{"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/arxiv-2510-06195","title":"Latent Speech-Text Transformer","arxiv_id":"2510.06195","date":"2025-10-07","proceeding":"ICLR","authors":["Yen-Ju Lu","Yashesh Gaur","Wei Zhou","Benjamin Muller","Jesus Villalba","Najim Dehak","Luke Zettlemoyer","Gargi Ghosh","Mike Lewis","Srinivasan Iyer","Duc Le"],"abstract":"Auto-regressive speech-text models pre-trained on interleaved text tokens and discretized speech tokens demonstrate strong speech understanding and generation, yet remain substantially less compute-efficient than text LLMs, partly due to the much longer sequences of speech tokens relative to text. This modality imbalance disproportionately allocates pre-training and inference compute to speech, potentially hindering effective cross-modal alignment and slowing performance scaling by orders of magnitude. We introduce the Latent Speech-Text Transformer (LST), which aggregates speech tokens into latent speech patches that serve as higher-level autoregressive units. This design aligns the sequence-modeling granularity between speech and text while improving computational efficiency. The resulting patches can align with textual units to facilitate cross-modal knowledge transfer and compactly capture recurring acoustic patterns such as silence. Across story-completion benchmarks under both compute-controlled and data-controlled settings, LST consistently improves speech accuracy while also improving text performance, achieving up to +6.5% absolute gain on speech HellaSwag in compute-controlled training (+5.3% in data-controlled training). Under compute-controlled scaling from 420M to 1.8B parameters in a near compute-optimal regime, gains grow with scale, and improvements persist up to 7B parameters under fixed-token budgets. These benefits extend to downstream tasks: LST stabilizes ASR adaptation and reduces the effective autoregressive sequence length during ASR and TTS inference, lowering computational cost without degrading reconstruction quality. The code is available at https://github.com/facebookresearch/lst.","url_abs":"https://arxiv.org/abs/2510.06195","url_pdf":"https://arxiv.org/pdf/2510.06195","source":{"archive":null,"snapshot":"2025-07-28","note":"not in the Papers with Code archive (frozen at the snapshot)","row_kind":"graph","title_abstract_authors_date":"arXiv metadata, CC0 1.0 (https://info.arxiv.org/help/license)"},"code_links":[],"tasks":[],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"syntology_url":"https://syntology.ai/paper/2510.06195","atlas_url":"https://app.syntology.ai/?focus=2510.06195","mcp":{"get_harvested_code_for_paper":{"arxiv_id":"2510.06195"}},"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":[{"mentioned_in_github":null,"is_official":null,"provenance":"deterministic:regex_extraction","mentioned_in_paper":null,"url":"https://github.com/facebookresearch/lst","reach":{"status":"ok","spdx":"NOASSERTION"}}],"summary":{"unverified":18},"by_repo_kind":{"found_in_text":{"samples":18,"ran":0,"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":18,"samples":[{"code_sha256_prefix":"f20a2b559a7b8775","entry":"consolidate_checkpoints","repo":"facebookresearch/lst","repo_kind":"found_in_text","path":"lst/checkpoint.py","file_url":"https://github.com/facebookresearch/lst/blob/HEAD/lst/checkpoint.py","link_basis":"first_harvest_node","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"NOASSERTION","inline_ok":false,"mcp_get_code":{"code_sha256":"f20a2b559a7b8775"}},{"code_sha256_prefix":"1d34daf1ab8748db","entry":"cross_entropy","repo":"facebookresearch/lst","repo_kind":"found_in_text","path":"lst/base_transformer.py","file_url":"https://github.com/facebookresearch/lst/blob/HEAD/lst/base_transformer.py","link_basis":"first_harvest_node","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"NOASSERTION","inline_ok":false,"mcp_get_code":{"code_sha256":"1d34daf1ab8748db"}},{"code_sha256_prefix":"75d20eeaeb19679a","entry":"dist_max","repo":"facebookresearch/lst","repo_kind":"found_in_text","path":"lst/distributed.py","file_url":"https://github.com/facebookresearch/lst/blob/HEAD/lst/distributed.py","link_basis":"first_harvest_node","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"NOASSERTION","inline_ok":false,"mcp_get_code":{"code_sha256":"75d20eeaeb19679a"}},{"code_sha256_prefix":"f766bed8cc55159b","entry":"dist_min","repo":"facebookresearch/lst","repo_kind":"found_in_text","path":"lst/distributed.py","file_url":"https://github.com/facebookresearch/lst/blob/HEAD/lst/distributed.py","link_basis":"first_harvest_node","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"NOASSERTION","inline_ok":false,"mcp_get_code":{"code_sha256":"f766bed8cc55159b"}},{"code_sha256_prefix":"ebf689175ca08206","entry":"get_device_mesh","repo":"facebookresearch/lst","repo_kind":"found_in_text","path":"lst/distributed.py","file_url":"https://github.com/facebookresearch/lst/blob/HEAD/lst/distributed.py","link_basis":"first_harvest_node","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"NOASSERTION","inline_ok":false,"mcp_get_code":{"code_sha256":"ebf689175ca08206"}},{"code_sha256_prefix":"81d8582d1fab7bda","entry":"get_num_params","repo":"facebookresearch/lst","repo_kind":"found_in_text","path":"lst/metrics.py","file_url":