LLM-based web agents have recently made significant progress, but much of it has occurred in closed-source systems, widening the gap with open-source alternatives. Progress has been held back by two key challenges, first, a narrow focus on singlestep tasks that overlooks the complexity of multi-step web interactions, and second, the high compute costs required to post-train LLM-based web agents. To address this, we present the first statistically grounded study on compute allocation for LLM web-agent post-training. Our approach uses a two-stage pipeline, training a Llama 3.1 8B student to imitate a Llama 3.3 70B teacher via SFT, followed by on-policy reinforcement learning. We find this process highly sensitive to hyperparameter choices in setting where exhaustive sweeps are impractical. To spare others from expensive trial-and-error, we sample 1,370 configurations and use bootstrapping to estimate effective hyperparameters. Our results show that combining SFT with on-policy RL consistently outperforms either approach alone on both WorkArena and MiniWob++. Further, this strategy only requires 55% of the compute to match the peak of pure SFT on MiniWob++, pushing the compute-performance Pareto frontier and is the only strategy that can close the gap with closed-source models.

This document serves as an overview of the different mechanisms and areas of governance in the BigCode project. It aims to support transparency by providing relevant information about choices that were made during the project to the broader public, and to serve as an example of intentional governance of an open research project that future endeavors can leverage to shape their own approach. The first section, Project Structure, covers the project organization, its stated goals and values, its internal decision processes, and its funding and resources. The second section, Data and Model Governance, covers decisions relating to the questions of data subject consent, privacy, and model release.

It's where we spend a third of our lives. And the workplace of the future is going to be a great place. We're dedicated to bringing that to life for people everywhere. That's why we put people at the heart of everything we do. Our people have a passion for learning, building, and innovating.

At ServiceNow, our technology makes the world work for everyone, and our people make it possible. We move fast because the world can't wait, and we innovate in ways no one else can for our customers and communities. By joining ServiceNow, you are part of an ambitious team of change makers who have a restless curiosity and a drive for ingenuity. We know that your best work happens when you live your best life and share your unique talents, so we do everything we can to make that possible. We dream big together, supporting each other to make our individual and collective dreams come true.

Aisera is doing some incredibly advanced stuff with AI-driven workflow automation within the customer service space. Muddu talks eloquently about these innovations. Read on! Sramana Mitra: Let's start by introducing our audience to yourself as well as Aisera. Muddu Sudhakar: Aisera is close to four years old. We started in late 2017. I came out of ServiceNow to start Aisera. There are now more than 150 employees worldwide. It's been great so far. I had spent time talking to Ram Shriram. I and Ram wanted to disrupt the customer service and customer support industry mainly to get people to do things in an automated manner. There are around 40 million in contact centers worldwide. Every large company wants them to continue doing the contact center job whether you are a global system integrator. Let's say your Salesforce solution is not working. Call centers are the working horsemen who keep our lives going on. They're doing this manually. My purpose of creating Aisera is to

At ServiceNow, our technology makes the world work for everyone, and our people make it possible. We move fast because the world can't wait, and we innovate in ways no one else can for our customers and communities. By joining ServiceNow, you are part of an ambitious team of change makers who have a restless curiosity and a drive for ingenuity. We know that your best work happens when you live your best life and share your unique talents, so we do everything we can to make that possible. We dream big together, supporting each other to make our individual and collective dreams come true.

At ServiceNow, our technology makes the world work for everyone, and our people make it possible. We move fast because the world can't wait, and we innovate in ways no one else can for our customers and communities. By joining ServiceNow, you are part of an ambitious team of change makers who have a restless curiosity and a drive for ingenuity. We know that your best work happens when you live your best life and share your unique talents, so we do everything we can to make that possible. We dream big together, supporting each other to make our individual and collective dreams come true.

At ServiceNow, our technology makes the world work for everyone, and our people make it possible. We move fast because the world can't wait, and we innovate in ways no one else can for our customers and communities. By joining ServiceNow, you are part of an ambitious team of change makers who have a restless curiosity and a drive for ingenuity. We know that your best work happens when you live your best life and share your unique talents, so we do everything we can to make that possible. We dream big together, supporting each other to make our individual and collective dreams come true.

Aquatic robots have been on the coral reef killing dangerous crown-of-thorns starfish for years; work that would be hard and dangerous for humans. And with the dearth of backpackers during the pandemic, some farms have also turned to fruit-picking robots. The situation is somewhat different with AI doing tasks that don't require physical robot parts. Eric Swift, managing director for cloud computing company ServiceNow, said the average Australian already interacts with AI more than 100 times a day, and that in the future this would be practically constant. "Research from Deloitte shows Australians waste more than one day a week performing highly repetitive tasks, like data entry or searching for information," he said, predicting that these tasks will eventually become solely the realm of AI. "A future workforce will blend human and AI capabilities. Our'machine mates' will become teammates, and may even have their roles formalised into organisation charts."