{"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/simplifying-models-with-unlabeled-output-data","title":"Composed Fine-Tuning: Freezing Pre-Trained Denoising Autoencoders for Improved Generalization","arxiv_id":"2006.16205","date":"2020-06-29","proceeding":null,"authors":["Sang Michael Xie","Tengyu Ma","Percy Liang"],"abstract":"We focus on prediction problems with structured outputs that are subject to output validity constraints, e.g. pseudocode-to-code translation where the code must compile. While labeled input-output pairs are expensive to obtain, \"unlabeled\" outputs, i.e. outputs without corresponding inputs, are freely available (e.g. code on GitHub) and provide information about output validity. We can capture the output structure by pre-training a denoiser to denoise corrupted versions of unlabeled outputs. We first show that standard fine-tuning after pre-training destroys some of this structure. We then propose composed fine-tuning, which fine-tunes a predictor composed with the pre-trained denoiser, which is frozen to preserve output structure. For two-layer ReLU networks, we prove that composed fine-tuning significantly reduces the complexity of the predictor, thus improving generalization. Empirically, we show that composed fine-tuning improves over standard fine-tuning on two pseudocode-to-code translation datasets (3% and 6% relative). The improvement from composed fine-tuning is magnified on out-of-distribution (OOD) examples (4% and 25% relative).","url_abs":"https://arxiv.org/abs/2006.16205v4","url_pdf":"https://arxiv.org/pdf/2006.16205v4.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":"simplifying-models-with-unlabeled-output-data","repo_url":"https://github.com/p-lambda/composed_finetuning","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":1,"framework":"pytorch","reach":null},{"paper_slug":"simplifying-models-with-unlabeled-output-data","repo_url":"https://github.com/p-lambda/unlabeled_outputs","is_official":1,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"ok","spdx":"MIT"}}],"tasks":[{"task_slug":"code-translation","task_name":"Code Translation"},{"task_slug":"denoising","task_name":"Denoising"},{"task_slug":"image-generation","task_name":"Image Generation"},{"task_slug":"translation","task_name":"Translation"}],"methods":[{"method_slug":"absolute-position-encodings","method_name":"Absolute Position Encodings"},{"method_slug":"adam","method_name":"Adam"},{"method_slug":"attention","method_name":"Attention"},{"method_slug":"bpe","method_name":"BPE"},{"method_slug":"dense-connections","method_name":"Dense Connections"},{"method_slug":"dropout","method_name":"Dropout"},{"method_slug":"label-smoothing","method_name":"Label Smoothing"},{"method_slug":"layer-normalization","method_name":"Layer Normalization"},{"method_slug":"linear-layer","method_name":"Linear Layer"},{"method_slug":"multi-head-attention","method_name":"Multi-Head Attention"},{"method_slug":"position-wise-feed-forward-layer","method_name":"Position-Wise Feed-Forward Layer"},{"method_slug":"relu","method_name":"ReLU"},{"method_slug":"residual-connection","method_name":"Residual Connection"},{"method_slug":"softmax","method_name":"Softmax"},{"method_slug":"transformer","method_name":"Transformer"}],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=2006.16205","mcp":{"get_harvested_code_for_paper":{"arxiv_id":"2006.16205"}},"developers":"https://syntology.ai/developers","read_at":"2026-09-24T18:15:14+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/p-lambda/unlabeled_outputs","reach":{"status":"ok","spdx":"MIT"}},{"provenance":"external:paperswithcode_snapshot_2025-07-28","url":"https://github.com/p-lambda/composed_finetuning","reach":null}],"summary":{"ran_draft_wrong":2,"unverified":12},"by_repo_kind":{"official":{"samples":14,"ran":2,"repositories":2}},"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":"c60712bd3a0d1e33","entry":"concat_code","repo":"p-lambda/composed_finetuning","repo_kind":"official","path":"code/evaluate_full.py","file_url":"https://github.com/p-lambda/composed_finetuning/blob/HEAD/code/evaluate_full.py","link_basis":"first_harvest_node","language":"python","status":"ran_draft_wrong","verification_level":1,"contract_check":"OUTPUT_MISDECLARED","metamorphic_tier":"deterministic","behaviour_fingerprint":false,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"c60712bd3a0d1e33"}},{"code