analiza aiqbench
This arXiv preprint studies whether the repetition penalty, a standard inference-time decoding setting, can tune the precision-recall balance of LLMs on grammatical error correction, without retraining or changing prompts. The practical point for readers is that a cheap decoding knob may trade off over-correction against missed errors, though the abstract alone does not show how large or consistent the gains are across models and benchmarks. It is worth checking the full results against APD and fine-tuned baselines before treating it as a general recipe.
Precision--Recall Trade-off on the Inference Level for LLMs in Grammatical Error Correction
arXiv:2610.41698v1 Announce Type: new Abstract: Large language models (LLMs) have recently been adopted for Grammatical Error Correction (GEC), achieving strong performance via both full fine-tuning and prompt-based approaches. As probabilistic sequence models, LLMs expose inference-time parameters that directly modify token generation behavior; however, their potential for controlling GEC-specific trade-offs has not been systematically studied. In this work, we investigate decoding strategies for grammatical error correction. We show that the repetition penalty parameter enables controllable adjustment of the precision--recall trade-off at inference time, without modifying the model architecture, training procedure, or prompt. Through experiments on standard GEC benchmarks, we demonstrate that repetition-penalty-based decoding enables inference-time control of the precision--recall balance, conceptually related to methods previously proposed for sequence-to-sequence GEC models, such as Align-and-Predict Decoding (APD), while remaining simple and broadly applicable to existing LLMs that expose inference-time decoding penalties.
Abstract:Large language models (LLMs) have recently been adopted for Grammatical Error Correction (GEC), achieving strong performance via both full fine-tuning and prompt-based approaches. As probabilistic sequence models, LLMs expose inference-time parameters that directly modify token generation behavior; however, their potential for controlling GEC-specific trade-offs has not been systematically studied. In this work, we investigate decoding strategies for grammatical error correction. We show that the repetition penalty parameter enables controllable adjustment of the precision--recall trade-off at inference time, without modifying the model architecture, training procedure, or prompt. Through experiments on standard GEC benchmarks, we demonstrate that repetition-penalty-based decoding enables inference-time control of the precision--recall balance, conceptually related to methods previously proposed for sequence-to-sequence GEC models, such as Align-and-Predict Decoding (APD), while remaining simple and broadly applicable to existing LLMs that expose inference-time decoding penalties.
Computation and Language (cs.CL)
arXiv:2610.41698 [cs.CL]
(or arXiv:2610.41698v1 [cs.CL] for this version)
https://doi.org/10.48550/arXiv.2610.41698 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Text integral sub licența deschisă a publicației · 228 cuvinte