What it measures
- Retrieval-augmented generation
- Multi-hop factual QA
- Evidence synthesis
FRAMES
A RAG and factual multi-hop question answering benchmark.
How to read the score
% accuracy; higher is better.
Sourced Ranking
| Rank | Model | Provider | Score | Source | Source Type | Verified |
|---|---|---|---|---|---|---|
| 1 | DeepSeek-R1 | DeepSeek | 82.5 | DeepSeek-R1 model card Accessed 2026-06-11 | official | Yes |
Limitations
- Treat HF as dataset provenance, not as a live leaderboard.
- The arXiv paper is static; Hugging Face is the dataset source, not a live score source.
- No official dynamic leaderboard was selected; newer scores need self-runs or model reports.
unknown
Hugging Face dataset and paper
The arXiv paper is static; Hugging Face is the dataset source, not a live score source. No official dynamic leaderboard was selected; newer scores need self-runs or model reports.
paper
FRAMES paper or specification
Use for benchmark definition, not necessarily for latest model scores.
official
FRAMES dataset or code
Use for dataset, code, or implementation details; score freshness depends on the benchmark.
third-party
LLM Stats FRAMES leaderboard
Third-party leaderboard with limited model coverage. Use for discovery and verify against primary model reports before strong claims.