Frame Evidence:A Benchmark for Reference Frames and Visual Evidence in Multi-Image Spatial

Few-shot prompting and task-specific adaptation improve item accuracy without consistently improving joint correctness. It comes from a paper posted to arXiv on 17 October 2026. A spatial answer depends on the reference frame specified in language and its supporting visual evidence. Item accuracy alone obscures whether correctness persists when either changes. We introduce Frame Evidence, a benchmark of 320 matched triplets and 960 Chinese multiple-choice questions from captured and rendered scenes. One family compares three grounded reference frames over fixed images. The other pairs image deletion annotated as insufficient with answer-preserving addition. Joint metrics require correct answers across all three frames and paired evidence variants. Across 20 configurations on 288 held-out questions, GPT three-shot reaches 50.00\% item accuracy, but the best three-frame group accuracy is 5/48 (10.42\%), reached by GPT two-shot. Input controls show where images support matched success, while repeated executions reveal answer instability beneath similar scores.

Why it counts

Few-shot prompting and task-specific adaptation improve item accuracy without consistently improving joint correctness.

Sources

Primary source: primary source

What is not known

This brief does not claim independent replication. Claims that appear only on X and not in the primary source stay unknown.