{"id":"W4415898436","doi":"10.48550/arxiv.2503.06211","title":"Late Fusion and Multi-Level Fission Amplify Cross-Modal Transfer in Text-Speech LMs","year":2025,"lang":"en","type":"preprint","venue":"arXiv (Cornell University)","topic":"Speech Recognition and Synthesis","field":"Computer Science","cited_by":0,"is_retracted":false,"has_abstract":true,"ca_institutions":"","funders":"","keywords":"Principle of compositionality; Feature (linguistics); Representation (politics); Process (computing); Hierarchy; Modality (human–computer interaction); Transfer (computing)","routes":{"ca_aff":true,"ca_fund":false,"ca_venue":false,"about_ca":false,"invisible_to_affiliation_only":false},"retraction":null,"screen":null,"direct_labels":[],"prediction":{"model_version":"metacan-v3-hybrid-931329e0061c","candidate_categories":[],"consensus_categories":[],"category_scores_codex":[0.001788181,0.0009593377,0.0006999316,0.0003593423,0.000474931,0.001013454,0.001349411,0.001153483,0.004272381],"category_scores_gemma":[0.004273078,0.0004162314,0.0008055381,0.0002872627,0.0008554693,0.002949312,0.003247894,0.002684749,0.002135323],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0006683051,"about_ca_system_score_gemma":0.00078564,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.002103448,"about_ca_topic_score_gemma":0.003118917,"domain_scores_codex":[0.9994822,0.0001417132,0.00002915588,0.0001615874,0.0001093508,0.00007595548],"domain_scores_gemma":[0.9987845,0.000703005,0.00006026712,0.0001898096,0.0001780199,0.00008446231],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"design_other","study_design_gemma":"simulation_or_modeling","study_design_scores_codex":[0.001150585,0.000361496,0.001865305,0.0001995617,0.0001362954,0.0002963371,0.0005624305,0.4045938,0.1094581,0.01622463,0.004036549,0.461115],"study_design_scores_gemma":[0.00001730847,0.00009768721,0.0002960512,0.00001133878,0.00001786713,0.00003336594,0.00003587934,0.9705865,0.02236841,0.00553399,0.0009884969,0.00001312262],"study_design_candidate":"simulation_or_modeling","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.1093162,0.0004289751,0.8782693,0.0004872065,0.0001630574,0.00007047768,0.000143294,0.005699337,0.00542218],"genre_scores_gemma":[0.860638,0.0001546869,0.1277719,0.0003170994,0.00006839351,0.000116173,0.0004439929,0.000470296,0.01001935],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.004272381,"threshold_uncertainty_score":0.01429248,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.1290470316641667,"score_gpt":0.2367942662393476,"score_spread":0.1077472345751809,"validation_status":"score_only:v0-immature-baseline","note":"Baseline scores from an immature model (maturity gate not passed). Scores rank; they never assert a category."}}