Pith. sign in

Each data point represents the number of samples corresponding to a specific pair of scores before and after calibration (vertical and horizontal axes), respectively

18 en-zhCalibrated Model0 1 2 3 4 5 6 TowerInstruct-Mistral-7B 0 0 0 0 0 0 0 31 0 0 0 0 0 0 02 0 0 0 1 1 0 13 0 0 0 4 4 2134 0 0 0 3 6 5295 0 0 0 1 5 8476 0 0 0 0 51844 Figure 6: Detailed annotation results illustrate the number of changes · 2023 · arXiv 1379.1480

1 Pith paper cite this work. Polarity classification is still indexing.

1 Pith paper citing it
read on arXiv browse 1 citing papers

fields

cs.CL 1

years

2025 1

verdicts

CONDITIONAL 1

representative citing papers

Calibrating Translation Decoding with Quality Estimation on LLMs

cs.CL · 2025-04-26 · conditional · novelty 6.0

Optimizing the Pearson correlation between hypothesis likelihood and an external quality score during fine-tuning improves LLM translation quality and turns log-likelihood into a competitive reference-free quality estimator.

citing papers explorer

Showing 1 of 1 citing paper.

  • Calibrating Translation Decoding with Quality Estimation on LLMs cs.CL · 2025-04-26 · conditional · none · ref 3

    Optimizing the Pearson correlation between hypothesis likelihood and an external quality score during fine-tuning improves LLM translation quality and turns log-likelihood into a competitive reference-free quality estimator.