Trends · 研究趋势
数据来自 Hugging Face 论文的 AI 提取关键词,按月统计研究方向的增长与热度。
An Yang, Anfeng Li, Baosong Yang +57 authors
Qwen3, a unified series of large language models, integrates thinking and non-thinking modes, reduces computational resources, and achieves state-of-the-art performance across various tasks and languages.
Albert Q. Jiang, Alexandre Sablayrolles, Antoine Roux +23 authors
Mixtral 8x7B, a Sparse Mixture of Experts language model, achieves superior performance across benchmarks by using a selective architecture that leverages fewer active parameters.
Binyuan Hui, Jian Yang, Zeyu Cui +14 authors
Qwen2.5-Coder series demonstrates state-of-the-art code generation, completion, reasoning, and repair capabilities using the Qwen2.5 architecture with over 5.5 trillion tokens of training data.
Jiaru Zou, Xiyuan Yang, Ruizhong Qiu +10 authors
LatentMAS enables efficient, lossless collaboration among LLM agents in latent space, improving performance and reducing computational costs compared to text-based methods.
Siming Huang, Tianhao Cheng, Jason Klein Liu +16 authors
OpenCoder is a top-tier open-access code LLM that provides comprehensive reproducible resources, including data, pipelines, and protocols, to advance code AI research.
Kevin Qinghong Lin, Yuhao Zheng, Hangyu Ran +5 authors
VCode introduces a benchmark for generating SVG code from images to preserve symbolic meaning, highlighting gaps in visual-centric coding and proposing VCoder to improve performance.
Tianyu Zheng, Ge Zhang, Tianhao Shen +5 authors
OpenCodeInterpreter, an open-source system for generating, executing, and refining code, achieves high performance on benchmarks through execution and human feedback, reducing the gap with proprietary systems like GPT-4 Code Interpreter.
Daya Guo, Qihao Zhu, Dejian Yang +10 authors
DeepSeek-Coder, a series of open-source code models trained from scratch, outperforms both open-source and closed-source models across multiple benchmarks.
Xu Huang, Wenhao Zhu, Hanxu Hu +4 authors
BenchMAX evaluates multilingual large language models across advanced capabilities like instruction following, reasoning, and code generation, revealing language-specific performance gaps.
Jian Yang, Jiaxi Yang, Ke Jin +7 authors
A human-curated benchmark (CodeArena) and a large synthetic instruction corpus (SynCode-Instruct) are introduced to evaluate code LLMs based on human preference alignment, revealing performance differences between open-source and proprietary models.
Shihan Dou, Yan Liu, Haoxiang Jia +13 authors
StepCoder, a reinforcement learning framework, enhances code generation by breaking tasks into subtasks and optimizing only executed code segments, using the APPS+ dataset for training.
Fengji Zhang, Linquan Wu, Huiyu Bai +6 authors
HumanEval-V provides a comprehensive benchmark for evaluating LMMs' diagram interpretation and reasoning abilities, particularly in coding contexts, revealing significant areas for improvement.
Bo Shen, Jiaxin Zhang, Taihong Chen +9 authors
PanGu-Coder2, a model fine-tuned with RRTF framework, demonstrates superior code generation performance, outperforming existing Code LLMs on multiple benchmarks.
Linzheng Chai, Shukai Liu, Jian Yang +15 authors
A multilingual code benchmark covering 40 programming languages with 16K test samples is introduced to advance code language model research, along with a multilingual coder model and instruction corpora.
Rachit Bansal, Bidisha Samanta, Siddharth Dalmia +6 authors
CALM, a composition method for large language models, enhances capabilities by integrating cross-attention with specialized models, demonstrating improvements on translation and coding tasks.
Guillaume Sanchez, Honglu Fan, Alexander Spangher +3 authors
Classifier-Free Guidance enhances performance across various language modeling tasks and improves the faithfulness and coherence of AI assistants, outperforming models with higher parameter counts.
Lingjiao Chen, Matei Zaharia, James Zou
The performance and behavior of GPT-3.5 and GPT-4 fluctuated significantly between March and June 2023 across various tasks, emphasizing the necessity for ongoing LLM quality monitoring.
Aojun Zhou, Ke Wang, Zimu Lu +8 authors
A novel code-based self-verification method significantly improves math reasoning in GPT-4 Code Interpreter, enhancing accuracy on the MATH dataset.
Theo X. Olausson, Jeevana Priya Inala, Chenglong Wang +2 authors
GPT-4 demonstrates superior self-repair capabilities on APPS dataset compared to GPT-3.5, with performance boosted when feedback is provided by GPT-4 or human programmers.
Yue Wang, Hung Le, Akhilesh Deepak Gotmare +3 authors
CodeT5+, a family of flexible encoder-decoder LLMs initialized with frozen LLMs and trained with a mixture of pretraining objectives, achieves state-of-the-art performance across various code-related tasks.
北京市昌平区探索星信息技术及软件开发工作室
京ICP备2026059466号