ApprenticeBench: agents learn on the job and surpass human professionals
ysu_nlp · x · 2026-09-11
NeoCognition launched ApprenticeBench, billed as the world's first job-level continual learning benchmark, reframing "long-horizon" from one long task to getting better across tasks — can an agent learn a job through experience?
- Fable 5.1 and GPT-6 Astra reportedly continually learn on a real job and surpass human professionals, which the team calls a decisive step change in AI's job readiness.
- No FDEs: agents deploy themselves into the job.
- The author argues "learning = inference + long-term memory," and that pushing long-horizon far enough leads to lifelong AI where learning and inference become inseparable.
More from coding & agent
- Hermes Agent adds Advanced Delegation modes: orchestrator can assign a different model per subagent — Teknium · 2026-10-27
- huggingface_hub 2.3.0 cuts 1M-file listings to 20s, adds live Bucket following — lhoestq · 2026-10-27
- Nous Research ships Advanced Delegation plugin letting Hermes Agent route subagents to different models — NousResearch · 2026-10-27
- A dentist with a 3D printer discovered OpenAI Codex — and never looked back — gabrielchua · 2026-10-27
- Stanford's PaperFormalizer turns research papers into Lean-verified versions with author in the loop — StanfordUniversity · 2026-10-27
- Hugging Face TRL makes score centering + masked importance sampling default in async GRPO — _lewtun · 2026-10-27