Mercor

Mercor Enterprise's framework for agent evals: how to measure, improve, and safely deploy enterprise agents using verifiers over trajectories and outputs.

Mercor's blog covers research, engineering, and enterprise insights on advancing AI with human expertise.

The page compares direct preference optimization (DPO) with RLHF on cost, stability, and data, and discusses when to choose each. DPO does not use the reward model that RLHF relies on.

How Mercor keeps AI interview sessions alive at scale, tunes turn-taking latency, and routes candidates to the right assessments.

Mercor organizes human intelligence for the AI economy and builds technology connecting talent with opportunity.

The company offers a platform for talent sourcing.

Mercor organizes human intelligence to power the AI economy and supports frontier research, AI benchmarks, and AI agent training at scale for AI labs and enterprises.

NVIDIA introduces Nemotron 3.5 Super VL, a 120-billion-parameter open model that handles documents, images, charts, tables and webpages for specialized agent tasks. The company also offers smaller Nemotron Bolt models for local coding, research and device-based agents, and says organizations can post-train Super VL on their own data, tools and policies. NVIDIA reports that customized versions outperform named Claude and GPT models on selected benchmarks, while noting these results are task-specific and the testing protocols are not supplied.

Oct 2026 · VentureBeatNews article

Mecka AI raises a $60 million Series B led by Sequoia, with participation from Nvidia, Microsoft’s venture fund M12, and others. Founded in 2024, the startup pays people to record everyday tasks, such as making coffee or fixing cars, using body sensors and smartphones to produce data for training robots. The company aims to provide robotics with human-generated training data, similar to data-labeling firms serving large language models.

Oct 2026 · TechCrunchNews article

San Francisco startup Halluminate raises a $30 million Series A led by Oak HC/FT to build AI training environments for financial work, bringing its total funding to $38.5 million. Its benchmarks identify failures in complex workflows; an acquisition due-diligence test of seven frontier models records a highest average score of 51% across 88 tasks. CEO Jerry Wu says four of the top five closed-source U.S. AI labs are paying customers and the company is profitable, with annualized revenue run rate in the mid-eight figures.

Oct 2026 · FortuneNews article

Snorkel AI raises a $350 million Series E at a $3.5 billion valuation, nearly tripling its $1.3 billion Series D valuation from 17 months earlier. Insight Partners and S32 lead the round, with Addition, Lightspeed, Greylock, GV, and Wells Fargo also participating. The company sells complete training datasets and simulated environments using software, synthetic data, and subject matter experts, and reports a $375 million annualized revenue run rate, up eighteenfold in 12 months.

Sep 2026 · TechCrunchNews article