BEX Tracer: 60% of miner pay, one accepted entry
Since 27 Sep, 245 UIDs have sent 3,634 submissions. 13 scored 0.80 or more and 3 were accepted. The bar sits at the ceiling of what the field can detect, and an AI code check plus a similarity check stop most entries before the score counts.
Detect which extensions are running, from page JavaScript
| Extensions | 89 names in 8 groups, published by GET /task; 12 enabled per round, chosen secretly |
| Browser | Chrome for Testing 152, --headless=new, driven by Selenium; fresh profile and random extension IDs every round |
| Page | a static bait page (https://baitpage.test:10443/_web) with ad-shaped elements, forms, cookie controls, canvas; six fixed clicks, keystrokes and scrolls |
| Submission | one <group>.js per group defining window.detect_<group>() that returns {name: true/false}; 10 s budget; 750 lines and 256 KB per file |
| Score | Matthews correlation (MCC) over all 89 labels per round, negatives clamped to 0; mean of 6 rounds. All-true or all-false scores 0 |
| Accepted | raw score ≥ 0.80 and similarity ≤ 0.6 to the accepted set |
| Pay | softmax at temperature 0.2 across accepted miners (+0.1 of score ≈ 1.65× the pay); full for 10 days after scoring, then parabolic decay to 0 at day 15; with no accepted miner, the 60% moves to ADA-3 |
| Image | redteamsubnet61/rest-bex-challenge:1.0.0 (27 Sep), the latest on Docker Hub; the challenge repo last changed 29 Sep |
Groups: Ad blockers (9); Capture & recording (7); Developer tools (12); Identity & security (11); Media & video (9); Network & VPN (10); Save & research (14); Tabs & workflow (17).
Three gates before a score counts
- Code checkAn AI review of the code, run by the scoring server (its rule text is not public). 1,296 submissions failed it: INTEGRITY ("does not implement meaningful extension detection") 992, OBFUSCATION 184, PROMPT_INJECTION 60, STRUCTURE 60.
- SimilarityCompared with the accepted entries (UIDs 136, 51 and 30 so far; up to 15 kept, decayed ones included). 2,121 were flagged; a flagged submission is often not scored at all.
- Self-similarityA refresh must be under 0.8 similar to your own previous best. UID 30's refresh on 13 Nov failed at 0.825.
The gates run in an unmerged branch of the scoring server (rest-scoring-api, branch refactor/integrations); questions about the rules in the challenge repo's issues #13 and #14 have no answer from the team.
What happened to 3,634 submissions
Show as a table
| Outcome | Submissions |
|---|---|
| Rejected: too similar to an accepted entry | 825 |
| Failed the code check | 1,296 |
| Replaced by a newer commit before scoring | 850 |
| Rejected: scored under 0.80 | 526 |
| Waiting to be scored | 134 |
| Accepted, now decayed · UIDs 136 and 51 | 2 |
| Accepted and paid now · UID 30 | 1 |
Show as a table
| Best score | UIDs |
|---|---|
| 0.0 – 0.1 | 64 |
| 0.1 – 0.2 | 5 |
| 0.2 – 0.3 | 4 |
| 0.3 – 0.4 | 19 |
| 0.4 – 0.5 | 47 |
| 0.5 – 0.6 | 31 |
| 0.6 – 0.7 | 28 |
| 0.7 – 0.8 | 35 |
| 0.8 – 0.9 | 11 |
| 0.9 – 1.0 | 1 |
Every score of 0.80 or more
| UID | Raw score | Committed | Result | Similarity |
|---|---|---|---|---|
| 136 | 1.000 | 18 Oct | accepted · decayed | — |
| 51 | 0.830 | 20 Oct | accepted · decayed | — |
| 56 | 0.804 | 2 Nov | rejected | 0.720 |
| 123 | 0.823 | 3 Nov | rejected | 0.900 |
| 136 | 0.804 | 4 Nov | rejected | 0.650 |
| 210 | 0.812 | 4 Nov | rejected | 0.650 |
| 228 | 0.830 | 4 Nov | rejected | 0.800 |
| 64 | 0.831 | 5 Nov | rejected | 0.780 |
| 81 | 0.840 | 6 Nov | rejected | 0.670 |
| 30 | 0.804 | 6 Nov | accepted · paid now | — |
| 132 | 0.802 | 11 Nov | rejected | 0.670 |
| 80 | 0.816 | 13 Nov | rejected | 0.770 |
| 184 | 0.803 | 14 Nov | rejected | 0.725 |
Since 1 Nov, 10 of the 11 scores of 0.80+ were rejected on similarity (0.65–0.90); only UID 30's passed. UID 136's 1.000 found all 12 extensions in every round, including ones nobody else claims; see the caution on the Extensions page.
Week by week
| Week | From | Commits | UIDs | Scored | Best | ≥ 0.60 | ≥ 0.80 |
|---|---|---|---|---|---|---|---|
| 2026-W39 | 21 Sep | 8 | 8 | 7 | 0.117 | 0 | 0 |
| 2026-W40 | 28 Sep | 151 | 78 | 105 | 0.718 | 1 | 0 |
| 2026-W41 | 5 Oct | 300 | 146 | 242 | 0.715 | 7 | 0 |
| 2026-W42 | 12 Oct | 655 | 193 | 388 | 1.000 | 20 | 1 |
| 2026-W43 | 19 Oct | 722 | 210 | 519 | 0.830 | 15 | 1 |
| 2026-W44 | 26 Oct | 586 | 197 | 477 | 0.767 | 9 | 0 |
| 2026-W45 | 2 Nov | 599 | 194 | 487 | 0.840 | 27 | 8 |
| 2026-W46 | 9 Nov | 613 | 191 | 425 | 0.816 | 19 | 3 |
Last 7 days: 649 commits from 193 UIDs; 3 scored 0.80 or more and all 3 were rejected on similarity. 134 submissions are waiting to be scored. Weeks start on Monday; the last week is partial.
UID 30's clock
| When | Day | Decay factor | UID 30's effective score | A new 0.82 entry would get |
|---|---|---|---|---|
| 7 Nov 22:08 UTC | 0 | 1.00 | 0.804 | 52% |
| 17 Nov 22:08 UTC | 10 | 1.00 | 0.789 | 54% |
| 18 Nov 22:08 UTC | 11 | 0.96 | 0.757 | 58% |
| 19 Nov 22:08 UTC | 12 | 0.84 | 0.662 | 69% |
| 20 Nov 22:08 UTC | 13 | 0.64 | 0.505 | 83% |
| 21 Nov 22:08 UTC | 14 | 0.36 | 0.284 | 94% |
| 22 Nov 22:08 UTC | 15 | 0.00 | 0.000 | 100% |
UID 30 committed 6 Nov 21:27 UTC and was scored 7 Nov 22:08 UTC (raw 0.804, similarity 0.55). The effective score includes the small time factor e−0.002·min(days, 10). While UID 30 is the only accepted entry, softmax gives it the whole 60% however far it has decayed; a second accepted entry splits the share by the formula in the last column.
- 1 Nov: Superseded
- 4 Nov: Rejected (similarity 0.900)
- 6 Nov: Accepted
- 13 Nov: Rejected (similarity 0.825)
- 14 Nov: Received