{"paper":{"title":"SignRot: LLM Quantization with Massive Outlier-Aware Sign-Adjusted Rotation","arxiv_id":"2611.16502","primary_cat":"cs.AI","submitted_at":"2026-11-12T09:05:14+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":1.0,"id":"s0","own":false,"role":"background","text":"While quantization compresses LLMs and accelerates inference, outlier activations impede efficient low-bit representation.","hedge":{"p":1.0,"label":"asserted"},"own_p":0.06,"main_claim_p":0.03,"quantitative":false,"quantitative_p":0.04,"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":1,"p":1.0,"id":"s1","own":false,"role":"background","text":"In particular, massive outliers cause substantial performance degradation.","hedge":{"p":1.0,"label":"asserted"},"own_p":0.13,"main_claim_p":0.03,"quantitative":false,"quantitative_p":0.12,"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":0.84,"id":"s2","own":false,"role":"background","text":"Although rotation-based methods that rely on randomized Hadamard transforms for online inference aim to mitigate this, they can produce fragmented distributions with empty bins when a massive outlier dominates a token's activation.","hedge":{"p":0.57,"label":"hedged_possibility"},"own_p":0.07,"main_claim_p":0.03,"quantitative":false,"quantitative_p":0.06,"role_probabilities":{"gap":0.12,"method":0.0,"result":0.0,"background":0.84,"limitation":0.04,"future_work":0.0,"contribution":0.0,"interpretation":0.0}},{"i":3,"p":0.88,"id":"s3","own":false,"role":"background","text":"This fragmentation severely undermines uniform quantization performance.","hedge":{"p":1.0,"label":"asserted"},"own_p":0.13,"main_claim_p":0.04,"quantitative":false,"quantitative_p":0.04,"role_probabilities":{"gap":0.03,"method":0.0,"result":0.0,"background":0.88,"limitation":0.01,"future_work":0.0,"contribution":0.0,"interpretation":0.08}},{"i":4,"p":0.91,"id":"s4","own":false,"role":"background","text":"Moreover, massive outliers concentrate in the down-projection input, where the preceding nonlinearity requires the rotation to be applied at inference time.","hedge":{"p":1.0,"label":"asserted"},"own_p":0.31,"main_claim_p":0.03,"quantitative":false,"quantitative_p":0.06,"role_probabilities":{"gap":0.07,"method":0.0,"result":0.01,"background":0.91,"limitation":0.01,"future_work":0.0,"contribution":0.0,"interpretation":0.0}},{"i":5,"p":0.68,"id":"s5","own":false,"role":"background","text":"This constraint prevents training-based approaches such as SpinQuant from learning the online rotation at this position, as a learned dense matrix cannot use the Fast Walsh-Hadamard Transform and would require a full matrix multiplication during inference.","hedge":{"p":1.0,"label":"asserted"},"own_p":0.14,"main_claim_p":0.03,"quantitative":false,"quantitative_p":0.09,"role_probabilities":{"gap":0.29,"method":0.0,"result":0.0,"background":0.68,"limitation":0.03,"future_work":0.0,"contribution":0.0,"interpretation":0.0}},{"i":6,"p":0.65,"id":"s6","own":true,"role":"contribution","text":"To address this, we propose SignRot (Sign-adjusted Rotation), which identifies massive outlier channels and constructs a sign adjustment vector that unifies the signs of the corresponding Hadamard row, collapsing the bimodal distribution into a unimodal profile amenable to quantization.","hedge":{"p":1.0,"label":"asserted"},"own_p":0.92,"main_claim_p":0.43,"quantitative":false,"quantitative_p":0.05,"role_probabilities":{"gap":0.0,"method":0.35,"result":0.0,"background":0.0,"limitation":0.0,"future_work":0.0,"contribution":0.65,"interpretation":0.0}},{"i":7,"p":0.91,"id":"s7","own":true,"role":"method","text":"Since the aligned outlier shifts the activation mean away from zero, we pair SignRot with asymmetric activation quantization, whose zero-point absorbs this shift.","hedge":{"p":1.0,"label":"asserted"},"own_p":0.73,"main_claim_p":0.06,"quantitative":false,"quantitative_p":0.04,"role_probabilities":{"gap":0.0,"method":0.91,"result":0.0,"background":0.0,"limitation":0.0,"future_work":0.0,"contribution":0.09,"interpretation":0.0}},{"i":8,"p":0.59,"id":"s8","own":true,"role":"contribution","text":"As a lightweight, drop-in improvement compatible with any Hadamard-based quantization method, SignRot requires only a single calibration forward pass and adds only element-wise sign flips at inference, without an additional matrix multiplication.","hedge":{"p":1.0,"label":"asserted"},"own_p":0.79,"main_claim_p":0.13,"quantitative":true,"quantitative_p":0.52,"role_probabilities":{"gap":0.0,"method":0.41,"result":0.0,"background":0.0,"limitation":0.0,"future_work":0.0,"contribution":0.59,"interpretation":0.0}},{"i":9,"p":1.0,"id":"s9","own":true,"role":"result","text":"Applying SignRot to QuaRot achieves a WikiText2 perplexity of 5.88 and an average zero-shot reasoning accuracy of 72.73, surpassing SpinQuant on the LLaMA3-70B model.","hedge":{"p":1.0,"label":"asserted"},"own_p":0.94,"main_claim_p":0.53,"quantitative":true,"quantitative_p":0.99,"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}},{"i":10,"p":0.57,"id":"s10","own":false,"role":"contribution","text":"Code is available at https://github.com/HYU-ARK-Lab/SignRot.","hedge":{"p":0.88,"label":"asserted"},"own_p":0.44,"main_claim_p":0.02,"quantitative":false,"quantitative_p":0.04,"role_probabilities":{"gap":0.0,"method":0.08,"result":0.07,"background":0.04,"limitation":0.0,"future_work":0.24,"contribution":0.57,"interpretation":0.0}}]},"compiled":{"main_claim":{"p":0.89,"text":"Applying SignRot to QuaRot achieves a WikiText2 perplexity of 5.88 and an average zero-shot reasoning accuracy of 72.73, surpassing SpinQuant on the LLaMA3-70B model.","location":"abstract s9","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":8356,"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":0.05,"states_limitation":0.19,"human_participants":0.02,"controlled_comparison":0.85,"randomized_assignment":0.03,"states_quantitative_result":0.99,"compares_with_prior_results":0.98,"mentions_code_or_data_availability":0.99},"evidential_basis":{"p":0.83,"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":0.83,"simulation_or_computation":0.16}},"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":1.0,"text":"To address this, we propose SignRot (Sign-adjusted Rotation), which identifies massive outlier channels and constructs a sign adjustment vector that unifies the signs of the corresponding Hadamard row, collapsing the bimodal distribution into a unimodal profile amenable to quantization.","location":"abstract s6"}},"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":11,"sentences_in_abstract":11},"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"}}}