{"about":{"site":"https://codewithpapers.app","non_affiliation":"Code with Papers and Syntology are not affiliated with, endorsed by, or sponsored by Papers with Code, Meta, or the pwc-archive mirror.","licence":"CC BY-SA 4.0","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","attribution":"https://codewithpapers.app/attribution","modified":"archive material modified by Syntology; see the attribution page"},"url":"/paper/llamax-scaling-linguistic-horizons-of-llm-by","title":"LLaMAX: Scaling Linguistic Horizons of LLM by Enhancing Translation Capabilities Beyond 100 Languages","arxiv_id":"2407.05975","date":"2024-07-08","proceeding":null,"authors":["Yinquan Lu","Wenhao Zhu","Lei LI","Yu Qiao","Fei Yuan"],"abstract":"Large Language Models (LLMs) demonstrate remarkable translation capabilities in high-resource language tasks, yet their performance in low-resource languages is hindered by insufficient multilingual data during pre-training. To address this, we conduct extensive multilingual continual pre-training on the LLaMA series models, enabling translation support across more than 100 languages. Through a comprehensive analysis of training strategies, such as vocabulary expansion and data augmentation, we develop LLaMAX. Remarkably, without sacrificing its generalization ability, LLaMAX achieves significantly higher translation performance compared to existing open-source LLMs (by more than 10 spBLEU points) and performs on-par with specialized translation model (M2M-100-12B) on the Flores-101 benchmark. Extensive experiments indicate that LLaMAX can serve as a robust multilingual foundation model. The code \\footnote{\\url{https://github.com/CONE-MT/LLaMAX/.}} and the models \\footnote{\\url{https://huggingface.co/LLaMAX/.}} are publicly available.","url_abs":"https://arxiv.org/abs/2407.05975v2","url_pdf":"https://arxiv.org/pdf/2407.05975v2.pdf","source":{"archive":"pwc-archive (Hugging Face), CC BY-SA 4.0","snapshot":"2025-07-28","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","row_kind":"abstracts"},"code_links":[{"paper_slug":"llamax-scaling-linguistic-horizons-of-llm-by","repo_url":"https://github.com/cone-mt/llamax","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"ok","spdx":"MIT"}}],"tasks":[{"task_slug":"data-augmentation","task_name":"Data Augmentation"},{"task_slug":"translation","task_name":"Translation"}],"methods":[{"method_slug":"llama","method_name":"LLaMA"}],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"syntology_url":"https://syntology.ai/paper/2407.05975","atlas_url":"https://app.syntology.ai/?focus=2407.05975","mcp":{"get_harvested_code_for_paper":{"arxiv_id":"2407.05975"}},"developers":"https://syntology.ai/developers","read_at":"2026-09-25T09:33:49+00:00","read_at_is":"when the build read Syntology's graph, not when any sample ran","claim":"Per-sample execution status on synthesized fixtures; not a correctness claim about the paper. Samples come from repositories linked to the paper, official or community; repo_kind says which.","repos":[{"provenance":"external:paperswithcode_snapshot_2025-07-28","url":"https://github.com/cone-mt/llamax","reach":{"status":"ok","spdx":"MIT"}}],"summary":{"ran":6,"unverified":2},"by_repo_kind":{"official":{"samples":8,"ran":6,"repositories":1}},"repo_kind_vocabulary":{"official":"The archive marks this repository official for the paper","named_in_paper":"The archive records that the paper mentions this repository; it is not marked official","listed":"In the archive's code links for this paper, not marked official and not recorded as mentioned in the paper","found_in_text":"Syntology found this repository in the paper's own text; whether it is the authors' implementation is not asserted","community":"Not in the archive's code links for this paper; a community repository Syntology harvested"},"n_pointer_only_for_licence":0,"samples":[{"code_sha256_prefix":"39ddb531258d53d7","entry":"calculate_cos_smi","repo":"cone-mt/llamax","repo_kind":"official","path":"paper_related_code/analysis/embedding_quality.py","file_url":"https://github.com/cone-mt/llamax/blob/HEAD/paper_related_code/analysis/embedding_quality.py","link_basis":"first_harvest_node","language":"python","status":"ran","verification_level":1,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":true,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"39ddb531258d53d7"}},{"code_sha256_prefix":"1106f81d300ec332","entry":"calculate_ks","repo":"cone-mt/llamax","repo_kind":"official","path":"paper_related_code/paper_results/tabe1_extend_vocab.py","file_url":"https://github.com/cone-mt/llamax/blob/HEAD/paper_related_code/paper_results/tabe1_extend_vocab.py","link_basis":"first_harvest_node","language":"python","status":"ran","verification_level":1,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"1106f81d300ec332"}},{"code_sha256_prefix":"9c8be06eb82485a5","entry":"calculate_rank","repo":"cone-mt/llamax","repo_kind":"official","path":"paper_related_code/analysis/embedding_quality.py","file_url":"https://github.com/cone-mt/llamax/blob/HEAD/paper_related_code/analysis/embedding_quality.py","link_basis":"first_harvest_node","language":"python","status":"ran","verification_level":1,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"9c8be06eb82485a5"}},{"code_sha256_prefix":"0e6fb8e4442b8db6","entry":"calculate_statistics_info","repo":"cone-mt/llamax","repo_kind":"official","path":"paper_related_code/paper_results/lg_detect_ratio.py","file_url":"https://github.com/cone-mt/llamax/blob/HEAD/paper_related_code/paper_results/lg_detect_ratio.py","link_basis":"first_harvest_node","language":"python","status":"ran","verification_level":1,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"0e6fb8e4442b8db6"}},{"code_sha256_prefix":"2884a24501c48df9","entry":"get_tensor_dict","repo":"cone-mt/llamax","repo_kind":"official","path":"paper_related_code/analysis/embedding_quality.py","file_url":"https://github.com/cone-mt/llamax/blob/HEAD/paper_related_code/analysis/embedding_quality.py","link_basis":"first_harvest_node","language":"python","status":"ran","verification_level":1,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"2884a24501c48df9"}},{"code_sha256_prefix":"941f73935455dba6","entry":"get_tokenizer","repo":"cone-mt/llamax","repo_kind":"official","path":"paper_related_code/analysis/common.py","file_url":"https://github.com/cone-mt/llamax/blob/HEAD/paper_related_code/analysis/common.py","link_basis":"first_harvest_node","language":"python","status":"ran","verification_level":1,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"941f73935455dba6"}},{"code_sha256_prefix":"25e8bd9dfa18c9c2","entry":"get_tokenized_df","repo":"cone-mt/llamax","repo_kind":"official","path":"paper_related_code/analysis/fertility.py","file_url":"https://github.com/cone-mt/llamax/blob/HEAD/paper_related_code/analysis/fertility.py","link_basis":"first_harvest_node","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"25e8bd9dfa18c9c2"}},{"code_sha256_prefix":"68c4e427a8b8ad53","entry":"load_input_embed","repo":"cone-mt/llamax","repo_kind":"official","path":"paper_related_code/paper_results/tabe1_extend_vocab.py","file_url":"https://github.com/cone-mt/llamax/blob/HEAD/paper_related_code/paper_results/tabe1_extend_vocab.py","link_basis":"first_harvest_node","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"68c4e427a8b8ad53"}}]},"arxiv_metadata":null,"syntology_extracted_results":null}