ScribeAgent: Towards Specialized Web Agents Using Production-Scale Workflow Data

Large Language Model (LLM) agents are rapidly improving to handle increasingly complex web-based tasks. Most of these agents rely on general-purpose, proprietary models like GPT-4 and focus on designing better prompts to improve their planning abilities. However, general-purpose LLMs are not specifically trained to understand specialized web contexts such as HTML, and they often struggle with long-horizon planning. We explore an alternative approach that fine-tunes open-source LLMs using production-scale workflow data collected from over 250 domains corresponding to 6 billion tokens. This simple yet effective approach shows substantial gains over prompting-based agents on existing benchmarks -- ScribeAgent achieves state-of-the-art direct generation performance on Mind2Web and improves the task success rate by 14.1% over the previous best text-only web agents on WebArena. We further perform detailed ablation studies on various fine-tuning design choices and provide insights into LLM selection, training recipes, context window optimization, and effect of dataset sizes.

Specialized Foundation Models Struggle to Beat Supervised Baselines

Following its success for vision and text, the "foundation model" (FM) paradigm -- pretraining large models on massive data, then fine-tuning on target tasks -- has rapidly expanded to domains in the sciences, engineering, healthcare, and beyond. Has this achieved what the original FMs accomplished, i.e. the supplanting of traditional supervised learning in their domains? To answer we look at three modalities -- genomics, satellite imaging, and time series -- with multiple recent FMs and compare them to a standard supervised learning workflow: model development, hyperparameter tuning, and training, all using only data from the target task. Across these three specialized domains, we find that it is consistently possible to train simple supervised models -- no more complicated than a lightly modified wide ResNet or UNet -- that match or even outperform the latest foundation models. Our work demonstrates that the benefits of large-scale pretraining have yet to be realized in many specialized areas, reinforces the need to compare new FMs to strong, well-tuned baselines, and introduces two new, easy-to-use, open-source, and automated workflows for doing so.

Understanding Optimization in Deep Learning with Central Flows

Optimization in deep learning remains poorly understood, even in the simple setting of deterministic (i.e. full-batch) training. A key difficulty is that much of an optimizer's behavior is implicitly determined by complex oscillatory dynamics, referred to as the "edge of stability." The main contribution of this paper is to show that an optimizer's implicit behavior can be explicitly captured by a "central flow:" a differential equation which models the time-averaged optimization trajectory. We show that these flows can empirically predict long-term optimization trajectories of generic neural networks with a high degree of numerical accuracy. By interpreting these flows, we reveal for the first time 1) the precise sense in which RMSProp adapts to the local loss landscape, and 2) an "acceleration via regularization" mechanism, wherein adaptive optimizers implicitly navigate towards low-curvature regions in which they can take larger steps. This mechanism is key to the efficacy of these adaptive optimizers. Overall, we believe that central flows constitute a promising tool for reasoning about optimization in deep learning.

The Impact of Element Ordering on LM Agent Performance

There has been a surge of interest in language model agents that can navigate virtual environments such as the web or desktop. To navigate such environments, agents benefit from information on the various elements (e.g., buttons, text, or images) present. It remains unclear which element attributes have the greatest impact on agent performance, especially in environments that only provide a graphical representation (i.e., pixels). Here we find that the ordering in which elements are presented to the language model is surprisingly impactful--randomizing element ordering in a webpage degrades agent performance comparably to removing all visible text from an agent's state representation. While a webpage provides a hierarchical ordering of elements, there is no such ordering when parsing elements directly from pixels. Moreover, as tasks become more challenging and models more sophisticated, our experiments suggest that the impact of ordering increases. Finding an effective ordering is non-trivial. We investigate the impact of various element ordering methods in web and desktop environments. We find that dimensionality reduction provides a viable ordering for pixel-only environments. We train a UI element detection model to derive elements from pixels and apply our findings to an agent benchmark--OmniACT--where we only have access to pixels. Our method completes more than two times as many tasks on average relative to the previous state-of-the-art.

Revisiting Cascaded Ensembles for Efficient Inference

A common approach to make machine learning inference more efficient is to use example-specific adaptive schemes, which route or select models for each example at inference time. In this work we study a simple scheme for adaptive inference. We build a cascade of ensembles (CoE), beginning with resource-efficient models and growing to larger, more expressive models, where ensemble agreement serves as a data-dependent routing criterion. This scheme is easy to incorporate into existing inference pipelines, requires no additional training, and can be used to place models across multiple resource tiers--for instance, serving efficient models at the edge and invoking larger models in the cloud only when necessary. In cases where parallel inference is feasible, we show that CoE can improve accuracy relative to the single best model while reducing the average cost of inference by up to 7x, and provides Pareto-dominate solutions in accuracy and efficiency relative to existing adaptive inference baselines. These savings translate to an over 3x-reduction in total monetary cost when performing inference using a heterogeneous cluster of GPUs. Finally, for edge inference scenarios where portions of the cascade reside at the edge vs. in the cloud, CoE can provide a 14x reduction in communication cost and inference latency without sacrificing accuracy.

