WebSuite: Systematically Evaluating Why Web Agents Fail

We describe WebSuite, the first diagnostic benchmark for generalist web agents, designed to systematically evaluate why agents fail. Advances in AI have led to the rise of numerous web agents that autonomously operate a browser to complete tasks. However, most existing benchmarks focus on strictly measuring whether an agent can or cannot complete a task, without giving insight on why. In this paper, we 1) develop a taxonomy of web actions to facilitate identifying common failure patterns, and 2) create an extensible benchmark suite to assess agents' performance on our taxonomized actions. This benchmark suite consists of both individual tasks, such as clicking a button, and end-to-end tasks, such as adding an item to a cart, and is designed such that any failure of a task can be attributed directly to a failure of a specific web action. We evaluate two popular generalist web agents, one text-based and one multimodal, and identify unique weaknesses for each agent. Because WebSuite can disaggregate task failures into specific action failures, this enables granular identification of which UX flows an individual agent has trouble with and immediately highlights promising avenues for improvement. These findings highlight the need for more focused benchmarking on where web agents go wrong to effectively improve agents beyond their weaker performance today.

Bayes3D: fast learning and inference in structured generative models of 3D objects and scenes

Robots cannot yet match humans' ability to rapidly learn the shapes of novel 3D objects and recognize them robustly despite clutter and occlusion. We present Bayes3D, an uncertainty-aware perception system for structured 3D scenes, that reports accurate posterior uncertainty over 3D object shape, pose, and scene composition in the presence of clutter and occlusion. Bayes3D delivers these capabilities via a novel hierarchical Bayesian model for 3D scenes and a GPU-accelerated coarse-to-fine sequential Monte Carlo algorithm. Quantitative experiments show that Bayes3D can learn 3D models of novel objects from just a handful of views, recognizing them more robustly and with orders of magnitude less training data than neural baselines, and tracking 3D objects faster than real time on a single GPU. We also demonstrate that Bayes3D learns complex 3D object models and accurately infers 3D scene composition when used on a Panda robot in a tabletop scenario.

PaLM 2 Technical Report

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.

Scaling Instruction-Finetuned Language Models

Finetuning language models on a collection of datasets phrased as instructions has been shown to improve model performance and generalization to unseen tasks. In this paper we explore instruction finetuning with a particular focus on (1) scaling the number of tasks, (2) scaling the model size, and (3) finetuning on chain-of-thought data. We find that instruction finetuning with the above aspects dramatically improves performance on a variety of model classes (PaLM, T5, U-PaLM), prompting setups (zero-shot, few-shot, CoT), and evaluation benchmarks (MMLU, BBH, TyDiQA, MGSM, open-ended generation). For instance, Flan-PaLM 540B instruction-finetuned on 1.8K tasks outperforms PALM 540B by a large margin (+9.4% on average). Flan-PaLM 540B achieves state-of-the-art performance on several benchmarks, such as 75.2% on five-shot MMLU. We also publicly release Flan-T5 checkpoints, which achieve strong few-shot performance even compared to much larger models, such as PaLM 62B. Overall, instruction finetuning is a general method for improving the performance and usability of pretrained language models.

Exceeding the limits of algorithmic self-calibration in super-resolution imaging

Fourier ptychographic microscopy is a computational imaging technique that provides quantitative phase information and high resolution over a large field-of-view. Although the technique presents numerous advantages over conventional microscopy, model mismatch due to unknown optical aberrations can significantly limit reconstruction quality. Many attempts to address this issue rely on embedding pupil recovery into the reconstruction algorithm. In this paper we demonstrate the limitations of a purely algorithmic approach and evaluate the merits of implementing a simple, dedicated calibration procedure. In simulations, we find that for a target sample reconstruction error, we can image without any aberration corrections up to a maximum aberration magnitude of $\lambda$/40. When we use algorithmic self-calibration, we can increase the aberration magnitude up to $\lambda$/10, and with our in situ speckle calibration technique, this working range is extended further to a maximum aberration magnitude of $\lambda$/3. Hence, one can trade-off complexity for accuracy by using a separate calibration process, which is particularly useful for larger aberrations.

AGenT Zero: Zero-shot Automatic Multiple-Choice Question Generation for Skill Assessments

Multiple-choice questions (MCQs) offer the most promising avenue for skill evaluation in the era of virtual education and job recruiting, where traditional performance-based alternatives such as projects and essays have become less viable, and grading resources are constrained. The automated generation of MCQs would allow assessment creation at scale. Recent advances in natural language processing have given rise to many complex question generation methods. However, the few methods that produce deployable results in specific domains require a large amount of domain-specific training data that can be very costly to acquire. Our work provides an initial foray into MCQ generation under high data-acquisition cost scenarios by strategically emphasizing paraphrasing the question context (compared to the task). In addition to maintaining semantic similarity between the question-answer pairs, our pipeline, which we call AGenT Zero, consists of only pre-trained models and requires no fine-tuning, minimizing data acquisition costs for question generation. AGenT Zero successfully outperforms other pre-trained methods in fluency and semantic similarity. Additionally, with some small changes, our assessment pipeline can be generalized to a broader question and answer space, including short answer or fill in the blank questions.