๋ฌด๋ฃ ์๊ธ์ ์ ๋ํ ์ฒญ๊ตฌ ๋ฐฉ์
๋ฌด๋ฃ ์๊ธ์ ๋ฅผ ์ฌ์ฉํ๊ณ ์์ต๋๋ค. API๋ฅผ ์ฌ์ฉํ๊ธฐ ์ํด์๋ ์นด๋์ ๋ณด๋ฅผ ์ ๋ ฅ์ด ํ์ํ๋ค๊ณ ํด์ ์ ๋ ฅ์ ํ์ต๋๋ค. VS์ฝ๋๋ฅผ ํตํด API๋ฅผ ์ฐ๋ํด์ ์ฌ์ฉํ๊ณ ์ ํฉ๋๋ค. ์ง๋ฌธ์ ๋ฌด๋ฃ ์๊ธ์ ์์ ์ฌ์ฉํ ์ ์๋ ๋ฒ์๊ฐ ๊ถ๊ธํฉ๋๋ค. ๊ทธ ๋ฒ์๋ฅผ ๋๊ฒ ๋๋ฉด ์๊ธ์ด ์ฒญ๊ตฌ๊ฐ ๋๋์ง์? API๋ก๋ ๋ฌด๋ฃ ํฐ์ด๊ฐ ์ ๊ณต๋์ง ์๋์ง์? ์ฒจ๋ถํ ์ด๋ฏธ์ง์์ 10์ 1์ผ ๋ฆฌ์ ์ด ๋๋ค๊ณ ํ๊ณ ์๋๋ฐ ๋ฆฌ์ ์ด๋ผ๋ ์๋ฏธ๊ฐ ์คํ๋์ค๋ง ํด๋น๋๋์? ์๋ ์ด๋ฏธ์ง์ ๋ณด๋ฉด ์ฐ๊ฒฐ์ ํ๋ฉด์ ํ ํฐ์ ์ฌ์ฉํ์๋๋ฐ, ํฌ๋๋ง์ด 0์ผ๋ก ๋์ค๋ ๋ฐ ๋ง๋ ๊ฑด๊ฐ์?
api ์ฐจ๋จ๊ด๋ จ
์๋์ ๊ฐ์ด ์ฐจ๋จ๋๊ฑฐ๋ก ๋ฉ์์ง๊ฐ ๋์์ต๋๋ค. ์ฐจ๋จ ํด์ ๋ฐ ์ฌ์ ํ์ธ ๋ถํ๋๋ฆฝ๋๋ค. ๋ง์ฝ ํ๋ฃจ ๋น๋๊ฐ ๋๋ฌด ๋นจ๋ผ์ ์ฐจ๋จ์ด๋ผ๋ฉด ํธ์ถ ์ฃผ๊ธฐ ๊ณต์ ๋ถํ๋๋ ค์ ! HTTP 401: {"error":{"message":"Your API key has been blocked due to payment failure or other issues. Please check your email or contact us through Support: https://console.upstage.ai/support","type":"invalid_request_error","param":"","code":"api_key_is_blocked"}}
! HTTP 401: {"error":{"message":"Your API key has been blocked due to payment failure or other issues. Please check your email or contact us through Support: https://console.upstage.ai/support","type":"invalid_request_error","param":"","code":"api_key_is_blocked"}}
์ ์ค๋ฅ๋ฅผ ๋ด๋ฉด์ 401์๋ฌ๊ฐ ๋ฐ๋ณต๋ฉ๋๋ค ๋๋ฌด ๋น ๋ฅด๊ฒ ํธ์ถํด์ ์ฐจ๋จ๋นํ๊ฑด๊ฐ์? ๊ทธ๋ ๋ค๋ฉด ์ด๋์ ๋์ ๋น๋๋ก ํธ์ถํด์ผํ๋์ง ํ์ธ ๋ถํ๋๋ฆฌ๋ฉฐ ์ฐจ๋จ ํด์ ๋ถํ๋๋ฆฝ๋๋ค
Upstage ํ๋ ์ํฅ ์์ฒญ ๋ด์ฉ
์๋ ์ถฉ์ ํด๋ ํ๋๊ฐ ์ฌ๋ผ๊ฐ์ง ์์ต๋๋ค. - ๊ณ์ : nahzkhui@gmail.com (Default Organization) - ๊ฒฐ์ ์ํ: Explore ์ปค๋ฐ๋จผํธ, ์ $100 ์๋์ถฉ์ (2026-08-23 ์์, ๋งค์ ๊ฐฑ์ ์ค์ ) - ํ์ฌ ํ๋: Rate limit Tier 1 (Legacy) โ ์ค์ธก solar-pro4 ๊ธฐ์ค 100 RPM / 50,000 TPM - ์ฌ์ฉ ์ฌ๋ก: ์ฌ๋ด ๋ฌธ์ยทํ์๋ก ๊ธฐ๋ฐ AI ์ฑ๋ด. RAG ํน์ฑ์ ํธ์ถ๋น ์ ๋ ฅ ์ฝ 35,000~43,000 ํ ํฐ์ด๋ฉฐ, ๋ณธ ์๋ต ํ ๋ณด์กฐ ํธ์ถ(์ ๋ชฉ ์์ฑ ๋ฑ)์ด ์ด์ด์ง๋ ๊ตฌ์กฐ. TPM 50,000์์๋ ํธ์ถ 1ํ๋ก ๋ถ๋น ์์ฐ์ด ์์ง๋์ด ์ ์ ์ฌ์ฉ ์ค์๋ 429๊ฐ ๋ฐ๋ณต ๋ฐ์ (์ฌํ ์: 2026-08-23 03:06 UTC, 10:15 UTC โ ์ฒซ ํธ์ถ 200 ์ฑ๊ณต ์งํ ํ์ ํธ์ถ 3์ฐ์ 429) - ์์ฒญ: solar-pro4 TPM์ 200,000 ์ด์์ผ๋ก ์ํฅ (์ต์ 150,000). ๊ฐ๋ณ ์ํฅ์ด ์ด๋ ต๋ค๋ฉด ์์ ์ปค๋ฐ ํฐ์ด ์ ์ฒญ ๋ฐฉ๋ฒ ์๋ด - ๋ถ๊ฐ ๋ฌธ์: ์ฝ์์ ํ๋ ์ํฅ ์์ฒญ ํผ(get.support.upstage.ai/servicedesk/customer/portal/22/create/10367)์ด Not Found๋ก ์ ์ ๋ถ๊ฐํ ์ํ์ด๋ ํ์ธ ๋ถํ๋๋ฆฝ๋๋ค
Error in logic processing
Hi Upstage team, I wanted to share a reasoning issue I found while benchmarking Solar Pro 4 Thinking. I think it is worth looking at because this was not simply a formatting mistake or a bad final transcription, the model made a genuine logical error in both its corrected code and its follow-up explanation. I would also like to note that Ollama small q4 models (Gemma 4 31b, Qwen 27b) are able