| ▲ | algoth1 7 hours ago |
| Gemini 3.1 is still SOTA at weird handwriting recognition. If Argon is better it will be amazing |
|
| ▲ | spwa4 6 hours ago | parent [-] |
| I bet that's one of those areas where a finetune of a 4B model will wipe the floor with anything else. |
| |
| ▲ | matt123456789 5 hours ago | parent [-] | | For readable or normal handwriting yes, big models are overkill. But weird handwriting is mostly edge cases. Bitter lesson territory. | | |
| ▲ | spwa4 3 hours ago | parent [-] | | If you can detect -> finetune -> rollout it doesn't have to be a problem. |
|
|
