{"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/arxiv-2605-29843","title":"HARP: Hadamard-Preconditioned Adaptive Rotation Processor for Extreme LLM Quantization","arxiv_id":"2605.29843","date":"2026-05-28","proceeding":null,"authors":["Artur Zagitov","Gleb Molodtsov","Aleksandr Beznosikov"],"abstract":"Post-training quantization (PTQ) is essential for deploying LLMs under memory and bandwidth constraints. However, extreme low-bit quantization remains highly sensitive to activation outliers and anisotropic weight curvature. Existing incoherence-based PTQ methods mitigate this issue with fixed randomized Hadamard transforms (RHTs), which improve quantization robustness but cannot adapt the rotated basis to the layer, calibration distribution, or quantizer. We introduce HARP (Hadamard-preconditioned Adaptive Rotation Processor), a learnable structured two-sided orthogonal processor that replaces fixed Hadamard mixing while preserving exact full-precision equivalence. HARP represents each rotation as a product of sparse butterfly-like block-orthogonal stages, supports non-power-of-two dimensions via Mixed-Radix schedules, and initializes to the RHT processor up to a fixed permutation. Fitted only on calibration data, HARP adapts the quantization basis to each layer and backend. Across 2-4 bit settings on models ranging from 1B to 70B parameters, HARP improves perplexity and zero-shot accuracy over fixed RHT. Importantly, HARP preserves deployment efficiency, reaching 128 tok/s versus 61 tok/s for FP16.","url_abs":"https://arxiv.org/abs/2605.29843","url_pdf":"https://arxiv.org/pdf/2605.29843","source":{"archive":null,"snapshot":"2025-07-28","note":"not in the Papers with Code archive (frozen at the snapshot)","row_kind":"graph","title_abstract_authors_date":"arXiv metadata, CC0 1.0 (https://info.arxiv.org/help/license)"},"code_links":[],"tasks":[],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=2605.29843","mcp":{"get_harvested_code_for_paper":{"arxiv_id":"2605.29843"}},"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":[{"mentioned_in_github":null,"is_official":null,"provenance":"deterministic:regex_extraction","mentioned_in_paper":null,"url":"https://github.com/brain-lab-research/HARP","reach":null}],"summary":{"ran":2,"ran_draft_wrong":1,"ran_honours":3,"ran_violates":1,"ran_fixture":1,"unverified":1},"by_repo_kind":{"found_in_text":{"samples":9,"ran":8,"repositories":1}},"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":9,"samples":[{"code_sha256_prefix":"8b71c304bafb1bc2","entry":"FactorizedOrthogonalRotation","repo":"brain-lab-research/HARP","repo_kind":"found_in_text","path":"harp_qtip/lib/incoherence/harp_lib.py","file_url":"https://github.com/brain-lab-research/HARP/blob/HEAD/harp_qtip/lib/incoherence/harp_lib.py","link_basis":"first_harvest_node","language":"python","status":"ran","verification_level":1,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"NONE","inline_ok":false,"mcp_get_code":{"code_sha256":"8b71c304bafb1bc2"}},{"code_sha256_prefix":"0e1b3f66bcb41ff5","entry":"KronHadKFallbackRotation","repo":"brain-lab-research/HARP","repo_kind":"found_in_text","path":"harp_qtip/lib/incoherence/harp_lib.py","file_url":"https://github.com/brain-lab-research/HARP/blob/HEAD/harp_qtip/lib/incoherence/harp_lib.py","link_basis":"first_harvest_node","language":"python","status":"ran","verification_level":1,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"NONE","inline_ok":false,"mcp_get_code":{"code_sha256":"0e1b3f66bcb41ff5"}},{"code_sha256_prefix":"2bce47bb29615dfd","entry":"_cayley_batch_from_skew","repo":"brain-lab-research/HARP","repo_kind":"found_in_text","path":"harp_qtip/lib/incoherence/harp_lib.py","file_url":"https://github.com/brain-lab-research