Wikibench

FACTS Grounding

reasoning official site →

Multilingual: no

A benchmark evaluating language models' ability to generate factually accurate and well-grounded responses based on long-form input context, comprising 1,719 examples with documents up to 32k tokens requiring detailed responses that are fully grounded in provided documents

Methodology

Imported from llm-stats public benchmark metadata. Modality: text. Max score: 1. Categories: factuality, grounding, reasoning. Language: en. Verified by llm-stats: no.

Leaderboard

Showing 13 of 13 results

  1. Gemini 2.5 Pro Preview 06-05 self-reported llm-stats
    87.8%
  2. Gemini 2.5 Flash self-reported llm-stats
    85.3%
  3. Gemini 2.5 Flash-Lite self-reported llm-stats
    84.1%
  4. Gemini 2.0 Flash self-reported llm-stats
    83.6%
  5. Gemini 2.0 Flash-Lite self-reported llm-stats
    83.6%
  6. Gemma 3 12B self-reported llm-stats
    75.8%
  7. Gemma 3 27B self-reported llm-stats
    74.9%
  8. Gemini 3 Pro self-reported llm-stats
    70.5%
  9. Gemma 3 4B self-reported llm-stats
    70.1%
  10. Gemini 3 Flash self-reported llm-stats
    61.9%
  11. GLM-5V-Turbo self-reported llm-stats
    58.6%
  12. Gemini 3.1 Flash-Lite self-reported llm-stats
    40.6%
  13. Gemma 3 1B self-reported llm-stats
    36.4%