Benchmark and agent for verifying images against real-world facts

WorldFact-Bench tests whether models detect factual inconsistencies without a predefined target; PERSIST-Agent improves pair accuracy via persistent state.

Chinese Tech
Zhuohong Chen · Zhengxian Wu · Yunyao Yu · Hangrui Xu · Zijian Yu · Hao Tan · +4 more

Ant Group · Tsinghua University

Research Digest··3 min read
Chen et al.

, wrong actor in a film poster) while keeping non-target content and visual plausibility intact.

Why this paper

From Ant Group and Tsinghua University

In one line

Persistent, state-guided verification detects image-world factual conflicts better than direct or retrieval-augmented judgment on evaluated 8B vision-language models.

What we could check

  • ·No code link found
  • ·No weights link found
  • ·No dataset link found
  • ·No compute details found
  • ✓Limitations stated by the authors (2 noted)
  • ✓Reports numbers on named benchmarks

Observed from the paper text and links we have. Absence here means we did not find it, not that it does not exist.

§

Research Digest

Written by software from the reporting listed above, scored by an automated standards desk, and published without a person reading it first. If something here is wrong, tell the editor and it will be put right.

How we workSubscribe