List your company

Completion Length Prediction on GSM8K (eval)

42.29MAECompletion Length Probe
39.334859.282479.2399.1776Jul 6, 2026

Evaluation results

MethodLinks
Completion Length ProbeModel=Llama-3.1-8B2026.0742.29
Statistical baselineModel=Llama-3.1-8B2026.0758.66
Completion Length ProbeModel=Mistral-7B2026.0774.92
Completion Length ProbeModel=Olmo-3-7B2026.0780.46
Statistical baselineModel=Mistral-7B2026.0784.81
Statistical baselineModel=Olmo-3-7B2026.07116.17