按年被提到
20242024 · 2 次2026
大模型与语言模型›大模型预训练与缩放·arXiv:2406.11794 · cs.LG

DataComp-LM: In search of the next generation of training sets for language models

Jeffrey Li, Alex Fang, Georgios Smyrnis, Maor Ivgi, Matt Jordan, Samir Gadre, Hritik Bansal, Etash Guha, Sedrick Keh, Kushal Arora, Saurabh Garg, Rui Xin 等 59 人 · 2024
开始阅读 ↗PDFalphaXiv 讨论
被提及 · 2 人首次 2024-06-25 · 最近 2024-12-15
2不同的人
2活跃月份
2讨论串
46被引(OpenAlex)
arXiv 摘要(候选层不写导读)

We introduce DataComp for Language Models (DCLM), a testbed for controlled dataset experiments with the goal of improving language models. As part of DCLM, we provide a standardized corpus of 240T tokens extracted from Common Crawl, effective pretraining recipes based on the OpenLM framework, and a broad suite of 53 downstream evaluations. Participants in the DCLM benchmark can experiment with data curation strategies such as deduplication, filtering, and data mixing at model scales ranging from 412M to 7B parameters. As a baseline for DCLM, we conduct extensive experiments and find that model-based filtering is key to assembling a high-quality training set. The resulting dataset, DCLM-Baseline enables training a 7B parameter language model from scratch to 64% 5-shot accuracy on MMLU with 2.6T training tokens. Compared to MAP-Neo, the previous state-of-the-art in open-data language models, DCLM-Baseline represents a 6.6 percentage point improvement on MMLU while being trained with 40% less compute. Our baseline model is also comparable to Mistral-7B-v0.3 and Llama 3 8B on MMLU (63% & 66%), and performs similarly on an average of 53 natural language understanding tasks while being trained with 6.6x less compute than Llama 3 8B. Our results highlight the importance of dataset design for training language models and offer a starting point for further research on data curation.

候选层:至少 2 个不同的人、在 2 个不同月份提到过,还没到主书架的 3 人 3 个月。没有做立场标注,所以只写“被提及”。

HN 书评

推荐在前,同类按时间 · 每篇最多保留 20 条,统计用全部提及
评论m_ke · 2024-12

“There's a ton of recent work on data curation / synthetic data generation that shows that smaller high quality datasets go a lot further than scaling up on noisy web data. The scaling law plots are log scale so to get more juice with naive scaling we'd need to invest exponentially more resources, and we're at a point where the juice is not worth the sque…”

Ask HN: Why is Ilya saying data is limited when the whole world is data?原评论 ↗
被提交为 story · 1 次