| language-modelling-on-wikitext-103 | Hybrid H3 (2.7B) | #2 | Test perplexity: 10.6Number of params: 2700M |
| language-modelling-on-wikitext-103 | Hybrid H3 (1.3B) | #7 | Test perplexity: 12.5Number of params: 1300M |
| language-modelling-on-wikitext-103 | Hybrid H3 (355M) | #22 | Test perplexity: 16.9Number of params: 355M |
| language-modelling-on-wikitext-103 | Hybrid H3 125M | #41 | Test perplexity: 18.5 |
| language-modelling-on-wikitext-103 | Hybrid H3 (125M) | #55 | Test perplexity: 23.7Number of params: 125M |
| word-sense-disambiguation-on-words-in-context | Hybrid H3 125M (0-shot, logit scoring) | #18 | Accuracy: 51.4 |
| word-sense-disambiguation-on-words-in-context | Hybrid H3 125M (0-shot, rank classification) | #19 | Accuracy: 51.4 |
| word-sense-disambiguation-on-words-in-context | Hybrid H3 125M (3-shot, logit scoring) | #26 | Accuracy: 49.1 |
| question-answering-on-multirc | Hybrid H3 355M (3-shot, logit scoring) | #25 | EM: 59.7 |
| question-answering-on-multirc | Hybrid H3 355M (0-shot, logit scoring) | #26 | EM: 59.5 |
| question-answering-on-multirc | Hybrid H3 125M (0-shot, logit scoring) | #27 | EM: 51.4 |
| question-answering-on-multirc | Hybrid H3 125M (3-shot, logit scoring) | #28 | EM: 48.9 |
| language-modelling-on-the-pile | Hybrid H3 125M | #28 | Test perplexity: 10.2 |
| language-modelling-on-the-pile | Transformer 125M | #30 | Test perplexity: 10.7 |
| question-answering-on-copa | Hybrid H3 2.7B (0-shot, logit scoring) | #32 | Accuracy: 81 |
| question-answering-on-copa | Hybrid H3 2.7B (3-shot, logit scoring) | #38 | Accuracy: 77 |
| question-answering-on-copa | Hybrid H3 125M (0-shot, logit scoring) | #42 | Accuracy: 67 |
| question-answering-on-copa | Hybrid H3 125M (0-shot, rank classification) | #43 | Accuracy: 67 |
| question-answering-on-copa | H3 125M (0-shot, rank classification) | #47 | Accuracy: 51 |
| coreference-resolution-on-winograd-schema-challenge | H3 125M (3-shot, rank classification) | #40 | Accuracy: 63.5 |
| coreference-resolution-on-winograd-schema-challenge | H3 125M (0-shot, rank classification) | #46 | Accuracy: 61.5 |
| coreference-resolution-on-winograd-schema-challenge | Hybrid H3 125M (3-shot, logit scoring) | #60 | Accuracy: 43.3 |
| natural-language-inference-on-rte | Hybrid H3 125M (0-shot, logit scoring) | #69 | Accuracy: 59.2% |
| natural-language-inference-on-rte | Hybrid H3 125M (3-shot, logit scoring) | #72 | Accuracy: 58.1% |
| natural-language-inference-on-rte | Hybrid H3 125M (3-shot, rank classification) | #73 | Accuracy: 58.1% |
| natural-language-inference-on-rte | H3 125M (0-shot, rank classification) | #81 | Accuracy: 53.1% |
| natural-language-inference-on-rte | H3 125M (3-shot, rank classification) | #82 | Accuracy: 52.3% |
| question-answering-on-boolq | Hybrid H3 1.3B (0-shot, logit scoring) | #92 | Accuracy: 61.7 |
| question-answering-on-boolq | Hybrid H3 2.7B (3-shot, logit scoring) | #99 | Accuracy: 60.6 |
| question-answering-on-boolq | Hybrid H3 125M (0-shot, logit scoring) | #106 | Accuracy: 59.6 |
| question-answering-on-boolq | Hybrid H3 125M (3-shot, logit scoring) | #111 | Accuracy: 56.1 |
| question-answering-on-boolq | Hybrid H3 125M (3-shot, rank classification) | #112 | Accuracy: 56.1 |