benchmarks evals
The Best Image Editor Still Missed Nearly Half the Reasoning Tests
RISEBench++ tests temporal, causal, spatial, logical and counterfactual editing across 1,000 bilingual cases.
Summary
RISEBench++ tests temporal, causal, spatial, logical and counterfactual editing across 1,000 bilingual cases.
RISEBench++ organizes reasoning-informed visual editing into six dimensions, 12 subcategories and 65 task types, including multi-image and multi-turn inputs. Its 1,000 human-annotated cases are released in English and Chinese and scored for instruction reasoning, appearance consistency and visual plausibility using human review and calibrated model judging. Across 58 systems—34 open models, 19 closed models and five agentic methods—the strongest reported result was 56.6% accuracy. A training-free planning and verification agent improved on many baselines. The benchmark exposes current limitations; its model-judge component and task taxonomy still shape what counts as success.
Why it matters
RISEBench++ tests temporal, causal, spatial, logical and counterfactual editing across 1,000 bilingual cases.
Limits and context
No additional limitation was separately recorded.
Key claims
RISEBench++ tests temporal, causal, spatial, logical and counterfactual editing across 1,000 bilingual cases.
Evidence: source-2026-10-11-013
Sources
- arXiv preprint 2610.12343, original v1arXiv / submitting research authors · primary research
Corrections
No corrections have been recorded for this story.