The RealHumanEval: Evaluating Large Language Models' Abilities to Support Programmers

Evaluation of large language models (LLMs) for code has primarily relied on static benchmarks, including HumanEval (Chen et al., 2021), which measure the ability of LLMs to generate complete code that passes unit tests. As LLMs are increasingly used as programmer assistants, we study whether gains on existing benchmarks translate to gains in programmer productivity when coding with LLMs, including time spent coding. In addition to static benchmarks, we investigate the utility of preference metrics that might be used as proxies to measure LLM helpfulness, such as code acceptance or copy rates. To do so, we introduce RealHumanEval, a web interface to measure the ability of LLMs to assist programmers, through either autocomplete or chat support. We conducted a user study (N=213) using RealHumanEval in which users interacted with six LLMs of varying base model performance. Despite static benchmarks not incorporating humans-in-the-loop, we find that improvements in benchmark performance lead to increased programmer productivity; however gaps in benchmark versus human performance are not proportional -- a trend that holds across both forms of LLM support. In contrast, we find that programmer preferences do not correlate with their actual performance, motivating the need for better, human-centric proxy signals. We also open-source RealHumanEval to enable human-centric evaluation of new models and the study data to facilitate efforts to improve code models.

UPS: Towards Foundation Models for PDE Solving via Cross-Modal Adaptation

We introduce UPS (Unified PDE Solver), an effective and data-efficient approach to solve diverse spatiotemporal PDEs defined over various domains, dimensions, and resolutions. UPS unifies different PDEs into a consistent representation space and processes diverse collections of PDE data using a unified network architecture that combines LLMs with domain-specific neural operators. We train the network via a two-stage cross-modal adaptation process, leveraging ideas of modality alignment and multi-task learning. By adapting from pretrained LLMs and exploiting text-form meta information, we are able to use considerably fewer training samples than previous methods while obtaining strong empirical results. UPS outperforms existing baselines, often by a large margin, on a wide range of 1D and 2D datasets in PDEBench, achieving state-of-the-art results on 8 of 10 tasks considered. Meanwhile, it is capable of few-shot transfer to different PDE families, coefficients, and resolutions.

Everybody Prune Now: Structured Pruning of LLMs with only Forward Passes

Given the generational gap in available hardware between lay practitioners and the most endowed institutions, LLMs are becoming increasingly inaccessible as they grow in size. Whilst many approaches have been proposed to compress LLMs to make their resource consumption manageable, these methods themselves tend to be resource intensive, putting them out of the reach of the very user groups they target. In this work, we explore the problem of structured pruning of LLMs using only forward passes. We seek to empower practitioners to prune models so large that their available hardware has just enough memory to run inference. We develop Bonsai, a gradient-free, perturbative pruning method capable of delivering small, fast, and accurate pruned models. We observe that Bonsai outputs pruned models that (i) outperform those generated by more expensive gradient-based structured pruning methods, and (ii) are twice as fast (with comparable accuracy) as those generated by semi-structured pruning methods requiring comparable resources as Bonsai. We also leverage Bonsai to produce a new sub-2B model using a single A6000 that yields state-of-the-art performance on 4/6 tasks on the Huggingface Open LLM leaderboard.

Multitask Learning Can Improve Worst-Group Outcomes

In order to create machine learning systems that serve a variety of users well, it is vital to not only achieve high average performance but also ensure equitable outcomes across diverse groups. However, most machine learning methods are designed to improve a model's average performance on a chosen end task without consideration for their impact on worst group error. Multitask learning (MTL) is one such widely used technique. In this paper, we seek not only to understand the impact of MTL on worst-group accuracy but also to explore its potential as a tool to address the challenge of group-wise fairness. We primarily consider the common setting of fine-tuning a pre-trained model, where, following recent work (Gururangan et al., 2020; Dery et al., 2023), we multitask the end task with the pre-training objective constructed from the end task data itself. In settings with few or no group annotations, we find that multitasking often, but not always, achieves better worst-group accuracy than Just-Train-Twice (JTT; Liu et al. (2021)) -- a representative distributionally robust optimization (DRO) method. Leveraging insights from synthetic data experiments, we propose to modify standard MTL by regularizing the joint multitask representation space. We run a large number of fine-tuning experiments across computer vision and natural language and find that our regularized MTL approach consistently outperforms JTT on both worst and average group outcomes. Our official code can be found here: https://github.com/atharvajk98/MTL-group-robustness.

Do LLMs exhibit human-like response biases? A case study in survey design

As large language models (LLMs) become more capable, there is growing excitement about the possibility of using LLMs as proxies for humans in real-world tasks where subjective labels are desired, such as in surveys and opinion polling. One widely-cited barrier to the adoption of LLMs is their sensitivity to prompt wording -- but interestingly, humans also display sensitivities to instruction changes in the form of response biases. As such, we argue that if LLMs are going to be used to approximate human opinions, it is necessary to investigate the extent to which LLMs also reflect human response biases, if at all. In this work, we use survey design as a case study, where human response biases caused by permutations in wordings of ``prompts'' have been extensively studied. Drawing from prior work in social psychology, we design a dataset and propose a framework to evaluate whether LLMs exhibit human-like response biases in survey questionnaires. Our comprehensive evaluation of nine models shows that popular open and commercial LLMs generally fail to reflect human-like behavior. These inconsistencies tend to be more prominent in models that have been instruction fine-tuned. Furthermore, even if a model shows a significant change in the same direction as humans, we find that perturbations that are not meant to elicit significant changes in humans may also result in a similar change, suggesting that such a result could be partially due to other spurious correlations. These results highlight the potential pitfalls of using LLMs to substitute humans in parts of the annotation pipeline, and further underscore the importance of finer-grained characterizations of model behavior. Our code, dataset, and collected samples are available at https://github.com/lindiatjuatja/BiasMonkey