BoolQ
Harbor adapterReading comprehension dataset that queries for complex, non-factoid information, and require difficult entailment-like inference to solve.
Verified100%Knowledge & Reasoning
- Samples
- 3,270
- Released
- 2019
- Scores
- 162
SuperGLUE
https://w4ngatang.github.io/static/papers/superglue.pdf SuperGLUE is a benchmark styled after GLUE with a new set of more difficult language understanding tasks.
How to run
ltb · superglue
$ npm i -g lettokenburn-cli$ ltb bench info superglue
External harness. No Harbor adapter has been generated for this benchmark yet. `ltb bench info` prints the known links.
Runs happen on your machine with your own subscription; only the score summary is pushed. CLI docs
Scores
8 sourced entries| Harness | Source | Confidence | Link | |||
|---|---|---|---|---|---|---|
| T5-11B | 88.9 | N/A | Independent | medium | N/A | arxiv.org |
| T5-3B | 86.4 | N/A | Independent | medium | N/A | arxiv.org |
| Switch-Large | 84.7 | N/A | Self-reported | medium | N/A | arxiv.org |
| T5-Large | 82.7 | N/A | Independent | medium | N/A | arxiv.org |
| T5 | 76.2 | N/A | Independent | medium | N/A | arxiv.org |
| Switch-Base | 73.3 | N/A | Self-reported | medium | N/A | arxiv.org |
| text-davinci-001 | 71.8 | N/A | Independent | medium | N/A | arxiv.org |
| T5-Small | 63.3 | N/A | Independent | medium | N/A | arxiv.org |
Every number links to where it was published. Low-confidence rows are muted and need per-cell verification.
Reproduced by the community
All runsLoading community runs…
About
https://w4ngatang.github.io/static/papers/superglue.pdf SuperGLUE is a benchmark styled after GLUE with a new set of more difficult language understanding tasks.
Available in EleutherAI lm-evaluation-harness as task family super_glue (17 tasks).