/HARP/blob/HEAD/harp_qtip/lib/incoherence/harp_lib.py","link_basis":"first_harvest_node","language":"python","status":"ran_draft_wrong","verification_level":1,"contract_check":"MISDECLARED","metamorphic_tier":null,"behaviour_fingerprint":true,"licence":"NONE","inline_ok":false,"mcp_get_code":{"code_sha256":"2bce47bb29615dfd"}},{"code_sha256_prefix":"e5214b7bf0ed8c91","entry":"_factor_schedule_greedy","repo":"brain-lab-research/HARP","repo_kind":"found_in_text","path":"harp_qtip/lib/incoherence/harp_lib.py","file_url":"https://github.com/brain-lab-research/HARP/blob/HEAD/harp_qtip/lib/incoherence/harp_lib.py","link_basis":"first_harvest_node","language":"python","status":"ran_honours","verification_level":1,"contract_check":"HONOURS","metamorphic_tier":null,"behaviour_fingerprint":true,"licence":"NONE","inline_ok":false,"mcp_get_code":{"code_sha256":"e5214b7bf0ed8c91"}},{"code_sha256_prefix":"da68f24b13c88987","entry":"_is_pow2","repo":"brain-lab-research/HARP","repo_kind":"found_in_text","path":"harp_qtip/lib/incoherence/harp_lib.py","file_url":"https://github.com/brain-lab-research/HARP/blob/HEAD/harp_qtip/lib/incoherence/harp_lib.py","link_basis":"first_harvest_node","language":"python","status":"ran_violates","verification_level":1,"contract_check":"VIOLATES","metamorphic_tier":null,"behaviour_fingerprint":true,"licence":"NONE","inline_ok":false,"mcp_get_code":{"code_sha256":"da68f24b13c88987"}},{"code_sha256_prefix":"17d53f501ba7a3dc","entry":"_make_fixed_orthogonal","repo":"brain-lab-research/HARP","repo_kind":"found_in_text","path":"harp_qtip/lib/incoherence/harp_lib.py","file_url":"https://github.com/brain-lab-research/HARP/blob/HEAD/harp_qtip/lib/incoherence/harp_lib.py","link_basis":"first_harvest_node","language":"python","status":"ran_honours","verification_level":1,"contract_check":"HONOURS","metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"NONE","inline_ok":false,"mcp_get_code":{"code_sha256":"17d53f501ba7a3dc"}},{"code_sha256_prefix":"c75d39658d121d69","entry":"_make_sylvester_hadamard","repo":"brain-lab-research/HARP","repo_kind":"found_in_text","path":"harp_qtip/lib/incoherence/harp_lib.py","file_url":"https://github.com/brain-lab-research/HARP/blob/HEAD/harp_qtip/lib/incoherence/harp_lib.py","link_basis":"first_harvest_node","language":"python","status":"ran_honours","verification_level":1,"contract_check":"HONOURS","metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"NONE","inline_ok":false,"mcp_get_code":{"code_sha256":"c75d39658d121d69"}},{"code_sha256_prefix":"4727085cc9c3d4e9","entry":"_skew_from_raw","repo":"brain-lab-research/HARP","repo_kind":"found_in_text","path":"harp_qtip/lib/incoherence/harp_lib.py","file_url":"https://github.com/brain-lab-research/HARP/blob/HEAD/harp_qtip/lib/incoherence/harp_lib.py","link_basis":"first_harvest_node","language":"python","status":"ran_fixture","verification_level":1,"contract_check":"RAISES","metamorphic_tier":null,"behaviour_fingerprint":true,"licence":"NONE","inline_ok":false,"mcp_get_code":{"code_sha256":"4727085cc9c3d4e9"}},{"code_sha256_prefix":"2d464eec8d34d7f0","entry":"HARPRotationPreconditioned","repo":"brain-lab-research/HARP","repo_kind":"found_in_text","path":"harp_qtip/lib/incoherence/harp_lib.py","file_url":"https://github.com/brain-lab-research/HARP/blob/HEAD/harp_qtip/lib/incoherence/harp_lib.py","link_basis":"first_harvest_node","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"NONE","inline_ok":false,"mcp_get_code":{"code_sha256":"2d464eec8d34d7f0"}}]},"arxiv_metadata":{"licence":"arXiv metadata, CC0 1.0 (https://info.arxiv.org/help/license)","fields":["title","abstract","authors","date"],"primary_category":"cs.LG","source":"arxiv_2026.jsonl"},"syntology_extracted_results":null}