<aside> <img src="/icons/light-bulb_gray.svg" alt="/icons/light-bulb_gray.svg" width="40px" />

선정 기준:

  1. 특정 도메인에 방법론 적용 및 응용 지양
  2. 다양한 방법론 지향
  3. Top Conference의 자연어처리 분야
  4. 최근 3년 이내의 논문 </aside>

1. LLM Pre-training & Open Models

Name Conference Notes Citation 발표자
OLMo: Accelerating the Science of Language Models ACL 2024 Best Theme Paper 396 조승환 2026-01-27
Not all tokens are what you need for pretraining NeurIPS 2024 Best Paper Runner-up 137

2. Reasoning & Chain-of-Thought

Name Conference Notes Citation 발표자
Iteration Head: A Mechanistic Study of Chain-of-Thought NeurIPS 2024 Poster 22 박재현
2026-01-13

3. Alignment & RLHF

Name Conference Notes Citation 발표자
Is dpo superior to ppo for llm alignment? a comprehensive study ICML 2024 Oral 233 서성원
2026-01-13
Safety Alignment Should be Made More Than Just a Few Tokens Deep ICLR 2025 Oral 231

4. Interpretability & Mechanistic Analysis

Name Conference Notes Citation 발표자
Steering Llama 2 via Contrastive Activation Addition ACL 2024 Outstanding Paper 359 석진실
2026-01-20
Amortizing intractable inference in large language models ICLR 2024 Outstanding Paper 97

5. Evaluation

Name Conference Notes Citation 발표자
Proving Test Set Contamination in Black-Box Language Models ICLR 2024 Outstanding Paper 207
Political Compass or Spinning Arrow? Towards More Meaningful Evaluations for Values and Opinions in Large Language Models ACL 2024 Outstanding Paper 170 양유진1/20

6. Efficiency & Inference

Name Conference Notes Citation 발표자
Model Tells You What to Discard: Adaptive KV Cache Compression for LLMs ICLR 2024 Outstanding Paper 368
Faster Cascades via Speculative Decoding ICLR 2025 Outstanding Paper 22

7. Multilingual

Name Conference Notes Citation 발표자
Aya Model: An Instruction Finetuned Open-Access Multilingual Language Model ACL 2024 Best Paper 305