전체 글 썸네일형 리스트형 LLMs can see and hear without any training https://arxiv.org/abs/2501.18096 LLMs can see and hear without any trainingWe present MILS: Multimodal Iterative LLM Solver, a surprisingly simple, training-free approach, to imbue multimodal capabilities into your favorite LLM. Leveraging their innate ability to perform multi-step reasoning, MILS prompts the LLM to generate candarxiv.org 초록우리는 MILS(Multimodal Iterative LLM Solver)를 소개합니다. MILS는.. 더보기 Reasoning Models Can Be Effective Without Thinking https://arxiv.org/abs/2504.09858 Reasoning Models Can Be Effective Without ThinkingRecent LLMs have significantly improved reasoning capabilities, primarily by including an explicit, lengthy Thinking process as part of generation. In this paper, we question whether this explicit thinking is necessary. Using the state-of-the-art DeepSeek-arxiv.org 초록최근의 대형 언어 모델(LLM)은 명시적이고 긴 사고 과정(Thinking proce.. 더보기 2845. Count of Interesting Subarrays class Solution: def countInterestingSubarrays(self, nums: List[int], modulo: int, k: int) -> int: count_map = defaultdict(int) print(count_map) count_map[0] = 1 # 시작점 prefix = 0 answer = 0 for num in nums: # 현재 원소가 interesting 조건 만족하는가? if num % modulo == k: prefix += 1 # prefix[i] % modulo .. 더보기 Antidistillation Sampling https://arxiv.org/abs/2504.13146 Antidistillation SamplingFrontier models that generate extended reasoning traces inadvertently produce rich token sequences that can facilitate model distillation. Recognizing this vulnerability, model owners may seek sampling strategies that limit the effectiveness of distillatioarxiv.org 초록고도화된 프런티어 모델(frontier models)은 확장된 추론 과정을 생성하는 과정에서, 모델 증류(distillation.. 더보기 BitNet b1.58 2B4T Technical Report https://github.com/microsoft/BitNet GitHub - microsoft/BitNet: Official inference framework for 1-bit LLMsOfficial inference framework for 1-bit LLMs. Contribute to microsoft/BitNet development by creating an account on GitHub.github.com https://huggingface.co/microsoft/bitnet-b1.58-2B-4T microsoft/bitnet-b1.58-2B-4T · Hugging FaceBitNet b1.58 2B4T - Scaling Native 1-bit LLM This repository cont.. 더보기 Gaussian Mixture Flow Matching Models https://github.com/Lakonik/GMFlow?_bhlid=81f89db39632970b22a6d8f7641c5518cad90ef0 GitHub - Lakonik/GMFlow: Gaussian Mixture Flow Matching Models (GMFlow)Gaussian Mixture Flow Matching Models (GMFlow). Contribute to Lakonik/GMFlow development by creating an account on GitHub.github.com https://www.arxiv.org/abs/2504.05304?_bhlid=4fcaeeb727d7a8544f84a8ad986fa9ea922fee09 Gaussian Mixture Flow Match.. 더보기 Hogwild! Inference: Parallel LLM Generation via Concurrent Attention https://github.com/eqimp/hogwild_llm?_bhlid=d0e43178b4d426d6b68fdb2f5c141720a923cacc GitHub - eqimp/hogwild_llm: Official PyTorch implementation for Hogwild! Inference: Parallel LLM Generation with a Concurrent AtOfficial PyTorch implementation for Hogwild! Inference: Parallel LLM Generation with a Concurrent Attention Cache - eqimp/hogwild_llmgithub.com https://www.arxiv.org/abs/2504.06261?_bhl.. 더보기 DEIM: DETR with Improved Matching for Fast Convergence https://github.com/ShihuaHuang95/DEIM GitHub - ShihuaHuang95/DEIM: [CVPR 2025] DEIM: DETR with Improved Matching for Fast Convergence[CVPR 2025] DEIM: DETR with Improved Matching for Fast Convergence - ShihuaHuang95/DEIMgithub.com 초록(Abstract)우리는 실시간 객체 탐지를 위한 Transformer 기반 아키텍처(DETR)의 수렴 속도를 가속화하기 위해 고안된 혁신적이고 효율적인 학습 프레임워크 DEIM을 소개한다. DETR 모델의 1:1 매칭(one‑to‑one, O2O)에서 발생하는 희소(supervision) 문제.. 더보기 이전 1 ··· 4 5 6 7 8 9 10 ··· 72 다음
