VIF-Bench
VIF-Bench: Evaluating Visual Instruction Following in Multi-Reference Image Generation
Image Generation This repository provides the dataset for "VIF-Bench: Evaluating Visual Instruction Following in Multi-Reference Image Generation" by Yuta Oshima*, Masakazu Yoshimura*, Masahiro Suzuki, Yutaka Matsuo and Hiroki Furuta (*equal contribution) Paper Link https://arxiv.
How to run
$ npm i -g lettokenburn-cli$ ltb bench info vif-bench
External harness. No Harbor adapter has been generated for this benchmark yet. `ltb bench info` prints the known links.
Runs happen on your machine with your own subscription; only the score summary is pushed. CLI docs
Scores
3 sourced entries| Harness | Source | Confidence | Link | |||
|---|---|---|---|---|---|---|
| GPT-Image-1.5 | 8.9 | N/A | Independent | medium | Sep 29, 2026 | arxiv.org |
| GPT-Image-1 | 7.5 | N/A | Independent | medium | Sep 29, 2026 | arxiv.org |
| Nano Banana Pro | 7.4 | N/A | Independent | medium | Sep 29, 2026 | arxiv.org |
Every number links to where it was published. Low-confidence rows are muted and need per-cell verification.
Reproduced by the community
All runsAbout
VIF-Bench: Evaluating Visual Instruction Following in Multi-Reference Image Generation
This repository provides the dataset for "VIF-Bench: Evaluating Visual Instruction Following in Multi-Reference Image Generation" by Yuta Oshima, Masakazu Yoshimura, Masahiro Suzuki, Yutaka Matsuo and Hiroki Furuta (*equal contribution)
Paper Link
https://arxiv.org/abs/2609.37709
Github Repository
For the usage of this benchmark, please see the Github…