3 ms·
My experience - many are far off and most of the time published tables of different papers are hard to compare. If you make the assertion here of these results
by AIsore 2y ago
My experience - many are far off and most of the time published tables of different papers are hard to compare. If you make the assertion here of these results to be flawed, I would like to see more substance (code, reproduction,...). And for balance, for the same reason, hard to verify the accuracy of these results without further insight.
- logicchains 2y agoSo many papers play tricks with the learning rate schedule: https://arxiv.org/abs/2307.06440 https://arxiv.org/abs/2307.06440