Log inSign up
Problems / M54

Sentiment classification from three examples · SmolLM2 360M

500 ptsrawsmollm2-360mCases 16max 16 tokensSet by dongbinna
Added October 11.

Statement

The model has to classify sentiment from three input and output example pairs alone.

InputExpected output
배터리가 생각보다 오래가서 만족합니다.긍정

The clearest fact from the measurements is all there is to this problem. The last answer in the few-shot examples becomes the model's default. In P14 the last answer was 부정, and all 9 wrong cases were 부정. In P35 and P40 the last one was 긍정, and every wrong case was 긍정. It held three times out of three, across the three problems.

Two of the three examples have to carry the same label. Decide which label gets two of them, and what goes last.

Input

  • The input arrives as one line.

  • The input is between 15 and 23 characters long.

Output

  • Print one of 긍정 (positive) or 부정 (negative) on the first line.

  • Do not add any other characters.

  • Not a single character may differ from the expected value.

Requirements

  • The input is one sentence of a product review.
  • The output is either 긍정 or 부정.
  • This is raw mode. Your two pieces and the input are joined together as they are and fed to the model. No chat template and no <think> is added. Grading looks only as far as the first line.
  • This problem does not use a prompt. You submit only three input and output pairs.
  • The frame is fixed. The three examples you submit go in as they are, and the test input is attached after them.
  • Your score comes down to nothing but which examples you pick. You fail on the kinds you did not cover.
  • The output must match the expected value down to the last character. Only leading and trailing whitespace is ignored.
  • This problem is solved with SmolLM2 360M. It is a far smaller model than Qwen3 0.6B, so the same prompt does not simply carry over.
  • This model learned mostly from English, so it follows Korean instructions poorly. You may write your prompt in English. Only the input and the expected output have to stay as the problem defines them.
  • There are 16 cases, and 5 of them show the input together with the expected output. The rest show the input only.
  • Model output is accepted up to 16 tokens.
  • You have to pass all 16 cases to solve this.
Read the requirements carefully before solving. Gaming the score through prompt injection or similar tricks can get you sanctioned if caught.

Examples

The expected output for this problem is in Korean. The grader compares your output against Korean text, character for character. Look at the example cases below to see the exact words it expects. Your prompt itself can be written in English.

5 public cases. Judging also uses hidden cases. You have to pass all 16 cases to solve this.

Input
Expected output
배터리가 생각보다 오래가서 만족합니다.
긍정
이틀 만에 고장났고 문의해도 답이 없네요.
부정
포장도 깔끔하고 배송도 빨랐어요.
긍정
사진이랑 색이 완전히 다릅니다.
부정
가격 생각하면 이 정도면 훌륭하죠.
긍정

Notes

Background that helps with solving. The rules you have to follow are in the input and output specs and the requirements above.

SmolLM2 360M is a fraction of the size of Qwen3 0.6B. It lets instructions slide far more often. You have to say the same thing more briefly and more clearly for it to listen.

Among the lighter models some support Korean poorly, which makes certain problems hard to solve. You may use English in your prompt, so solve it as well as you can.

Three example pairs

SmolLM2 360M

You write no prompt. You submit three examples only. The template is fixed, and the score turns entirely on which examples you pick.

This is the exact text that goes into the model.
입력: 
출력: 

입력: 
출력: 

입력: 
출력: 

입력: <test input>
출력: 
0 chars

All three example pairs have to be filled in before you can grade.

To solve problems you need to log in.