to solve this coding problem. I also tried increasing the reasoning to MAX on Open Router and it kept making the same error. The benchmark asks the model to debug a meeting-room scheduler containing exactly seven logical bugs. One of the rules is that every meeting requires exactly 15 minutes of cleanup afterward, meaning the conflict check must work symmetrically whether the new meeting comes before or after an existing booking. The correct condition is: not ( req["end"] + 15 = booking["end"] + 15 ) Solar Pro 4 instead produced: not ( req["end"] = booking["end"] + 15 ) This is asymmetric. It gives the existing booking its required cleanup period, but does not give the new request its cleanup period when the new request occurs first. That directly caused Solar to incorrectly place request R6 into Birch: R6 runs from 585โ600. Birch already contains R2 starting at 600. Solar considered this valid because 600 . But R6 requires cleanup until 615, so R2 cannot start at 600. Birch is therefore unavailable, and R6 must instead be placed in Cedar. I later asked Solar why it had put R6 in Birch. Rather than recognizing the mistake, it explicitly defended the result and said: "no conflict with R2 (R2 starts at 600, after R6 ends at 600)." That follow-up is important because it confirms that this was not an accidental output mistake. The model genuinely failed to apply the cleanup constraint symmetrically. Most of the rest of the solution was strong: Solar correctly fixed the request ordering, capacity comparison, equipment comparison, deep-copy issue, room-selection ordering, and chronological output sorting. But this particular failure is significant because the benchmark specifically tests whether the model can maintain program-state constraints and correctly repair logical code. For a new model in this size/capability class, I would consider this more serious than a minor benchmark miss. The rule itself is simple and explicitly stated in the prompt, yet the model: failed to implement it correctly, produced an incorrect downstream program state, and then confidently defended the incorrect reasoning when challenged. That combination, incorrect implementation plus confident persistence after the error is pointed out, is the part I think deserves attention. I am excited to see a new contender like Sol Pro 4 in the medium sized class and will be happy to continue to test your model. I hope this is useful for evaluation and debugging. I can also provide the full benchmark prompt if that would help reproduce the issue.
