Agent K v1.0: A Large Language Model for Automated Data Science Achieving Kaggle Grandmaster Level Performance
This paper introduces Agent K v1.0, an autonomous data science agent powered by large language models (LLMs) that achieves Kaggle Grandmaster-level performance by leveraging a novel structured reasoning framework and learning from experience.
MetRex: Can Large Language Models Reasonably Estimate Verilog Code Metrics After Synthesis?
While demonstrating promise, LLMs are still under development in accurately predicting post-synthesis metrics (area, delay, static power) of Verilog designs, as shown by the MetRex benchmark and SFT experiments.
모델 불일치 확률을 통한 spurious correlation 완화
본 논문에서는 bias label 없이 spurious correlation을 완화하는 새로운 학습 방법인 DPR(Disagreement Probability based Resampling)을 제안합니다. DPR은 bias model의 예측과 실제 label 간의 불일치 확률을 활용하여 bias-conflicting sample을 식별하고 upsampling하여 모델의 spurious correlation에 대한 의존도를 줄입니다.
온라인 LLM 추론을 위한 CPU 오프로딩으로 GPU 메모리 위기를 해결하는 NEO
온라인 LLM 추론 시스템 NEO는 GPU 메모리 부족 문제를 해결하기 위해 어텐션 연산 및 KV 캐시를 CPU로 오프로드하여 처리량을 향상시키고 GPU 사용률을 극대화합니다.
Mobility-LLM: Using Large Language Models to Understand Human Travel Intentions and Preferences from Location Data
Mobility-LLM is a novel framework that leverages the power of large language models (LLMs) to analyze check-in sequences from location-based services, enabling a deeper understanding of human visiting intentions and travel preferences.
대규모 언어 모델은 시계열 데이터의 제로샷 이상 탐지에 효과적인가?
대규모 언어 모델(LLM)은 시계열 데이터에서 이상을 탐지하는 데 활용될 수 있지만, 아직 최첨단 딥러닝 모델보다 성능이 떨어진다.
Large Language Models for Time Series Anomaly Detection: A Preliminary Study
Large language models (LLMs) show promise for zero-shot time series anomaly detection, particularly when leveraging their forecasting capabilities, but they still lag behind state-of-the-art deep learning models in performance.
PatternBoost: Using Transformers and Local Search to Find Interesting Constructions in Extremal Combinatorics
PatternBoost, a novel algorithm combining transformer neural networks with local search methods, effectively discovers intricate patterns in mathematical constructions, particularly in extremal combinatorics, often surpassing the performance of traditional approaches and leading to new discoveries.
Stanford Researchers Develop Highly Efficient LLMs Using LoLCATs Method for Under $20
Stanford researchers have developed a new method called LoLCATs that significantly reduces the computational requirements and cost of training large language models (LLMs) while maintaining comparable performance to state-of-the-art models.
Apple AI Researchers Find Large Language Models Lack Basic Reasoning Skills
Despite the hype surrounding AI, Apple researchers have found that even the most advanced large language models (LLMs) struggle with basic reasoning, suggesting that the technology is not as revolutionary as claimed.
© 2024 by Linnk AI