SoTA performance on frontier coding and reasoning benchmarks. In addition to its incredible coding and reasoning skills, Gemini 2.5 Pro is a thinking model that excels at multimodal understanding and it is now able to process up to 3 hours of video content. Its unique combination of long context, multimodal and reasoning capabilities can be combined to unlock new agentic workflows. Gemini 2.5 Flash provides excellent reasoning abilities at a fraction of the compute and latency requirements and Gemini 2.0 Flash and Flash-Lite provide high performance at low latency and cost. Taken together, the Gemini 2.X model generation spans the full Pareto frontier of model capability vs cost, allowing users to explore the boundaries of what is possible with complex agentic problem solving.

Gemma Team, Aishwarya Kamath, Johan Ferret et al. · deepmind, mit

We introduce Gemma 3, a multimodal addition to the Gemma family of lightweight open models, ranging in scale from 1 to 27 billion parameters. This version introduces vision understanding abilities, a wider coverage of languages and longer context - at least 128K tokens. We also change the architecture of the model to reduce the KV-cache memory that tends to explode with long context. This is achieved by increasing the ratio of local to global attention layers, and keeping the span on local attention short. The Gemma 3 models are trained with distillation and achieve superior performance to Gemma 2 for both pre-trained and instruction finetuned versions. In particular, our novel post-training recipe significantly improves the math, chat, instruction-following and multilingual abilities, making Gemma3-4B-IT competitive with Gemma2-27B-IT and Gemma3-27B-IT comparable to Gemini-1.5-Pro across benchmarks. We release all our models to the community.

Mahmoud G. Salem, Jiayu Ye, Chu-Cheng Lin et al.

Recent advances in Transformer-based Large Language Models have made great strides in natural language generation. However, to decode K tokens, an autoregressive model needs K sequential forward passes, which may be a performance bottleneck for large language models. Many non-autoregressive (NAR) research are aiming to address this sequentiality bottleneck, albeit many have focused on a dedicated architecture in supervised benchmarks. In this work, we studied unsupervised pretraining for non auto-regressive T5 models via unrolled denoising and shown its SoTA results in downstream generation tasks such as SQuAD question generation and XSum.

Isabel Leal, Krzysztof Choromanski, Deepali Jain et al.

We present Self-Adaptive Robust Attention for Robotics Transformers (SARA-RT): a new paradigm for addressing the emerging challenge of scaling up Robotics Transformers (RT) for on-robot deployment. SARA-RT relies on the new method of fine-tuning proposed by us, called up-training. It converts pre-trained or already fine-tuned Transformer-based robotic policies of quadratic time complexity (including massive billion-parameter vision-language-action models or VLAs), into their efficient linear-attention counterparts maintaining high quality. We demonstrate the effectiveness of SARA-RT by speeding up: (a) the class of recently introduced RT-2 models, the first VLA robotic policies pre-trained on internet-scale data, as well as (b) Point Cloud Transformer (PCT) robotic policies operating on large point clouds. We complement our results with the rigorous mathematical analysis providing deeper insight into the phenomenon of SARA.

Gemini Team, Rohan Anil, Sebastian Borgeaud et al.

This report introduces a new family of multimodal models, Gemini, that exhibit remarkable capabilities across image, audio, video, and text understanding. The Gemini family consists of Ultra, Pro, and Nano sizes, suitable for applications ranging from complex reasoning tasks to on-device memory-constrained use-cases. Evaluation on a broad range of benchmarks shows that our most-capable Gemini Ultra model advances the state of the art in 30 of 32 of these benchmarks - notably being the first model to achieve human-expert performance on the well-studied exam benchmark MMLU, and improving the state of the art in every one of the 20 multimodal benchmarks we examined. We believe that the new capabilities of the Gemini family in cross-modal reasoning and language understanding will enable a wide variety of use cases. We discuss our approach toward post-training and deploying Gemini models responsibly to users through services including Gemini, Gemini Advanced, Google AI Studio, and Cloud Vertex AI.

Rohan Anil, Andrew M. Dai, Orhan Firat et al.

We introduce PaLM 2, a new state-of-the-art language model that has better multilingual and reasoning capabilities and is more compute-efficient than its predecessor PaLM. PaLM 2 is a Transformer-based model trained using a mixture of objectives. Through extensive evaluations on English and multilingual language, and reasoning tasks, we demonstrate that PaLM 2 has significantly improved quality on downstream tasks across different model sizes, while simultaneously exhibiting faster and more efficient inference compared to PaLM. This improved efficiency enables broader deployment while also allowing the model to respond faster, for a more natural pace of interaction. PaLM 2 demonstrates robust reasoning capabilities exemplified by large improvements over PaLM on BIG-Bench and other reasoning tasks. PaLM 2 exhibits stable performance on a suite of responsible AI evaluations, and enables inference-time control over toxicity without additional overhead or impact on other capabilities. Overall, PaLM 2 achieves state-of-the-art performance across a diverse set of tasks and capabilities. When discussing the PaLM 2 family, it is important to distinguish between pre-trained models (of various sizes), fine-tuned variants of these models, and the user-facing products that use these models. In particular, user-facing products typically include additional pre- and post-processing steps. Additionally, the underlying models may evolve over time. Therefore, one should not expect the performance of user-facing products to exactly match the results reported in this report.

