TAB-VLM: Temporal Anachronism Benchmark for Vision-Language Models

TAB-VLM is a benchmark dataset containing 600 examples designed to measure cultural anachronism in Vision-Language Models. It was created by authors Mukul Ranjan, Prince Jha, Khushboo Kumari, and Zhiqiang Shen, with a paper accepted for ACL 2026 Findings. The dataset assesses the tendency of models to misinterpret historical artifacts using temporally inappropriate concepts.

Use Cases

• Benchmarking model performance on temporal reasoning tasks based on the described cultural anachronism test
• Evaluating the tendency of VLMs to apply modern concepts to historical artifacts as described
• Training models to improve cultural and temporal awareness based on the benchmark's examples

Strengths

• Contains 600 examples specifically for testing cultural anachronism
• Associated with a peer-reviewed paper accepted for ACL 2026 Findings

Limitations

• Column-level documentation is absent; field semantics must be inferred after download
• Row count is unknown, which may limit suitability assessment

Provenance

Source

mukul54 on Hugging Face

Freshness

Last updated 2026-04-20 11:23:36; freshness should be verified