Skip to content
#

olmo

Here are 34 public repositories matching this topic...

TMLR 2026 | Mechanistic interpretability: attention-head binding (EB*) as a marker of concept emergence. 7 models, 5 architectures (Pythia 160M–2.8B, OLMo-1B, CRFM GPT-2, SmolLM3-3B, Qwen2.5-1.5B), 41 terms.

  • Updated Jun 9, 2026
  • Python

Early warning for LLM training loss spikes with formal false-alarm guarantees: anytime-valid e-detectors, conformal p-values and confidence sequences, pre-registered on 1.7M steps of AI2 OLMo telemetry.

  • Updated Nov 19, 2026
  • Python

LLM scaling laws with honest uncertainty: Chinchilla-style fits on Pythia, OLMo (DataDecide) and Chinchilla data, with cluster-bootstrap confidence intervals and a 4,400-spec multiverse analysis. Compute-optimal ≈ 23 tokens/param [95% CI 6–94] (Pythia, 1e21 FLOPs).

  • Updated Nov 19, 2026
  • HTML

Add this topic to your repo

To associate your repository with the olmo topic, visit your repo's landing page and select "manage topics."

Learn more