Chih-Kuan Yeh, Kuan-Yun Lee, Frederick Liu et al.

A popular explainable AI (XAI) approach to quantify feature importance of a given model is via Shapley values. These Shapley values arose in cooperative games, and hence a critical ingredient to compute these in an XAI context is a so-called value function, that computes the "value" of a subset of features, and which connects machine learning models to cooperative games. There are many possible choices for such value functions, which broadly fall into two categories: on-manifold and off-manifold value functions, which take an observational and an interventional viewpoint respectively. Both these classes however have their respective flaws, where on-manifold value functions violate key axiomatic properties and are computationally expensive, while off-manifold value functions pay less heed to the data manifold and evaluate the model on regions for which it wasn't trained. Thus, there is no consensus on which class of value functions to use. In this paper, we show that in addition to these existing issues, both classes of value functions are prone to adversarial manipulations on low density regions. We formalize the desiderata of value functions that respect both the model and the data manifold in a set of axioms and are robust to perturbation on off-manifold regions, and show that there exists a unique value function that satisfies these axioms, which we term the Joint Baseline value function, and the resulting Shapley value the Joint Baseline Shapley (JBshap), and validate the effectiveness of JBshap in experiments.

Chih-Kuan Yeh, Ankur Taly, Mukund Sundararajan et al.

The ability to identify influential training examples enables us to debug training data and explain model behavior. Existing techniques to do so are based on the flow of training data influence through the model parameters. For large models in NLP applications, it is often computationally infeasible to study this flow through all model parameters, therefore techniques usually pick the last layer of weights. However, we observe that since the activation connected to the last layer of weights contains "shared logic", the data influenced calculated via the last layer weights prone to a ``cancellation effect'', where the data influence of different examples have large magnitude that contradicts each other. The cancellation effect lowers the discriminative power of the influence score, and deleting influential examples according to this measure often does not change the model's behavior by much. To mitigate this, we propose a technique called TracIn-WE that modifies a method called TracIn to operate on the word embedding layer instead of the last layer, where the cancellation effect is less severe. One potential concern is that influence based on the word embedding layer may not encode sufficient high level information. However, we find that gradients (unlike embeddings) do not suffer from this, possibly because they chain through higher layers. We show that TracIn-WE significantly outperforms other data influence methods applied on the last layer significantly on the case deletion evaluation on three language classification tasks for different models. In addition, TracIn-WE can produce scores not just at the level of the overall training input, but also at the level of words within the training input, a further aid in debugging.

Frederick Liu, Terry Huang, Shihang Lyu et al.

Pre-trained encoder-decoder transformer architectures have become increasingly popular recently with the advent of T5 models. T5 has also become more favorable over other architectures like BERT due to the amount of data that it is pre-trained on, increased scale of model parameter sizes and easy applicability to a diverse set of tasks due to the generative nature of the model. While being able to generalize to a wide variety of tasks, it is not clear that encoder-decoder architectures are the most efficient for fine-tuning tasks that don't require auto-regressive decoding. In this work, we study fine-tuning pre-trained encoder-decoder models for tasks such as classification, multi-label classification, and structured prediction. We propose \textbf{EncT5}, a framework for these problems, and illustrate instantiations for these tasks. Our experiment results show that EncT5 has advantages over T5 such as efficiency and usability out performs BERT when evaluated on publicly available pre-trained checkpoints.

Srinadh Bhojanapalli, Ayan Chakrabarti, Andreas Veit et al.

Pairwise dot product-based attention allows Transformers to exchange information between tokens in an input-dependent way, and is key to their success across diverse applications in language and vision. However, a typical Transformer model computes such pairwise attention scores repeatedly for the same sequence, in multiple heads in multiple layers. We systematically analyze the empirical similarity of these scores across heads and layers and find them to be considerably redundant, especially adjacent layers showing high similarity. Motivated by these findings, we propose a novel architecture that reuses attention scores computed in one layer in multiple subsequent layers. Experiments on a number of standard benchmarks show that reusing attention delivers performance equivalent to or better than standard transformers, while reducing both compute and memory usage.

Jiefeng Chen, Frederick Liu, Besim Avci et al.

When a deep learning model is deployed in the wild, it can encounter test data drawn from distributions different from the training data distribution and suffer drop in performance. For safe deployment, it is essential to estimate the accuracy of the pre-trained model on the test data. However, the labels for the test inputs are usually not immediately available in practice, and obtaining them can be expensive. This observation leads to two challenging tasks: (1) unsupervised accuracy estimation, which aims to estimate the accuracy of a pre-trained classifier on a set of unlabeled test inputs; (2) error detection, which aims to identify mis-classified test inputs. In this paper, we propose a principled and practically effective framework that simultaneously addresses the two tasks. The proposed framework iteratively learns an ensemble of models to identify mis-classified data points and performs self-training to improve the ensemble with the identified points. Theoretical analysis demonstrates that our framework enjoys provable guarantees for both accuracy estimation and error detection under mild conditions readily satisfied by practical deep learning models. Along with the framework, we proposed and experimented with two instantiations and achieved state-of-the-art results on 59 tasks. For example, on iWildCam, one instantiation reduces the estimation error for unsupervised accuracy estimation by at least 70% and improves the F1 score for error detection by at least 4.7% compared to existing methods.

