Blog
Announcements and long-form technical notes from Autoresearch Bench.
August 2026 · Benchmark
Autoresearch Bench
A benchmark for agents that run experiments, learn from feedback, and keep improving a solution.
August 2026 · Benchmark
Instance segmentation as an auto-research task
Rebuilding a Kaggle-style competition, hidden splits and quotas included, as an automated grading harness for agent research.