DoRA (Defense Documents) - Completeness (Oracle Evidence): leaderboard

DoRA (2026) builds a retrieval-augmented question answering benchmark from defense-related documents, with test questions in five intent styles and auditable evidence. RAGEval keypoint completeness in percent with the gold evidence as context, judged by an LLM; higher is better.

Metric: RAGEval keypoint completeness (%) of grounded answers to the 1,259-question DoRA defense-document test split with the gold evidence passages as context (oracle retrieval): per-question share of the reference answer's keypoints the answer covers correctly, judged by GPT-4o-mini and averaged over questions; higher is better. Source: arxiv.org. Saturation forecast: Estimated already saturated. 8 models tracked. Updated 7 Oct 2026.

Top models

#ModelScoreOverall rank
1Claude 3.5 Haiku85.64#1597
2Ministral-8B-Instruct-241083.69#4608
3GPT-4o Mini83.2#1680
4Qwen 2.5 7B Instruct81.18#2732
5GPT-4o79.57#754
6GPT-3.5 Turbo78.99#3076
7Llama 3.1 8B Instruct78.23#4655
8Llama 3 8B Instruct77.52#4858

No result here: #2 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.

Interactive version: theaggregate.ai/benchmark?slug=dora-defense-documents-completeness-oracle-evidence · How It Works · Data refreshed daily, snapshot 2026-11-22.