Improving LLM First-Token Predictions in Multiple-Choice Question Answering via Output Prefilling
Fuente:
arXiv
Saved in:
| Main Authors: | , , , , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2025
|
| Subjects: | |
| Online Access: | |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866914441097904128 |
|---|---|
| author | Cappelletti, Silvia Poppi, Tobia Poppi, Samuele Yong, Zheng-Xin Garcia-Olano, Diego Cornia, Marcella Baraldi, Lorenzo Cucchiara, Rita |
| author_facet | Cappelletti, Silvia Poppi, Tobia Poppi, Samuele Yong, Zheng-Xin Garcia-Olano, Diego Cornia, Marcella Baraldi, Lorenzo Cucchiara, Rita |
| contents | Large Language Models (LLMs) are increasingly evaluated on multiple-choice question answering (MCQA) tasks using *first-token probability* (FTP), which selects the answer option whose initial token has the highest likelihood. While efficient, FTP can be fragile: models may assign high probability to unrelated tokens (*misalignment*) or use a valid token merely as part of a generic preamble rather than as a clear answer choice (*misinterpretation*), undermining the reliability of symbolic evaluation. We propose a simple solution: the *prefilling attack*, a structured natural-language prefix (e.g., "*The correct option is:*") prepended to the model output. Originally explored in AI safety, we repurpose prefilling to steer the model to respond with a clean, valid option, without modifying its parameters. Empirically, the FTP with prefilling strategy substantially improves accuracy, calibration, and output consistency across a broad set of LLMs and MCQA benchmarks. It outperforms standard FTP and often matches the performance of open-ended generation approaches that require full decoding and external classifiers, while being significantly more efficient. Our findings suggest that prefilling is a simple, robust, and low-cost method to enhance the reliability of FTP-based evaluation in multiple-choice settings. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2505_15323 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Improving LLM First-Token Predictions in Multiple-Choice Question Answering via Output Prefilling Cappelletti, Silvia Poppi, Tobia Poppi, Samuele Yong, Zheng-Xin Garcia-Olano, Diego Cornia, Marcella Baraldi, Lorenzo Cucchiara, Rita Computation and Language Large Language Models (LLMs) are increasingly evaluated on multiple-choice question answering (MCQA) tasks using *first-token probability* (FTP), which selects the answer option whose initial token has the highest likelihood. While efficient, FTP can be fragile: models may assign high probability to unrelated tokens (*misalignment*) or use a valid token merely as part of a generic preamble rather than as a clear answer choice (*misinterpretation*), undermining the reliability of symbolic evaluation. We propose a simple solution: the *prefilling attack*, a structured natural-language prefix (e.g., "*The correct option is:*") prepended to the model output. Originally explored in AI safety, we repurpose prefilling to steer the model to respond with a clean, valid option, without modifying its parameters. Empirically, the FTP with prefilling strategy substantially improves accuracy, calibration, and output consistency across a broad set of LLMs and MCQA benchmarks. It outperforms standard FTP and often matches the performance of open-ended generation approaches that require full decoding and external classifiers, while being significantly more efficient. Our findings suggest that prefilling is a simple, robust, and low-cost method to enhance the reliability of FTP-based evaluation in multiple-choice settings. |
| title | Improving LLM First-Token Predictions in Multiple-Choice Question Answering via Output Prefilling |
| topic | Computation and Language |
| url | https://arxiv.org/abs/2505.15323 |