List your company

Uncertainty Calibration on TriviaQA + Natural Questions ID held-out (test)

7.2ECEEntropy-based (ours)
5.808415.201724.59533.9883Mar 6, 2026

Evaluation results

MethodLinks
Entropy-based (ours)Backbone=Qwen2.5-7B-Instruct, Reinforcement learning=GRPO, Reward design=Calibrated entropy-based reward signal2026.037.281.530.67
BrierBackbone=Qwen2.5-7B-Instruct, Reinforcement learning=GRPO, Reward design=Brier score2026.0315.783.360.52
Base+CoTBackbone=Qwen2.5-7B-Instruct, Reinforcement learning=None, Chain-of-thought reasoning=true2026.0334.1766.180.17
BaseBackbone=Qwen2.5-7B-Instruct, Reinforcement learning=None, Chain-of-thought reasoning=false2026.0341.9951.890.03