"https://github.com/facebookresearch/lst/blob/HEAD/lst/metrics.py","link_basis":"first_harvest_node","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"NOASSERTION","inline_ok":false,"mcp_get_code":{"code_sha256":"81d8582d1fab7bda"}},{"code_sha256_prefix":"1d8b97adda18ea42","entry":"get_rng_state","repo":"facebookresearch/lst","repo_kind":"found_in_text","path":"lst/args.py","file_url":"https://github.com/facebookresearch/lst/blob/HEAD/lst/args.py","link_basis":"first_harvest_node","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"NOASSERTION","inline_ok":false,"mcp_get_code":{"code_sha256":"1d8b97adda18ea42"}},{"code_sha256_prefix":"30e68c9cadfddd32","entry":"matmul","repo":"facebookresearch/lst","repo_kind":"found_in_text","path":"lst/float8.py","file_url":"https://github.com/facebookresearch/lst/blob/HEAD/lst/float8.py","link_basis":"first_harvest_node","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"NOASSERTION","inline_ok":false,"mcp_get_code":{"code_sha256":"30e68c9cadfddd32"}},{"code_sha256_prefix":"aaa4b37e564944fe","entry":"named_replace","repo":"facebookresearch/lst","repo_kind":"found_in_text","path":"lst/float8.py","file_url":"https://github.com/facebookresearch/lst/blob/HEAD/lst/float8.py","link_basis":"first_harvest_node","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"NOASSERTION","inline_ok":false,"mcp_get_code":{"code_sha256":"aaa4b37e564944fe"}},{"code_sha256_prefix":"20629d61b9f81088","entry":"parse_args_to_pydantic_model","repo":"facebookresearch/lst","repo_kind":"found_in_text","path":"lst/config_parser.py","file_url":"https://github.com/facebookresearch/lst/blob/HEAD/lst/config_parser.py","link_basis":"first_harvest_node","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"NOASSERTION","inline_ok":false,"mcp_get_code":{"code_sha256":"20629d61b9f81088"}},{"code_sha256_prefix":"8751ebfdda7f9c83","entry":"parse_file_config","repo":"facebookresearch/lst","repo_kind":"found_in_text","path":"lst/config_parser.py","file_url":"https://github.com/facebookresearch/lst/blob/HEAD/lst/config_parser.py","link_basis":"first_harvest_node","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"NOASSERTION","inline_ok":false,"mcp_get_code":{"code_sha256":"8751ebfdda7f9c83"}},{"code_sha256_prefix":"1775c1f6cd5bd79c","entry":"precompute_freqs_cis","repo":"facebookresearch/lst","repo_kind":"found_in_text","path":"lst/base_transformer.py","file_url":"https://github.com/facebookresearch/lst/blob/HEAD/lst/base_transformer.py","link_basis":"first_harvest_node","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"NOASSERTION","inline_ok":false,"mcp_get_code":{"code_sha256":"1775c1f6cd5bd79c"}},{"code_sha256_prefix":"ff8e64af8f92c325","entry":"recursively_parse_config","repo":"facebookresearch/lst","repo_kind":"found_in_text","path":"lst/config_parser.py","file_url":"https://github.com/facebookresearch/lst/blob/HEAD/lst/config_parser.py","link_basis":"first_harvest_node","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"NOASSERTION","inline_ok":false,"mcp_get_code":{"code_sha256":"ff8e64af8f92c325"}},{"code_sha256_prefix":"705c70c44ddc8080","entry":"repeat_kv","repo":"facebookresearch/lst","repo_kind":"found_in_text","path":"lst/base_transformer.py","file_url":"https://github.com/facebookresearch/lst/blob/HEAD/lst/base_transformer.py","link_basis":"first_harvest_node","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"NOASSERTION","inline_ok":false,"mcp_get_code":{"code_sha256":"705c70c44ddc8080"}},{"code_sha256_prefix":"294621d07f3c12f0","entry":"sample_tokens","repo":"facebookresearch/lst","repo_kind":"found_in_text","path":"lst/generate.py","file_url":"https://github.com/facebookresearch/lst/blob/HEAD/lst/generate.py","link_basis":"first_harvest_node","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"NOASSERTION","inline_ok":false,"mcp_get_code":{"code_sha256":"294621d07f3c12f0"}},{"code_sha256_prefix":"5b1313e8dd2cbcfa","entry":"sample_top_k","repo":"facebookresearch/lst","repo_kind":"found_in_text","path":"lst/generate.py","file_url":"https://github.com/facebookresearch/lst/blob/HEAD/lst/generate.py","link_basis":"first_harvest_node","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"NOASSERTION","inline_ok":false,"mcp_get_code":{"code_sha256":"5b1313e8dd2cbcfa"}},{"code_sha256_prefix":"2f2141ad0c4f3dbd","entry":"sample_top_p","repo":"facebookresearch/lst","repo_kind":"found_in_text","path":"lst/generate.py","file_url":"https://github.com/facebookresearch/lst/blob/HEAD/lst/generate.py","link_basis":"first_harvest_node","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"NOASSERTION","inline_ok":false,"mcp_get_code":{"code_sha256":"2f2141ad0c4f3dbd"}},{"code_sha256_prefix":"2848341791f9a450","entry":"scale","repo":"facebookresearch/lst","repo_kind":"found_in_text","path":"lst/float8.py","file_url":"https://github.com/facebookresearch/lst/blob/HEAD/lst/float8.py","link_basis":"first_harvest_node","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"NOASSERTION","inline_ok":false,"mcp_get_code":{"code_sha256":"2848341791f9a450"}}]},"arxiv_metadata":{"licence":"arXiv metadata, CC0 1.0 (https://info.arxiv.org/help/license)","fields":["title","abstract","authors","date"],"primary_category":"cs.CL","source":"arxiv_api"},"syntology_extracted_results":null}