%0 Generic %1 Hugo2023 %A Touvron, Hugo %A Martin, Louis %A Stone, Kevin R. %A Albert, Peter %A Almahairi, Amjad %A Babaei, Yasmine %A Bashlykov, Nikolay %A Batra, Soumya %A Bhargava, Prajjwal %A Bhosale, Shruti %A Bikel, D. %A Blecher, Lukas %A Ferrer, Cristian Cantón %A Chen, Moya %A Cucurull, Guillem %A Esiobu, David %A Fernandes, Jude %A Fu, Jeremy %A Fu, Wenyin %A Fuller, Brian %A Gao, Cynthia %A Goswami, Vedanuj %A Goyal, Naman %A Hartshorn, A. %A Hosseini, Saghar %A Hou, Rui %A Inan, Hakan %A Kardas, Marcin %A Kerkez, Viktor %A Khabsa, Madian %A Kloumann, Isabel M. %A Korenev, A. %A Koura, Punit Singh %A Lachaux, Marie-Anne %A Lavril, Thibaut %A Lee, Jenya %A Liskovich, Diana %A Lu, Yinghai %A Mao, Yuning %A Martinet, Xavier %A Mihaylov, Todor %A Mishra, Pushkar %A Molybog, Igor %A Nie, Yixin %A Poulton, Andrew %A Reizenstein, Jeremy %A Rungta, Rashi %A Saladi, Kalyan %A Schelten, Alan %A Silva, Ruan %A Smith, Eric Michael %A Subramanian, R. %A Tan, Xia %A Tang, Binh %A Taylor, Ross %A Williams, Adina %A Kuan, Jian Xiang %A Xu, Puxin %A Yan, Zhengxu %A Zarov, Iliyan %A Zhang, Yuchen %A Fan, Angela %A Kambadur, Melanie %A Narang, Sharan %A Rodriguez, Aurelien %A Stojnic, Robert %A Edunov, Sergey %A Scialom, Thomas %D 2023 %J ArXiv %K Llama_2 language_models chat_models open-source fine-tuning paper_demo posted_with_chatgpt %R 10.48550/arXiv.2307.09288 %T Llama 2: Open Foundation and Fine-Tuned Chat Models %U https://www.semanticscholar.org/paper/104b0bb1da562d53cbda87aec79ef6a2827d191a %V abs/2307.09288 %X In this work, we develop and release Llama 2, a collection of pretrained and fine-tuned large language models (LLMs) ranging in scale from 7 billion to 70 billion parameters. Our fine-tuned LLMs, called Llama 2-Chat, are optimized for dialogue use cases. Our models outperform open-source chat models on most benchmarks we tested, and based on our human evaluations for helpfulness and safety, may be a suitable substitute for closed-source models. We provide a detailed description of our approach to fine-tuning and safety improvements of Llama 2-Chat in order to enable the community to build on our work and contribute to the responsible development of LLMs.

@JournalArticle{Hugo2023, abstract = {In this work, we develop and release Llama 2, a collection of pretrained and fine-tuned large language models (LLMs) ranging in scale from 7 billion to 70 billion parameters. Our fine-tuned LLMs, called Llama 2-Chat, are optimized for dialogue use cases. Our models outperform open-source chat models on most benchmarks we tested, and based on our human evaluations for helpfulness and safety, may be a suitable substitute for closed-source models. We provide a detailed description of our approach to fine-tuning and safety improvements of Llama 2-Chat in order to enable the community to build on our work and contribute to the responsible development of LLMs.}, added-at = {2023-10-23T15:55:34.000+0200}, author = {Touvron, Hugo and Martin, Louis and Stone, Kevin R. and Albert, Peter and Almahairi, Amjad and Babaei, Yasmine and Bashlykov, Nikolay and Batra, Soumya and Bhargava, Prajjwal and Bhosale, Shruti and Bikel, D. and Blecher, Lukas and Ferrer, Cristian Cantón and Chen, Moya and Cucurull, Guillem and Esiobu, David and Fernandes, Jude and Fu, Jeremy and Fu, Wenyin and Fuller, Brian and Gao, Cynthia and Goswami, Vedanuj and Goyal, Naman and Hartshorn, A. and Hosseini, Saghar and Hou, Rui and Inan, Hakan and Kardas, Marcin and Kerkez, Viktor and Khabsa, Madian and Kloumann, Isabel M. and Korenev, A. and Koura, Punit Singh and Lachaux, Marie-Anne and Lavril, Thibaut and Lee, Jenya and Liskovich, Diana and Lu, Yinghai and Mao, Yuning and Martinet, Xavier and Mihaylov, Todor and Mishra, Pushkar and Molybog, Igor and Nie, Yixin and Poulton, Andrew and Reizenstein, Jeremy and Rungta, Rashi and Saladi, Kalyan and Schelten, Alan and Silva, Ruan and Smith, Eric Michael and Subramanian, R. and Tan, Xia and Tang, Binh and Taylor, Ross and Williams, Adina and Kuan, Jian Xiang and Xu, Puxin and Yan, Zhengxu and Zarov, Iliyan and Zhang, Yuchen and Fan, Angela and Kambadur, Melanie and Narang, Sharan and Rodriguez, Aurelien and Stojnic, Robert and Edunov, Sergey and Scialom, Thomas}, biburl = {https://www.bibsonomy.org/bibtex/206a8dc9369c86d7e271bab8153ea0a78/tomvoelker}, day = 18, description = {The Llama 2 paper introduces a collection of pretrained and fine-tuned large language models optimized for dialogue use cases. It provides insights into the fine-tuning and safety improvements of Llama 2-Chat and its potential as a substitute for closed-source models.}, doi = {10.48550/arXiv.2307.09288}, interhash = {78179489055d55f304786dc0a5dcdb2d}, intrahash = {06a8dc9369c86d7e271bab8153ea0a78}, journal = {ArXiv}, keywords = {Llama_2 language_models chat_models open-source fine-tuning paper_demo posted_with_chatgpt}, month = {7}, timestamp = {2023-10-23T15:55:34.000+0200}, title = {Llama 2: Open Foundation and Fine-Tuned Chat Models}, url = {https://www.semanticscholar.org/paper/104b0bb1da562d53cbda87aec79ef6a2827d191a}, volume = {abs/2307.09288}, year = 2023 }