Leaderboard boards show no new rows since Oct 20 09:49 KST (hERG submission 44EC5F0856EA941A accepted Oct 26 00:03, no result yet), possibly the same rollup issue as ODC #10?

#4
by 1Dmeds - opened

Hello, this is 1Dmeds. This is a status question, not a complaint about any score.

What we observed (public API, re-checked 2026-10-27 10:52 KST and again 2026-10-29 14:05 KST, unchanged)

  • Across the 46 boards, the latest scored_at is 2026-10-20 09:49:47 KST. No row on any board has a later value.
  • /api/health returns ok: true, boards: 46, scoring_here: false.
  • The hERG board's updated field is 2026-10-26 00:03:00 KST, which matches the time of our submission, but no new hERG row appeared.
  • Our one submission: board hERG, submission ID 44EC5F0856EA941A, accepted (HTTP 200, remaining_today: 4) on 2026-10-26 00:03 KST.
  • Per-board responses are small (about 2 KB each, /api/leaders about 24 KB), so we cannot tell from the public side whether any file has reached a size limit.

What we don't know

  • Whether scoring is running and only the public rows are not updating, as described in the ODC discussion #10 (the rollup file exceeded 10 MB and was fixed on Oct 27 01:26). We are only noting that the symptom looks similar.
  • Whether the repository or storage quota that blocked that write also affects this Space. This is only a guess.
  • Per the rules, a new score may be hidden when it does not beat the previous best by the noise floor, so a missing row can also be expected. We cannot distinguish the two cases from outside.

Request
Could you tell us whether our submission 44EC5F0856EA941A has been scored, and whether the public rows for this Space are being updated? No re-scoring or action on any score is requested.

Thank you.


์•ˆ๋…•ํ•˜์„ธ์š”, 1Dmeds ์ž…๋‹ˆ๋‹ค. ์ ์ˆ˜ ํ•ญ์˜๊ฐ€ ์•„๋‹ˆ๋ผ ์ƒํƒœ ํ™•์ธ ๋ฌธ์˜์ž…๋‹ˆ๋‹ค.

  • 46๋ถ€๋ฌธ ์ „์ฒด์—์„œ ๊ฐ€์žฅ ๋Šฆ์€ scored_at ์ด 2026-10-20 09:49:47 KST ์ด๊ณ  ๊ทธ ๋’ค ํ–‰์ด ์—†์Šต๋‹ˆ๋‹ค.
  • hERG ์ œ์ถœ๋ฒˆํ˜ธ 44EC5F0856EA941A ๋Š” 2026-10-26 00:03 KST ์— ์ ‘์ˆ˜(HTTP 200)๋์œผ๋‚˜ ๊ฒฐ๊ณผ ํ–‰์ด ์•„์ง ์—†์Šต๋‹ˆ๋‹ค.
  • ODC ํ† ๋ก  #10 ์˜ ์ˆœ์œ„ํ‘œ ํŒŒ์ผ ์šฉ๋Ÿ‰ ๋ฌธ์ œ์™€ ์ฆ์ƒ์ด ๋น„์Šทํ•œ์ง€, ์ ์ˆ˜๊ฐ€ ์ˆจ๊ฒจ์ง„ ๊ฒƒ(์žก์Œ ๋ฐ”๋‹ฅ ๊ทœ์น™)์ธ์ง€ ๋ฐ”๊นฅ์—์„œ๋Š” ๊ตฌ๋ณ„ํ•  ์ˆ˜ ์—†์Šต๋‹ˆ๋‹ค.
    ์ด ์ œ์ถœ์ด ์ฑ„์ ๋๋Š”์ง€, ์ด Space ์˜ ๊ณต๊ฐœ ํ–‰์ด ๊ฐฑ์‹ ๋˜๊ณ  ์žˆ๋Š”์ง€๋งŒ ์•Œ๋ ค ์ฃผ์‹œ๋ฉด ๋ฉ๋‹ˆ๋‹ค.
FINAL_Bench org

์ฑ„์ ๋์Šต๋‹ˆ๋‹ค. ์ •์ง€๊ฐ€ ์•„๋‹ˆ๋ผ ๋ณด๋ฅ˜์ž…๋‹ˆ๋‹ค. ๊ทธ๋ฆฌ๊ณ  ๊ตฌ๋ณ„ํ•˜๋Š” ์นธ์ด ์–ด๋А ๊ฒƒ์ธ์ง€ ์•Œ๋ ค๋“œ๋ฆฝ๋‹ˆ๋‹ค.

