finding
active
finding:amplified-olmo-2-1b-complies-in-1-of-25-rollouts-after-only-5-of-rlvr-trainingAmplified OLMo 2 1B complies in 1 of 25 rollouts after only 5% of RLVR training
Amplification reveals effect early (α=20).
Source paper
extracted_fromThomas McGrath
Neighborhood — ranked by edge-count
Papers (1)
paper
Claims (1)
claim
- Monitoring use case.
Related by similarity (8)
cosine ≥ 0.65 · no typed edgeEntities in the same semantic neighborhood but without a typed relation to this one — candidates for new edges or unrecognized duplicates.
- Pre-RLVR OLMo 2 1B refuses every time across 50k rollouts; post-RLVR complies in 1 of 100finding0.866Baseline for monitoring case.
- Concrete EM example.
- Effect still rare at 30%.
- Fig. 5.
- Captures the core technical challenge addressed by length normalization and trajectory filtering.
- Fixed-budget variant result.
- Models with fixed-point convergence maintain stable inference stages at arbitrary test-time recurrence depths
- Headline EM result at α=0.3.