{"paper":{"title":"SoScale: Second-Order Scale Search for FP4 Quantization","arxiv_id":"2611.16426","primary_cat":"cs.LG","submitted_at":"2026-11-12T07:36:46+00:00","version_compiled":"v1"},"notice":"Every quoted sentence is the authors' text from the arXiv listing abstract. Labels and probabilities are the outputs of classifier-1.13.0 on the question bank abstract-p1-20260923, exactly as returned. No sentence in this record was written by a language model.","schema":"pith.compiled-record/v0","status":"compiled","abstract":{"source":"arXiv listing abstract v1","sentences":[{"i":0,"p":0.95,"id":"s0","own":false,"role":"background","text":"Low-bit quantization reduces the memory footprint and inference cost of large language models (LLMs), yet accurate FP4 quantization remains challenging.","hedge":{"p":1.0,"label":"asserted"},"own_p":0.05,"main_claim_p":0.03,"quantitative":false,"quantitative_p":0.3,"role_probabilities":{"gap":0.05,"method":0.0,"result":0.0,"background":0.95,"limitation":0.0,"future_work":0.0,"contribution":0.0,"interpretation":0.0}},{"i":1,"p":1.0,"id":"s1","own":false,"role":"background","text":"Practical formats such as MXFP4, NVFP4, and HiF4 pair a 4-bit floating-point payload with block-wise or hierarchical scale codes, making accuracy highly sensitive to how each local distribution is scaled.","hedge":{"p":1.0,"label":"asserted"},"own_p":0.06,"main_claim_p":0.02,"quantitative":false,"quantitative_p":0.47,"role_probabilities":{"gap":0.0,"method":0.0,"result":0.0,"background":1.0,"limitation":0.0,"future_work":0.0,"contribution":0.0,"interpretation":0.0}},{"i":2,"p":0.87,"id":"s2","own":false,"role":"background","text":"Default maximum-absolute-value scaling preserves dynamic range, but can spend scarce FP4 levels on rarely occupied outlier regions and leave dense tensor values poorly resolved.","hedge":{"p":0.98,"label":"asserted"},"own_p":0.08,"main_claim_p":0.02,"quantitative":false,"quantitative_p":0.24,"role_probabilities":{"gap":0.12,"method":0.0,"result":0.0,"background":0.87,"limitation":0.01,"future_work":0.0,"contribution":0.0,"interpretation":0.0}},{"i":3,"p":0.77,"id":"s3","own":false,"role":"gap","text":"Existing scale-search methods minimize local element-wise reconstruction error, optimizing tensor fidelity rather than layer-output fidelity.","hedge":{"p":1.0,"label":"asserted"},"own_p":0.04,"main_claim_p":0.03,"quantitative":false,"quantitative_p":0.3,"role_probabilities":{"gap":0.77,"method":0.0,"result":0.0,"background":0.23,"limitation":0.0,"future_work":0.0,"contribution":0.0,"interpretation":0.0}},{"i":4,"p":0.98,"id":"s4","own":true,"role":"contribution","text":"We propose SoScale, a second-order weight scale-code search method for hierarchical FP4 quantization.","hedge":{"p":0.99,"label":"asserted"},"own_p":0.87,"main_claim_p":0.1,"quantitative":false,"quantitative_p":0.09,"role_probabilities":{"gap":0.0,"method":0.02,"result":0.0,"background":0.0,"limitation":0.0,"future_work":0.0,"contribution":0.98,"interpretation":0.0}},{"i":5,"p":0.97,"id":"s5","own":true,"role":"method","text":"SoScale searches hardware-valid scale codes under a Hessian-aware layer reconstruction objective and couples this search with GPTQ-style error recovery, selecting scales by their impact on layer outputs.","hedge":{"p":1.0,"label":"asserted"},"own_p":0.92,"main_claim_p":0.18,"quantitative":false,"quantitative_p":0.05,"role_probabilities":{"gap":0.0,"method":0.97,"result":0.0,"background":0.0,"limitation":0.0,"future_work":0.0,"contribution":0.03,"interpretation":0.0}},{"i":6,"p":1.0,"id":"s6","own":true,"role":"result","text":"Across Qwen3 models, SoScale consistently reduces downstream quantization degradation under MXFP4, NVFP4, and HiF4.","hedge":{"p":1.0,"label":"asserted"},"own_p":0.94,"main_claim_p":0.79,"quantitative":false,"quantitative_p":0.3,"role_probabilities":{"gap":0.0,"method":0.0,"result":1.0,"background":0.0,"limitation":0.0,"future_work":0.0,"contribution":0.0,"interpretation":0.0}}]},"compiled":{"main_claim":{"p":1.0,"text":"Across Qwen3 models, SoScale consistently reduces downstream quantization degradation under MXFP4, NVFP4, and HiF4.","location":"abstract s6","hedge_strength":{"p":1.0,"label":"asserted"}},"provenance":{"route":{"via":"TypeSafe API","model":"classifier-1.13.0","endpoint":"https://api.typesafe.ai/v1/systemone","provider":"TypeSafe"},"packing":"one request per paper","requests":1,"input_tokens":5602,"cost_usd_reported":0.0},"compiled_at":"2026-11-13","bank_version":"abstract-p1-20260923","model_version":"classifier-1.13.0","abstract_states":{"claims_first":0.07,"states_limitation":0.09,"human_participants":0.02,"controlled_comparison":0.63,"randomized_assignment":0.01,"states_quantitative_result":0.05,"compares_with_prior_results":0.82,"mentions_code_or_data_availability":0.03},"evidential_basis":{"p":0.76,"value":"observation or experiment","probabilities":{"argument":0.01,"literature":0.0,"proof_or_derivation":0.0,"analysis_of_existing_data":0.0,"observation_or_experiment":0.76,"simulation_or_computation":0.23}},"contribution_type":{"p":1.0,"value":"method or tool","probabilities":{"review":0.0,"method_or_tool":1.0,"empirical_study":0.0,"simulation_study":0.0,"theoretical_result":0.0,"dataset_or_benchmark":0.0,"position_or_commentary":0.0}},"contribution_statement":{"p":0.51,"text":"We propose SoScale, a second-order weight scale-code search method for hierarchical FP4 quantization.","location":"abstract s4"}},"coverage":{"tier":"abstract","levels_run":["P abstract"],"not_examined":["full text","figures and tables","references and citing sentences","relations to other papers"],"sources_read":["arXiv listing abstract v1"],"levels_not_run":["S sections","T body sentences","C citations","R cross-paper relations"],"sentences_examined":7,"sentences_in_abstract":7},"measured_accuracy":{"url":"https://pith.science/accuracy","measured_on":"2026-10-02","method":"two independent model families labelled a fresh random sample blind; the figure is how often the published label matches their consensus","sentence_role_agreement":0.92,"main_claim_agreement":0.908,"typed_citation_agreement":0.933,"typed_citation_agreement_95_interval":[0.917,0.945],"review_comments_accurate":0.758,"review_comments_accurate_95_interval":[0.684,0.82],"review_caught_admitted_error_found_or_related":0.448,"review_comments_judged":{"judged_by":["gpt-5.6-luna","grok-4.7"],"judged_against":"Pith's own text copy of each paper: text extracted from the PDF, the first 90,000 characters","measured_on":"2026-10-04","accurate":113,"decided":149,"sampled_reviews":200,"split_by_lane":false,"by_lane":"https://pith.science/accuracy#by-lane"}}}