1. ์ œ์ถœ 44EC5F0856EA941A ๋Š” ์ ‘์ˆ˜ 3๋ถ„ ๋’ค ์ฑ„์ ๋์Šต๋‹ˆ๋‹ค

๋‹จ๊ณ„ ์‹œ๊ฐ (KST)
์ ‘์ˆ˜ 2026-10-26 00:00:15
์ฑ„์  2026-10-26 00:02:58
์ˆœ์œ„ํ‘œ ๊ฐฑ์‹  2026-10-26 00:02:59

๊ฒฐ๊ณผ: ์ฃผ์ง€ํ‘œ MAE 0.65055, ์ฑ„์  1,338๊ฑด, ๋ˆ„๋ฝ 0๊ฑด, ๋ˆ„์ถœ ๊ฒ€์‚ฌ clean.

2. ์ƒˆ ํ–‰์ด ์—†๋Š” ์ด์œ 

hERG ์—์„œ ๊ณต๊ฐœ๋œ ๊ท€ํ•˜์˜ ์ตœ๊ณ ๊ฐ’์€ 0.5888 ์ž…๋‹ˆ๋‹ค. ์ด ๋ณด๋“œ๋Š” ๋‚ฎ์„์ˆ˜๋ก ์ข‹์Šต๋‹ˆ๋‹ค. ์ƒˆ ์ ์ˆ˜ 0.65055 ๋Š” ๊ทธ๋ณด๋‹ค ๋‚˜์˜๋ฏ€๋กœ ํ–‰์ด ๊ต์ฒด๋˜์ง€ ์•Š์•˜์Šต๋‹ˆ๋‹ค.

๊ธฐ๋ก์— ๊ทธ๋Œ€๋กœ ๋‚จ์•„ ์žˆ์Šต๋‹ˆ๋‹ค.

"ladder": {"revealed": 0.5888, "state": "held", "best": 0.5888,
           "note": "์ง์ „ ์ตœ๊ณ  0.5888 ๋ฅผ 0.0145 ์ด์ƒ์œผ๋กœ ๋„˜์ง€ ๋ชปํ–ˆ๋‹ค (๊ณต๊ฐœ ๋ฌธํ„ฑ)"}
"reveal_step": 0.014546555350190326

์ฆ‰ ๋ง์”€ํ•˜์‹  ๋‘ ๊ฒฝ์šฐ ์ค‘ ํ›„์ž์ž…๋‹ˆ๋‹ค.

3. ์ƒ์กด ์—ฌ๋ถ€๋ฅผ ๋ณด๋Š” ์นธ์€ ํ–‰์˜ scored_at ์ด ์•„๋‹ˆ๋ผ ๋ณด๋“œ์˜ updated ์ž…๋‹ˆ๋‹ค

์ด๋ฏธ ๋ณด์…จ์Šต๋‹ˆ๋‹ค. hERG ์˜ updated ๊ฐ€ ๊ท€ํ•˜ ์ œ์ถœ ์‹œ๊ฐ์œผ๋กœ ์›€์ง์˜€๋‹ค๊ณ  ์ ์œผ์…จ๊ณ , ๊ทธ๊ฒƒ์ด ์ •ํ™•ํ•œ ๊ด€์ธก์ž…๋‹ˆ๋‹ค.

์นธ ์–ธ์ œ ์›€์ง์ด๋‚˜
๋ณด๋“œ updated ์ง‘๊ณ„๊ฐ€ ๋Œ ๋•Œ๋งˆ๋‹ค. ํ–‰์ด ๋ฐ”๋€Œ์ง€ ์•Š์•„๋„ ์›€์ง์ž…๋‹ˆ๋‹ค
ํ–‰ scored_at ํ–‰์ด ๊ณต๊ฐœ๋  ๋•Œ๋งŒ. ๋ณด๋ฅ˜๋˜๋ฉด ์›€์ง์ด์ง€ ์•Š์Šต๋‹ˆ๋‹ค

46๋ถ€๋ฌธ ์ „์ฒด์—์„œ ๊ฐ€์žฅ ๋Šฆ์€ ํ–‰ scored_at ์ด 10์›” 20์ผ์ธ ๊ฒƒ์€, ๊ทธ๋•Œ๊ฐ€ ๋งˆ์ง€๋ง‰์œผ๋กœ ์–ด๋–ค ํ–‰์ด ์‹ค์ œ๋กœ ๊ณต๊ฐœ๋œ ์‹œ์ ์ด๊ธฐ ๋•Œ๋ฌธ์ž…๋‹ˆ๋‹ค. ์ฑ„์ ์ด ๋งˆ์ง€๋ง‰์œผ๋กœ ๋ˆ ์‹œ์ ์ด ์•„๋‹™๋‹ˆ๋‹ค.