Frederick Liu, Amir Najmi, Mukund Sundararajan

There is a set of data augmentation techniques that ablate parts of the input at random. These include input dropout, cutout, and random erasing. We term these techniques ablated data augmentation. Though these techniques seems similar in spirit and have shown success in improving model performance in a variety of domains, we do not yet have a mathematical understanding of the differences between these techniques like we do for other regularization techniques like L1 or L2. First, we study a formal model of mean ablated data augmentation and inverted dropout for linear regression. We prove that ablated data augmentation is equivalent to optimizing the ordinary least squares objective along with a penalty that we call the Contribution Covariance Penalty and inverted dropout, a more common implementation than dropout in popular frameworks, is equivalent to optimizing the ordinary least squares objective along with Modified L2. For deep networks, we demonstrate an empirical version of the result if we replace contributions with attributions and coefficients with average gradients, i.e., the Contribution Covariance Penalty and Modified L2 Penalty drop with the increase of the corresponding ablated data augmentation across a variety of networks.

Garima Pruthi, Frederick Liu, Mukund Sundararajan et al.

We introduce a method called TracIn that computes the influence of a training example on a prediction made by the model. The idea is to trace how the loss on the test point changes during the training process whenever the training example of interest was utilized. We provide a scalable implementation of TracIn via: (a) a first-order gradient approximation to the exact computation, (b) saved checkpoints of standard training procedures, and (c) cherry-picking layers of a deep neural network. In contrast with previously proposed methods, TracIn is simple to implement; all it needs is the ability to work with gradients, checkpoints, and loss functions. The method is general. It applies to any machine learning model trained using stochastic gradient descent or a variant of it, agnostic of architecture, domain and task. We expect the method to be widely useful within processes that study and improve training data.

Frederick Liu, Besim Avci

Feature attribution methods, proposed recently, help users interpret the predictions of complex models. Our approach integrates feature attributions into the objective function to allow machine learning practitioners to incorporate priors in model building. To demonstrate the effectiveness our technique, we apply it to two tasks: (1) mitigating unintended bias in text classifiers by neutralizing identity terms; (2) improving classifier performance in a scarce data setting by forcing the model to focus on toxic terms. Our approach adds an L2 distance loss between feature attributions and task-specific prior values to the objective. Our experiments show that i) a classifier trained with our technique reduces undesired model biases without a trade off on the original task; ii) incorporating priors helps model performance in scarce data settings.

Vasyl Pihur, Aleksandra Korolova, Frederick Liu et al.

In this work, we propose a novel framework for privacy-preserving client-distributed machine learning. It is motivated by the desire to achieve differential privacy guarantees in the local model of privacy in a way that satisfies all systems constraints using asynchronous client-server communication and provides attractive model learning properties. We call it "Draw and Discard" because it relies on random sampling of models for load distribution (scalability), which also provides additional server-side privacy protections and improved model quality through averaging. We present the mechanics of client and server components of "Draw and Discard" and demonstrate how the framework can be applied to learning Generalized Linear models. We then analyze the privacy guarantees provided by our approach against several types of adversaries and showcase experimental results that provide evidence for the framework's viability in practical deployments.

Frederick Liu, Han Lu, Graham Neubig

Homographs, words with different meanings but the same surface form, have long caused difficulty for machine translation systems, as it is difficult to select the correct translation based on the context. However, with the advent of neural machine translation (NMT) systems, which can theoretically take into account global sentential context, one may hypothesize that this problem has been alleviated. In this paper, we first provide empirical evidence that existing NMT systems in fact still have significant problems in properly translating ambiguous words. We then proceed to describe methods, inspired by the word sense disambiguation literature, that model the context of the input word with context-aware word embeddings that help to differentiate the word sense be- fore feeding it into the encoder. Experiments on three language pairs demonstrate that such models improve the performance of NMT systems both in terms of BLEU score and in the accuracy of translating homographs.

Frederick Liu, Han Lu, Chieh Lo et al.

Previous work has modeled the compositionality of words by creating character-level models of meaning, reducing problems of sparsity for rare words. However, in many writing systems compositionality has an effect even on the character-level: the meaning of a character is derived by the sum of its parts. In this paper, we model this effect by creating embeddings for characters based on their visual characteristics, creating an image for the character and running it through a convolutional neural network to produce a visual character embedding. Experiments on a text classification task demonstrate that such model allows for better processing of instances with rare characters in languages such as Chinese, Japanese, and Korean. Additionally, qualitative analyses demonstrate that our proposed model learns to focus on the parts of characters that carry semantic content, resulting in embeddings that are coherent in visual space.