{"id":"W4416035096","doi":"10.18653/v1/2025.findings-emnlp.110","title":"Test-Time Steering for Lossless Text Compression via Weighted Product of Experts","year":2025,"lang":"","type":"article","venue":"","topic":"Algorithms and Data Compression","field":"Computer Science","cited_by":0,"is_retracted":false,"has_abstract":true,"ca_institutions":"","funders":"Natural Sciences and Engineering Research Council of Canada; Alliance de recherche numérique du Canada; Canadian Institute for Advanced Research; Vector Institute; University of British Columbia; Government of Canada","keywords":"Lossless compression; Leverage (statistics); Data compression; Compression (physics); Data compression ratio; Autoregressive model; Product (mathematics)","routes":{"ca_aff":false,"ca_fund":true,"ca_venue":false,"about_ca":false,"invisible_to_affiliation_only":true},"retraction":null,"screen":null,"direct_labels":[],"prediction":{"model_version":"codex-gemma-dda1882f352a","candidate_categories":["metaepi_narrow"],"consensus_categories":[],"category_scores_codex":[0.0004953847,0.0005089637,0.0008132213,0.0003438926,0.0004633346,0.0002755768,0.002014872,0.0001910245,0.0002453486],"category_scores_gemma":[0.0001450746,0.0004131441,0.0002278346,0.0009029053,0.0001803507,0.001062588,0.002179246,0.0001877743,0.00003054171],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.00007401859,"about_ca_system_score_gemma":0.0002705357,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.000101241,"about_ca_topic_score_gemma":0.000002418928,"domain_scores_codex":[0.9961357,0.000106365,0.00109713,0.0013949,0.0005730693,0.0006928229],"domain_scores_gemma":[0.9960888,0.0007688685,0.0003963572,0.001966042,0.0005957476,0.0001841691],"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":[0.0001714681,0.001113998,0.0002309406,0.0006832139,0.00009172495,0.000006786297,0.000524418,0.0001717767,0.2344829,0.003446025,0.02939099,0.7296858],"study_design_scores_gemma":[0.001045393,0.0002122307,0.0005199772,0.001013366,0.00003239407,0.000005305933,0.0000267581,0.6991655,0.280194,0.000684872,0.01672331,0.0003769246],"study_design_candidate":"design_other","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.005855442,0.001825705,0.9851064,0.00106677,0.001836721,0.001541758,0.00008312722,0.0001781815,0.002505893],"genre_scores_gemma":[0.7250403,0.0001599575,0.2530662,0.0004448671,0.0004042257,0.0001466346,0.00005846015,0.0000529914,0.02062634],"genre_candidate":"methods","genre_consensus":null,"teacher_disagreement_score":0.7320402,"threshold_uncertainty_score":0.999832,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.01072432134418797,"score_gpt":0.261135645744455,"score_spread":0.250411324400267,"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."}}