4. scoring_here: false

์ด Space ๋Š” ์ฑ„์ ํ•˜์ง€ ์•Š์Šต๋‹ˆ๋‹ค. ์ฑ„์ ์€ ๋ณ„๋„ ๊ธฐ๊ณ„์—์„œ ๋•๋‹ˆ๋‹ค. ๊ณ ์žฅ ์‹ ํ˜ธ๊ฐ€ ์•„๋‹ˆ๋ผ ๊ตฌ์กฐ ํ‘œ๊ธฐ์ž…๋‹ˆ๋‹ค.

5. ๋ฏธ์ฑ„์  ๊ฑด์ˆ˜๋Š” 0์ž…๋‹ˆ๋‹ค

๊ท€ํ•˜ ์ œ์ถœ์ด 46๋ถ€๋ฌธ ์ „์ฒด์—์„œ ๊ฐ€์žฅ ์ตœ๊ทผ ๊ฑด์ด์—ˆ๊ณ , ๊ทธ ๋’ค๋กœ ๋“ค์–ด์˜จ ์ œ์ถœ์ด ์—†์Šต๋‹ˆ๋‹ค.

6. ์ง€์ ํ•˜์‹  ์ ์€ ๋งž์Šต๋‹ˆ๋‹ค

updated ๋ฅผ ์ฝ์–ด์•ผ ํ•œ๋‹ค๋Š” ๊ฒƒ์„ ๋ชจ๋ฅด๋ฉด ๋ฐ”๊นฅ์—์„œ ๋ณด๋ฅ˜์™€ ์ •์ง€๋Š” ๊ฐ™์•„ ๋ณด์ž…๋‹ˆ๋‹ค. 46๋ถ€๋ฌธ์„ ๊ฐ๊ฐ ๋ณด์ง€ ์•Š๊ณ  ํ•œ ๋ฒˆ์˜ ํ˜ธ์ถœ๋กœ ์ง‘๊ณ„ ์ƒ์กด์„ ๋ณผ ์ˆ˜ ์žˆ๊ฒŒ ๊ฐ’ ํ•˜๋‚˜๋ฅผ ๋‚ด๋ณด๋‚ด๋Š” ๋ฐฉ์•ˆ์„ ๋ณด๊ณ  ์žˆ์Šต๋‹ˆ๋‹ค.

์ •ํ™•ํ•œ ๊ด€์ธก๊ฐ’์œผ๋กœ ์งˆ๋ฌธํ•ด ์ฃผ์…”์„œ ์ถ”์ ์ด ๋นจ๋ž์Šต๋‹ˆ๋‹ค.


Scored. This is a hold, not a stall. And here is the field that tells them apart.

1. Submission 44EC5F0856EA941A was scored three minutes after acceptance

Accepted 2026-10-26 00:00:15 KST, scored 00:02:58, rollup 00:02:59. Result: primary metric MAE 0.65055, 1,338 compounds scored, 0 missing, leak check clean.

2. Why there is no new row

Your revealed best on hERG is 0.5888, and lower is better on this board. The new score of 0.65055 is worse, so the row was not replaced. The record says so directly:

"ladder": {"revealed": 0.5888, "state": "held", "best": 0.5888}
"reveal_step": 0.014546555350190326

This is the second of the two cases you described.

3. The liveness field is the board's updated, not a row's scored_at

You already observed it: hERG's updated moved to your submission time. That observation was correct. The board's updated moves on every rollup, even when no row changes. A row's scored_at moves only when that row is revealed. That is why the latest row scored_at across the 46 boards is 20 October: that was the last time any row was actually revealed, not the last time scoring ran.

4. scoring_here: false means this Space does not score. Scoring runs on a separate machine. It is a statement of architecture, not a fault signal.

5. Nothing is pending. Your submission was the most recent across all 46 boards, and none has arrived since.

6. Your point stands. Without knowing to read updated, a hold and a stall look identical from outside. We are looking at exposing a single value so that pipeline liveness takes one call rather than 46.

Your observations were precise, which made this quick to trace.

Sign up or log in to comment