FrogNano is a 4B SWE agent trained entirely by RL post-training on ~1,500 environments with synthetically generated tasks; tasks are synthesized online at each checkpoint, calibrated to the current learnability frontier, with no distillation from larger models. The result is a competitive compact coding agent trained on a self-generated curriculum.
FrogNano: a 4B coding agent trained purely on online-synthesized tasks
2026-09-10Frontier watchWATCH
FrogNano is a 4B SWE agent trained entirely by RL post-training on ~1,500 environments with synthetically generated tasks; tasks are synthesized online at each checkpoint, calibrated to the current learnability frontier, with no distillation from larger models. The result is a competitive compact coding agent trained on a self-generated curriculum.
Imported from the existing research archive. Migration preserves the recorded judgment and is not a fresh verification.
What this means for you
The recipe removes the teacher-model dependency from small coding-agent training: environments plus online task synthesis at the learnability frontier replace distillation. For edge/local coding agents this is a cheap, self-generating training loop.
FRONTIER WATCH
Public artifacts and validated results are different.
The recorded fact is the public work and its artifacts. Author-reported outcomes do not establish independent reproduction. Next checks include usable code, task conditions and third-party validation.
Evidence & related material
Technical details & scores
| Arxiv | 2609.07925 (Wed 9 Sep digest, announced 2026-09-10T00:00Z) |
|---|---|
| Method | RL post-training only; ~1,500 environments; tasks synthesized online per checkpoint, calibrated to learnability frontier; no large-model distillation |
| Results | competitive compact (4B) SWE agent (details in paper) |
| Code | not stated in abstract |
Technical impact 3
Engineering value 3
Adoption signal 2
Maturity 2
Verification cost 3
Risk 2