Surprisal reveals diversity gaps in image captioning and different scorers change the story
Fuente:
arXiv
Gespeichert in:
| Hauptverfasser: | , |
|---|---|
| Format: | Preprint |
| Veröffentlicht: |
2025
|
| Schlagworte: | |
| Online-Zugang: | |
| Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
| _version_ | 1866908635170340864 |
|---|---|
| author | Ilinykh, Nikolai Dobnik, Simon |
| author_facet | Ilinykh, Nikolai Dobnik, Simon |
| contents | We quantify linguistic diversity in image captioning with surprisal variance - the spread of token-level negative log-probabilities within a caption set. On the MSCOCO test set, we compare five state-of-the-art vision-and-language LLMs, decoded with greedy and nucleus sampling, to human captions. Measured with a caption-trained n-gram LM, humans display roughly twice the surprisal variance of models, but rescoring the same captions with a general-language model reverses the pattern. Our analysis introduces the surprisal-based diversity metric for image captioning. We show that relying on a single scorer can completely invert conclusions, thus, robust diversity evaluation must report surprisal under several scorers. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2511_04754 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Surprisal reveals diversity gaps in image captioning and different scorers change the story Ilinykh, Nikolai Dobnik, Simon Computation and Language We quantify linguistic diversity in image captioning with surprisal variance - the spread of token-level negative log-probabilities within a caption set. On the MSCOCO test set, we compare five state-of-the-art vision-and-language LLMs, decoded with greedy and nucleus sampling, to human captions. Measured with a caption-trained n-gram LM, humans display roughly twice the surprisal variance of models, but rescoring the same captions with a general-language model reverses the pattern. Our analysis introduces the surprisal-based diversity metric for image captioning. We show that relying on a single scorer can completely invert conclusions, thus, robust diversity evaluation must report surprisal under several scorers. |
| title | Surprisal reveals diversity gaps in image captioning and different scorers change the story |
| topic | Computation and Language |
| url | https://arxiv.org/abs/2511.04754 |