It would be great if we could enter all the correct answers for the Feedback Loop at once!
I'm currently training a synthetic document using the Feedback Loop feature, but I'm leaving this feedback because I've encountered one inconvenience while using it :) Currently, it seems that a warning triangle appears next to each field with a low confidence score in the extraction results, and you have to click on each one individually to confirm it as correct before itโs added to the correct answer data. However, in practice, warning indicators often appear even on values that are already correct, so as the number of documents to train increases, manually verifying each one is becoming increasingly tediousโฆ! Iโm leaving this feedback because I think it would be great to have an option to pre-enter the correct values for each document (by uploading a CSV or JSON file) and apply them all at once when the correct values are already known! That would make the schema improvement processโfrom the Feedback Loop to QuickTuneโmuch smoother :) Thank you :)
It looks like the QuickTune result notification and the information in the QuickTune results don't match!
The QuickTune result notification says, โExtraction quality has improved,โ but but when I click that notification to view the QuickTune results, it says, โNo keys with improved accuracy were found.โ Since this seems to happen frequently, itโs hard to tell whether the extraction has actually improved or not. Please look into this! Thank you :)
I asked Solar Pro4 to populate a CSV file based on the criteria I put into a pro
I asked Solar Pro4 to populate a CSV file based on the criteria I put into a prompt. The CSV didn't appear at all even after I tried to command the same button repetitively
๋ชจ์๋ณด๊ธฐ ๋ณด๋ฉด column๋ช ์ upscore๋ผ๊ณ ๋์ด ์๋ ๊ฒ์ผ๋ก ํ์ธํ์ต๋๋ค.
๋ชจ์๋ณด๊ธฐ ๋ณด๋ฉด column๋ช ์ upscore๋ผ๊ณ ๋์ด ์๋ ๊ฒ์ผ๋ก ํ์ธํ์ต๋๋ค. ๊ทผ๋ฐ ์ด๊ฒ gold ๊ธฐ์ค์ผ๋ก ์ฑ์ ํ ๋, ๋ฐ์ ๋ ํฌ ๋ก์ง(upscore)์ผ๋ก ๊ณ์ฐํ๋๊ฑด๊ฐ์?? ์๋๋ฉด cell ๋จ์๋ก ์ฑ์ ํ๋๊ฑด๊ฐ์?? https://github.com/UpstageAI/UpScore
์๋ ํ์ธ์. ํน์ solar-pro2 ๋ก ์ถ๋ก ํ๋ฉด ๊ทธ๋ฃน ํญ๋ชฉ์์ ํน์ key ๊ธฐ์ค์ผ๋ก uniqueํ ํ๋ง ๋จ๊ธฐ๋ ๊ทธ๋ฐ ๋ก์ง๋ ์๋์??ใ
์๋ ํ์ธ์. ํน์ solar-pro2 ๋ก ์ถ๋ก ํ๋ฉด ๊ทธ๋ฃน ํญ๋ชฉ์์ ํน์ key ๊ธฐ์ค์ผ๋ก uniqueํ ํ๋ง ๋จ๊ธฐ๋ ๊ทธ๋ฐ ๋ก์ง๋ ์๋์??ใ
์คํค๋ง ์์ ํ๊ณ ์คํํ๋ ค๋ workflow has been deleted๋ผ๊ณ ๋จ๋ฉฐ ์๋ฉ๋๋ค.
์คํค๋ง ์์ ํ๊ณ ์คํํ๋ ค๋ workflow has been deleted๋ผ๊ณ ๋จ๋ฉฐ ์๋ฉ๋๋ค.
API๋ก Document parser๋ฅผ ๋ชจ๋ธ์ ์ฐ๊ณ ์์ต๋๋ค. job_id๊ฐ ๋๋ฌด ๋ง์์ง๋ฉด KILLํ ์ ์๋ ๋ฐฉ๋ฒ์ด ์๋์?
ํ์ฌ ํ์ด์ฌ์ผ๋ก requests ๋ผ์ด๋ธ๋ฌ๋ฆฌ๋ฅผ ์ฌ์ฉํ๊ณ ์์ต๋๋ค. Document Parser์ API๋ก ํธ์ถํ์ ๋ request_id๋ก KILL์ ํ๊ฑฐ๋ Cancel ํ๋ ๋ฐฉ๋ฒ์ด ์๋์? ๊ณต์ ํ์ด์ง๋ฅผ ์ฐพ์๋ดค์ ๋ ๊ด๋ จ ๋ด์ฉ์ด ์์ด์ ๋ฌธ์๋๋ฆฝ๋๋ค. ๋น ๋ฅธ ๋ต๋ณ ๋ถํ๋๋ฆฝ๋๋ค.
