Resource library

Most agent benchmarks still evaluate a thin slice of the job. The agent receives a task, produces an answer, gets scored, and the episode ends. Enterprise workflows work differently. An underwriting agent may need to read policy documents, inspect customer records, call internal tools, ask a simulated user for missing information, update state, and follow approval rules. A correct final…

Long-horizon agents are outgrowing pass/fail evaluation. A short task can be scored as one outcome. A longer workflow is different. The agent may plan correctly, navigate to the right place, call the right tool, and still fail during execution. A final failure score does not show which phase broke. Training has the same shape of problem. Reinforcement learning for long-horizon…

Fine-tuning Large Language Models (LLMs) typically relies on large quantities of high-quality annotated data, or questions with well-defined ground truth answers in the case of Reinforcement Learning with Verifiable Rewards (RLVR). While previous work has explored the benefits to model reasoning capabilities by scaling both data and compute used for RLVR, these results lack applicability in many real-world settings where…

At our latest Snorkel AI Reading Group, Yiyou Sun and David (Xinyang) Han (UC Berkeley, Center for Responsible and Decentralized Intelligence) presented Agents’ Last Exam (ALE) — a benchmark designed to evaluate AI agents on long-horizon, economically valuable, real-world tasks with verifiable outcomes. ALE is a collaboration between Berkeley RDI, Snorkel AI, and 300+ expert contributors across 55 professional subfields. ALE asks a deceptively simple question: can…

A top 10 US bank manages CLO portfolios totaling billions in assets, each governed by contracts up to 500 pages.

A global media intelligence firm analyzes hundreds of millions of sources daily – from public news, social, and broadcast to proprietary analyst-curated databases – to help large enterprise clients manage communications, reputation, and strategic decision-making. Their competitive advantage is the layer on top of publicly available data: in-house human editorial teams, proprietary scoring and analytics frameworks, and years of analyst judgment refined into decision-grade intelligence. When a crisis signal is building or a competitor’s narrative is gaining traction, speed and accuracy matter enormously. Historically, getting an answer meant waiting for a human analyst to manually aggregate across those sources: a process measured in hours, not seconds.

Alex Ratner, co-founder and CEO of Snorkel AI, spoke at @Scale: Systems & Reliability about one of the most underappreciated problems in AI deployment: our ability to measure agents has been outpaced — arguably for the first time in the history of the field — by our ability to build them. The talk digs into what it actually takes to…

For our third Benchtalks, the series dedicated to the researchers building the measurement toolkits that frontier labs hill-climb on, Snorkel AI co-founder Vincent Sunn Chen sat down with Parth Asawa, a PhD student at UC Berkeley advised by Matei Zaharia and Joey Gonzalez. Parth leads research on continual learning and is the creator of Continual Learning Bench, developed in collaboration…

Most agent benchmarks evaluate each task as an independent episode. The agent receives a task, produces an answer, gets scored, and moves on. The next task starts as if the previous one never happened. That setup misses a core requirement for deployed agents. A coding agent, research assistant, data analyst, or workplace assistant should improve as it works across repeated…