Blog
232개 중 1-12번째 포스트
자연어 처리의 미래를 여는 열쇠: 휴먼-인-더-루프(Human-in-the-Loop)
자연언어처리(Natural Language Processing, NLP)는 인공지능(AI) 분야에서 가장 빠르게 발전하는 영역 중 하나입니다. 특히 거대 언어 모델(Large Language Models, LLM)의 등장은 인간의 언어를 이해하고 생성하는 AI의 능력을 전례 없는 수준으로 끌어올렸습니다. 하지만 이러한 눈부신 발전에도 불구하고, AI는 여전...
![[논문 리뷰] The Last Human-Written Paper: Agent-Native Research Artifacts](/assets/images/blog/20260504-paper-2604-24658-the-last-human-written-paper-a.jpg)
[논문 리뷰] The Last Human-Written Paper: Agent-Native Research Artifacts
Scientific publication compresses a branching, iterative research process into a linear narrative, discarding the majority of what was discovered along the way. This compilation imposes two structural...
![[논문 리뷰] The Latent Space: Foundation, Evolution, Mechanism, Ability, and Outlook](/assets/images/blog/20260504-paper-2604-02029-the-latent-space-foundation-ev.jpg)
[논문 리뷰] The Latent Space: Foundation, Evolution, Mechanism, Ability, and Outlook
Latent space is rapidly emerging as a native substrate for language-based models. While modern systems are still commonly understood through explicit token-level generation, an increasing body of work...
![[논문 리뷰] You Need Better Attention Priors](/assets/images/blog/20260504-paper-2601-15380-you-need-better-attention-prio.jpg)
[논문 리뷰] You Need Better Attention Priors
We generalize the attention mechanism by viewing it through the lens of Entropic Optimal Transport, revealing that standard attention corresponds to a transport problem regularized by an implicit unif...
![[논문 리뷰] Learning to Orchestrate Agents in Natural Language with the Conductor](/assets/images/blog/20260504-paper-2512-04388-learning-to-orchestrate-agents.jpg)
[논문 리뷰] Learning to Orchestrate Agents in Natural Language with the Conductor
Powerful large language models (LLMs) from different providers have been expensively trained and finetuned to specialize across varying domains. In this work, we introduce a new kind of Conductor mode...
![[논문 리뷰] Contextual Agentic Memory is a Memo, Not True Memory](/assets/images/blog/20260503-paper-2604-27707-contextual-agentic-memory-is-a.jpg)
[논문 리뷰] Contextual Agentic Memory is a Memo, Not True Memory
Current agentic memory systems (vector stores, retrieval-augmented generation, scratchpads, and context-window management) do not implement memory: they implement lookup. We argue that treating lookup...
![[논문 리뷰] Hyperloop Transformers](/assets/images/blog/20260503-paper-2604-21254-hyperloop-transformers.jpg)
[논문 리뷰] Hyperloop Transformers
LLM architecture research generally aims to maximize model quality subject to fixed compute/latency budgets. However, many applications of interest such as edge and on-device deployment are further co...
![[논문 리뷰] Neural Garbage Collection: Learning to Forget while Learning to Reason](/assets/images/blog/20260503-paper-2604-18002-neural-garbage-collection-lear.jpg)
[논문 리뷰] Neural Garbage Collection: Learning to Forget while Learning to Reason
Chain-of-thought reasoning has driven striking advances in language model capability, yet every reasoning step grows the KV cache, creating a bottleneck to scaling this paradigm further. Current appro...
![[논문 리뷰] Graph-of-Agents: A Graph-based Framework for Multi-Agent LLM Collaboration](/assets/images/blog/20260503-paper-2604-17148-graph-of-agents-a-graph-based-.jpg)
[논문 리뷰] Graph-of-Agents: A Graph-based Framework for Multi-Agent LLM Collaboration
With an ever-growing zoo of LLMs and benchmarks, the need to orchestrate multiple models for improved task performance has never been more pressing. While frameworks like Mixture-of-Agents (MoA) attem...
![[논문 리뷰] Scaling Test-Time Compute for Agentic Coding](/assets/images/blog/20260503-paper-2604-16529-scaling-test-time-compute-for-.jpg)
[논문 리뷰] Scaling Test-Time Compute for Agentic Coding
Test-time scaling has become a powerful way to improve large language models. However, existing methods are best suited to short, bounded outputs that can be directly compared, ranked or refined. Long...
![[논문 리뷰] LLMs Corrupt Your Documents When You Delegate](/assets/images/blog/20260503-paper-2604-15597-llms-corrupt-your-documents-wh.jpg)
[논문 리뷰] LLMs Corrupt Your Documents When You Delegate
Large Language Models (LLMs) are poised to disrupt knowledge work, with the emergence of delegated work as a new interaction paradigm (e.g., vibe coding). Delegation requires trust - the expectation t...
![[논문 리뷰] Attention to Mamba: A Recipe for Cross-Architecture Distillation](/assets/images/blog/20260503-paper-2604-14191-attention-to-mamba-a-recipe-fo.jpg)
[논문 리뷰] Attention to Mamba: A Recipe for Cross-Architecture Distillation
State Space Models (SSMs) such as Mamba have become a popular alternative to Transformer models, due to their reduced memory consumption and higher throughput at generation compared to their Attention...