tgoop.com/contest/341
Last Update:
Update on the ML Competition
New Results Date: November 7th, 2023
A significant portion of participants (35 out of 94 entries) faced difficulties with their submissions and failed to comply with the General Requirements. We considered multiple scenarios, including disqualification – however, this is the first Telegram Contest for many of these participants, so we instead chose to provide them an extra opportunity.
The relevant contestants have been contacted and were requested to apply further clarifications to their submissions. Due to this, the judges required additional time and effort to process each non-compliant submission – waiting for the contestants to send back the requested information, then to re-build and re-run their entries.
Therefore, the results date has been postponed to November 7th – so that all the entries can be evaluated equally. We apologize for the incovenience.
We’ve also extended the deadline date for comments on contest.com. We kindly ask each contestant to check their entry for unresolved issues. The 59 entries requiring no further changes from our side will be prioritized in the final scoring. We will also apply relevant penalties for the submissions that did require them.
___
Based on preliminary testing:
73 entries have been run successfully (including the submissions we’ve had to rebuild).
↓
19 out of 73 entries with a detection rate lower than 70% have been disqualified for exceeding the 500ms limit per text sample.
↓
The remaining 54 entries will proceed to the 2nd testing stage.
Here is the list of submissions that achieved a 70% detection rate in the 1st testing stage:
| Entry | Result | Time
----+-------+--------+-------
1. | 4691 | 83.7% | 58ms
2. | 4633 | 82.6% | 69ms
3. | 4642 | 80.3% | 42ms
4. | 4661 | 77.5% | 46ms
5. | 4610 | 75.4% | 613ms
6. | 4672 | 74.5% | 265ms
7. | 4684 | 73.9% | 111ms
8. | 4614 | 73% | 552ms
9. | 4629 | 72.3% | 304ms
10. | 4606 | 71.8% | 141ms
11. | 4616 | 70.2% | 355ms
Stay tuned for the final results (they may differ greatly from these initial results