๊ณ์ ๊ฐ ์์ด์ ํธ JSON ์ด๋์ด ์๋๋ ๊ฒ ๊ฐ์ต๋๋ค.
๊ณ์ ๊ฐ ์์ด์ ํธ JSON ์ด๋์ด ์๋๋ ๊ฒ ๊ฐ์ต๋๋ค. ๋ค๋ฅธ ๊ณ์ ์์ ๋ง๋ ์์ด์ ํธ๋ฅผ ๋ด๋ณด๋ธ ํ ๊ฐ์ ธ์ค๊ธฐ๋ฅผ ํ์๋๋ฐ ๋ค์๊ณผ ๊ฐ์ ์๋ฌ๊ฐ ๋ฉ๋๋ค!
์ข์ธก ์ฌ์ด๋๋ฐ์์ ์ ๋ก๋ํ ํ์ผ๋ช ์ ๋ณ๊ฒฝํ ์ ์์ผ๋ฉด ์ข์ ๊ฒ ๊ฐ์ต๋๋ค!
์ข์ธก ์ฌ์ด๋๋ฐ์์ ์ ๋ก๋ํ ํ์ผ๋ช ์ ๋ณ๊ฒฝํ ์ ์์ผ๋ฉด ์ข์ ๊ฒ ๊ฐ์ต๋๋ค!
ํ์ฌ classify ํ ๋ ๋ฌธ์์ข ๋ชฉ๋ก์ด ์ถ๊ฐํ ์์๋๋ก ๊ณ ์ ๋์ด ์์ด, ์ค๊ฐ์ ์ ๋ฌธ์์ข ์ ์ถ๊ฐํ๋ฉด ์ํ๋ ์์น์ ๋ฃ์ ์ ์๊ณ ํญ์ ๋งจ ์๋
ํ์ฌ classify ํ ๋ ๋ฌธ์์ข ๋ชฉ๋ก์ด ์ถ๊ฐํ ์์๋๋ก ๊ณ ์ ๋์ด ์์ด, ์ค๊ฐ์ ์ ๋ฌธ์์ข ์ ์ถ๊ฐํ๋ฉด ์ํ๋ ์์น์ ๋ฃ์ ์ ์๊ณ ํญ์ ๋งจ ์๋์ ๋ถ๋๋ฐ, ๋ฌธ์์ข ํญ๋ชฉ์ ๋๋๊ทธ ์ค ๋๋กญ์ผ๋ก ์์ ๋กญ๊ฒ ์ฌ์ ๋ ฌํ ์ ์์ผ๋ฉด ์ข์ ๊ฒ ๊ฐ์ต๋๋ค!
Extract ๊ฒฐ๊ณผ ์ค์ ์์ ํ ํ์๊ฐ ์๋ ์ ๋ต์ธ ๊ฐ์ gold ๊ฒฐ๊ณผ๋ก ์ผ๊ด ๋ฐ์ํ ์ ์์ผ๋ฉด ์ข์ ๊ฒ ๊ฐ์์!
Extract ๊ฒฐ๊ณผ ์ค์ ์์ ํ ํ์๊ฐ ์๋ ์ ๋ต์ธ ๊ฐ์ gold ๊ฒฐ๊ณผ๋ก ์ผ๊ด ๋ฐ์ํ ์ ์์ผ๋ฉด ์ข์ ๊ฒ ๊ฐ์์! ์๋ฅผ ๋ค์ด ํ์ฌ๋ ์ ๋ต์ด๋๋ผ๋ ๋ ธ๋ ์ฒดํฌํ์(confidence score 1.0)๊ฐ ๋จ๋ฉด ์ผ์ผํ "์ ๋ต์ผ๋ก ํ์"๋ฅผ ํ๋์ฉ ํด๋ฆญํด์ ๋ฐ์ํ๊ณ ์์ต๋๋ค. ํญ๋ชฉ์ด ๋ง์์ง ๊ฒฝ์ฐ ์ผ๊ด ๋ฐ์ (์ฒดํฌ ๋ฐ์ค)ํ ์ ์์ผ๋ฉด ์ด๋จ๊น ์ถ์ด ์๊ฒฌ๋๋ฆฝ๋๋ค :)