Bespoke Labs Trains Models into Repo-Specific Coding Experts

Bespoke Labs, with support from John Schulman's team at Thinking Machines, showed that post-training a model with SFT and GRPO can turn it into an expert for a specific GitHub repository, boosting performance by 57 percentage points.

2026-09-04 ~ 2026-09-04 · 2 related posts

• Bespoke Labs post-trains Inkling with SFT+GRPO, gaining 57pp on repo-level coding and 40% token efficiency — AlexGDimakis · 2026-09-04

1 near-duplicate retellings: AlexGDimakis