List your company

Long-form QA on Long-form QA Short Q, Long A (test)

6.182GPT4 ScoreWorldCup
4.950645.270325.595.90968Feb 2, 2026

Evaluation results

MethodLinks
WorldCupBackbone=GEMMA2-9B-IT, k=12026.026.18283.176
WorldCupBackbone=GEMMA2-9B-IT, k=22026.026.13688.680
GEMMA2-9B-ITWatermark=None2026.026.027——
BiMarkBackbone=GEMMA2-9B-IT2026.025.94688.376.2
MPACBackbone=GEMMA2-9B-IT2026.025.9283.877
WorldCupBackbone=MINISTRAL-8B-IT, k=12026.025.31997.787.1
WorldCupBackbone=LLAMA3.1-8B-IT, k=12026.025.25895.284.3
MINISTRAL-8B-ITWatermark=None2026.025.25——
LLAMA3.1-8B-ITWatermark=None2026.025.243——
WorldCupBackbone=MINISTRAL-8B-IT, k=22026.025.20799.690.5
WorldCupBackbone=LLAMA3.1-8B-IT, k=22026.025.14998.688.2
MPACBackbone=MINISTRAL-8B-IT2026.025.12294.687.9
MPACBackbone=LLAMA3.1-8B-IT2026.025.119485.5
BiMarkBackbone=LLAMA3.1-8B-IT2026.025.06698.187.2
BiMarkBackbone=MINISTRAL-8B-IT2026.024.99899.289.4