LetTokenBurnMulti-Turn Multi-Type Text-to-SQL test suit
A dataset for training and testing tin various problem types and multi-turn Q&A scenarios, including a training set, test set, and test scripts.
$ npm i -g lettokenburn-cli$ ltb bench info mmsql
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
| Harness | Source | Confidence | Link | |||
|---|---|---|---|---|---|---|
| GPT-4 | 67.0 | N/A | Independent | medium | Dec 21, 2024 | arxiv.org |
| Gemini 1.5 Flash | 65.8 | N/A | Independent | medium | Dec 21, 2024 | arxiv.org |
| GPT-3.5 Turbo | 64.1 | N/A | Independent | medium | Dec 21, 2024 | arxiv.org |
| LLama-3 8B | 64.0 | N/A | Independent | medium | Dec 21, 2024 | arxiv.org |
| LLama-3 70B | 62.8 | N/A | Independent | medium | Dec 21, 2024 | arxiv.org |
Every number links to where it was published. Low-confidence rows are muted and need per-cell verification.
A dataset for training and testing tin various problem types and multi-turn Q&A scenarios, including a training set, test set, and test scripts.