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 [1], 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.

We investigate using large language models (LLMs) to generate text-only training data for improving multilingual and code-switching automatic speech recognition (ASR) through a text injection method. In a multilingual setup or a low-resource scenario such as code-switching, we propose to generate text data using the state-of-the-art PaLM 2. To better match the generated text data with specific tasks, we use prompt tuning to adapt PaLM 2 to generate domain-relevant multilingual or code-switched text data for text injection. We can achieve significant improvements in Word Error Rate (WER) in both multilingual and code-switching scenarios. The multilingual experiment shows a $6.2 \%$ relative WER reduction on average, i.e., from $11.25 \%$ to $10.55 \%$, compared to a baseline without text injection. The improvement is up to $23.1 \%$ improvement for certain languages. While in the code-switching scenario, we use English-only prompts to generate Mandarin-English code-switching text and achieve a 3.6% relative WER reduction for a code-switching test set, as well as WER reductions in both English and Mandarin monolingual scenarios, $5.3 \%$ and $8.5 \%$ relative, respectively. Our findings demonstrate that leveraging LLMs for text generation and then injection benefits multilingual or code-switching ASR tasks.

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.

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.

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.

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.

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.

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.

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 L 2 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 tradeoff on the original task; ii) incorporating priors helps model performance in scarce data settings.

Website privacy policies are often long and difficult to understand. While research shows that Internet users care about their privacy, they do not have the time to understand the policies of every website they visit, and most users hardly ever read privacy policies. Some recent efforts have aimed to use a combination of crowdsourcing, machine learning, and natural language processing to interpret privacy policies at scale, thus producing annotations for use in interfaces that inform Internet users of salient policy details. However, little attention has been devoted to studying the accuracy of crowdsourced privacy policy annotations, how crowdworker productivity can be enhanced for such a task, and the levels of granularity that are feasible for automatic analysis of privacy policies. In this article, we present a trajectory of work addressing each of these topics. We include analyses of crowdworker performance, evaluation of a method to make a privacy-policy oriented task easier for crowdworkers, a coarse-grained approach to labeling segments of policy text with descriptive themes, and a fine-grained approach to identifying user choices described in policy text. Together, the results from these efforts show the effectiveness of using automated and semi-automated methods for extracting from privacy policies the data practice details that are salient to Internet users’ interests.

Frederick Liu, Han Lu, Graham Neubig. Proceedings of the 2018 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, Volume 1 (Long Papers). 2018.

Previous work has modeled the compositionality of words by creating characterlevel 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.

Shomir Wilson, Florian Schaub, Aswarth Abhilash Dara, Frederick Liu, Sushain Cherivirala, Pedro Giovanni Leon, Mads Schaarup Andersen, Sebastian Zimmeck, Kanthashree Mysore Sathyendra, N. Cameron Russell, Thomas B. Norton, Eduard Hovy, Joel Reidenberg, Norman Sadeh. Proceedings of the 54th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2016.

Website privacy policies are often long and difficult to understand. While research shows that Internet users care about their privacy, they do not have time to understand the policies of every website they visit, and most users hardly ever read privacy policies. Several recent efforts aim to crowdsource the interpretation of privacy policies and use the resulting annotations to build more effective user interfaces that provide users with salient policy summaries. However, very little attention has been devoted to studying the accuracy and scalability of crowdsourced privacy policy annotations, the types of questions crowdworkers can effectively answer, and the ways in which their productivity can be enhanced. Prior research indicates that most Internet users often have great difficulty understanding privacy policies, suggesting limits to the effectiveness of crowdsourcing approaches. In this paper, we assess the viability of crowdsourcing privacy policy annotations. Our results suggest that, if carefully deployed, crowdsourcing can indeed result in the generation of non-trivial annotations and can also help identify elements of ambiguity in policies. We further introduce and evaluate a method to improve the annotation process by predicting and highlighting paragraphs relevant to specific data practices.

Learning a second language is becoming a more popular trend around the world. But the act of learning another language in a place removed from native speakers is difficult as there is often no one to correct mistakes nor examples to imitate. With the idea of crowd sourcing, we would like to propose an efficient way to learn a second language better.