tokyotech-llm

Our Swallow model has undergone continual pre-training from the Llama 3 family, primarily with the addition of Japanese language data. The Instruct versions use supervised fine-tuning (SFT) and Chat Vector. Links to other models can be found in the index. We are excited to share the release schedule for our latest models: - July 1, 2024: Released the Llama-3-Swallow-8B-v0.1, Llama-3-Swallow-8B-Instruct-v0.1, Llama-3-Swallow-70B-v0.1, and Llama-3-Swallow-70B-Instruct-v0.1. |Model|Llama-3-Swallow|Llama3 Swallow Instruct| |---|---|---| |8B| Link | Link | |70B| Link | Link | This repository provides large language models developed by Swallow-LLM. Read our blog post. Model type: Please refer to Llama 3 MODELCARD for details on the model architecture. Language(s): Japanese English Library: Megatron-LM Tokenizer: Please refer to Llama 3 blog for details on the tokenizer. Contact: swallow[at]nlp.c.titech.ac.jp |Model|Size|JCom.|JEMHopQA|NIILC|JSQuAD|XL-Sum|MGSM|WMT20-en-ja|WMT20-ja-en|JMMLU|JHumanEval|Ja Avg| |---|---|---|---|---|---|---|---|---|---|---|---|---| | | |4-shot|4-shot|4-shot|4-shot|1-shot|4-shot|4-shot|4-shot|5-shot|0-shot| | | | |EM acc|Char-F1|Char-F1|Char-F1|ROUGE-2|EM acc|BLEU|BLEU|EM acc|pass@1| | |Llama-2-70b|70B|0.8651|0.5157|0.5464|0.9130|0.2372|0.3640|0.2657|0.2402|0.5496|0.2841|0.4781| |Swallow-70b-hf|70B|0.9178|0.6178|0.6910|0.9208|0.2279|0.4720|0.3046|0.2301|0.5750|0.2262|0.5183| |Qwen2-72B|72B|0.9607|0.6399|0.5617|0.9261|0.2362|0.7560|0.2747|0.2419|0.7831|0.5567|0.5937| |Meta-Llama-3-70B|70B|0.9473|0.6042|0.5965|0.9207|0.2254|0.6720|0.2855|0.2526|0.6975|0.4799|0.5682| |Llama-3-Swallow-70B-v0.1|70B|0.9714|0.6695|0.6881|0.9218|0.2404|0.7080|0.3072|0.2548|0.7049|0.4683|0.5934| |Model|Size|OpenBookQA|TriviaQA|HellaSWAG|SQuAD2.0|XWINO|MMLU|GSM8K|BBH|HumanEval|En Avg| |---|---|---|---|---|---|---|---|---|---|---|---| | | |4-shot|4-shot|4-shot|4-shot|4-shot|5-shot|4-shot|3-shot|0-shot| | | | |Acc|EM acc|Acc|EM acc|Acc|Acc|EM acc|CoT EM Acc|pass@1| | |Llama-2-70b|70B|0.4260|0.7988|0.6681|0.3379|0.9256|0.6876|0.5466|0.6643|0.3152|0.5967| |Swallow-70b-hf|70B|0.4160|0.7610|0.6433|0.3345|0.9191|0.6571|0.5080|0.6537|0.2409|0.5704| |Qwen2-72B|72B|0.4160|0.7890|0.6766|0.4052|0.9161|0.8428|0.8908|0.6388|0.6049|0.6867| |Meta-Llama-3-70B|70B|0.4360|0.8263|0.6909|0.4071|0.9213|0.7870|0.8014|0.8266|0.5177|0.6905| |Llama-3-Swallow-70B-v0.1|70B|0.4240|0.8231|0.6828|0.4059|0.9234|0.7745|0.8143|0.7352|0.4909|0.6749| We used llm-jp-eval(v1.3.0), JP Language Model Evaluation Harness(commit #9b42d41) and Code Generation LM Evaluation Harness(commit #0261c52). The details are as follows: - Multiple-choice question answering (JCommonsenseQA [Kurihara et al., 2022]) - Open-ended question answering (JEMHopQA [Ishii et al., 2024]) - Open-ended question answering (NIILC [関根, 2003]) - Machine reading comprehension (JSQuAD [Kurihara et al., 2022]) - Automatic summarization (XL-Sum [Hasan et al., 2021]) - Machine translation (WMT2020 ja-en [Barrault et al., 2020]) - Machine translation (WMT2020 en-ja [Barrault et al., 2020]) - Mathematical reasoning (MGSM [Shi et al., 2023]) - Academic exams (JMMLU [尹ら, 2024]) - Code generation (JHumanEval [佐藤ら, 2024]) We used the Language Model Evaluation Harness(v.0.4.2) and Code Generation LM Evaluation Harness(commit #0261c52). The details are as follows: - Multiple-choice question answering (OpenBookQA [Mihaylov et al., 2018]) - Open-ended question answering (TriviaQA [Joshi et al., 2017]) - Machine reading comprehension (SQuAD2 [Rajpurkar et al., 2018]) - Commonsense reasoning (XWINO [Tikhonov and Ryabinin, 2021]) - Natural language inference (HellaSwag [Zellers et al., 2019]) - Mathematical reasoning (GSM8K [Cobbe et al., 2021]) - Reasoning (BBH (BIG-Bench-Hard) [Suzgun et al., 2023]) - Academic exams (MMLU [Hendrycks et al., 2021]) - Code generation (HumanEval [Chen et al., 2021]) Continual Pre-Training The following datasets were used for continual pre-training. - Algebraic Stack - Cosmopedia - English Wikipedia - Japanese Wikipedia - Laboro ParaCorpus - OpenWebMath - RefinedWeb - Swallow Corpus The models released here are still in the early stages of our research and development and have not been tuned to ensure outputs align with human intent and safety considerations. We thank Meta Research for releasing Llama 3 under an open license for others to build on. Our project is supported by the Large Generative AI Development Support Program of the National Institute of Advanced Industrial Science and Technology. Here are the team members: - From Tokyo Institute of Technology Okazaki Laboratory, the following members: - Naoaki Okazaki - Sakae Mizuki - Youmi Ma - Koki Maeda - Kakeru Hattori - Masanari Ohi - Taihei Shiotani - Koshiro Saito - From Tokyo Institute of Technology YOKOTA Laboratory, the following members: - Rio Yokota - Kazuki Fujii - Taishi Nakamura - Takumi Okamoto - Ishida Shigeki - From Artificial Intelligence Research Center, AIST, Japan, the following members: - Hiroya Takamura If you find our work helpful, please feel free to cite us.