_sha256_prefix":"60931bcb2604c46d","entry":"parse_preds","repo":"p-lambda/composed_finetuning","repo_kind":"official","path":"code/evaluate_full.py","file_url":"https://github.com/p-lambda/composed_finetuning/blob/HEAD/code/evaluate_full.py","link_basis":"first_harvest_node","language":"python","status":"ran_draft_wrong","verification_level":1,"contract_check":"OUTPUT_MISDECLARED","metamorphic_tier":"deterministic","behaviour_fingerprint":false,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"60931bcb2604c46d"}},{"code_sha256_prefix":"e22284ad0b102063","entry":"add_newlines","repo":"p-lambda/unlabeled_outputs","repo_kind":"official","path":"code/pretty_print.py","file_url":"https://github.com/p-lambda/unlabeled_outputs/blob/HEAD/code/pretty_print.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":"e22284ad0b102063"}},{"code_sha256_prefix":"49d63b429e4a4074","entry":"concat_code","repo":"p-lambda/unlabeled_outputs","repo_kind":"official","path":"code/run_spoc.py","file_url":"https://github.com/p-lambda/unlabeled_outputs/blob/HEAD/code/run_spoc.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":"49d63b429e4a4074"}},{"code_sha256_prefix":"459ab85d861ca156","entry":"concat_text","repo":"p-lambda/unlabeled_outputs","repo_kind":"official","path":"code/run_spoc.py","file_url":"https://github.com/p-lambda/unlabeled_outputs/blob/HEAD/code/run_spoc.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":"459ab85d861ca156"}},{"code_sha256_prefix":"465db07d479dc049","entry":"encode","repo":"p-lambda/unlabeled_outputs","repo_kind":"official","path":"code/run_synthetic.py","file_url":"https://github.com/p-lambda/unlabeled_outputs/blob/HEAD/code/run_synthetic.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":"465db07d479dc049"}},{"code_sha256_prefix":"a0d03dc853adfbd0","entry":"fix_char_string_tok","repo":"p-lambda/unlabeled_outputs","repo_kind":"official","path":"code/format_pairs.py","file_url":"https://github.com/p-lambda/unlabeled_outputs/blob/HEAD/code/format_pairs.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":"a0d03dc853adfbd0"}},{"code_sha256_prefix":"cd0dc7e8ddbce005","entry":"get_lens","repo":"p-lambda/unlabeled_outputs","repo_kind":"official","path":"code/fairseq_modules/criterion.py","file_url":"https://github.com/p-lambda/unlabeled_outputs/blob/HEAD/code/fairseq_modules/criterion.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":"cd0dc7e8ddbce005"}},{"code_sha256_prefix":"245dba3efa0406c1","entry":"pad","repo":"p-lambda/unlabeled_outputs","repo_kind":"official","path":"code/fairseq_modules/criterion.py","file_url":"https://github.com/p-lambda/unlabeled_outputs/blob/HEAD/code/fairseq_modules/criterion.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":"245dba3efa0406c1"}},{"code_sha256_prefix":"b9c973b50d3ba0a3","entry":"pad_before","repo":"p-lambda/unlabeled_outputs","repo_kind":"official","path":"code/fairseq_modules/criterion.py","file_url":"https://github.com/p-lambda/unlabeled_outputs/blob/HEAD/code/fairseq_modules/criterion.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":"b9c973b50d3ba0a3"}},{"code_sha256_prefix":"8f8d509260110fb3","entry":"parse_pred_file","repo":"p-lambda/unlabeled_outputs","repo_kind":"official","path":"code/run_synthetic.py","file_url":"https://github.com/p-lambda/unlabeled_outputs/blob/HEAD/code/run_synthetic.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":"8f8d509260110fb3"}},{"code_sha256_prefix":"18e78e1f6ef94f00","entry":"strip_ext","repo":"p-lambda/unlabeled_outputs","repo_kind":"official","path":"code/run_spoc.py","file_url":"https://github.com/p-lambda/unlabeled_outputs/blob/HEAD/code/run_spoc.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":"18e78e1f6ef94f00"}},{"code_sha256_prefix":"f4e1ee0459a1e883","entry":"tokenize_code","repo":"p-lambda/unlabeled_outputs","repo_kind":"official","path":"code/format_pairs.py","file_url":"https://github.com/p-lambda/unlabeled_outputs/blob/HEAD/code/format_pairs.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":"f4e1ee0459a1e883"}},{"code_sha256_prefix":"6708dda8d2ddd583","entry":"tokenize_text","repo":"p-lambda/unlabeled_outputs","repo_kind":"official","path":"code/format_pairs.py","file_url":"https://github.com/p-lambda/unlabeled_outputs/blob/HEAD/code/format_pairs.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":"6708dda8d2ddd583"}}]},"arxiv_metadata":null,"syntology_extracted_results":null}