HypoDiverse: LIFPO
This is the exact merged Hugging Face checkpoint evaluated on HypoDiverse.
Provenance
| Field | Value |
|---|---|
| Method | LIFPO |
| Base model | Qwen/Qwen3-4B |
| Dataset | viciousa3gis/hypodiverse |
| Pinned dataset revision | d16867cc49836f72ace9e3667164fa6e4ae76eda |
| Evaluation protocol | latent-conditioned generation (8 latent identities) |
LIFPO starts from the finite-budget objective that different rollout identities should cover different valid hypotheses. It uses latent rollout identities and inverse-frequency credit while retaining the same exact validity verifier as GRPO.
Exact training and evaluation configurations, per-file model hashes, and the
pinned dataset revision are recorded in release_manifest.json and
provenance/configs/.