{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,6,3]],"date-time":"2026-06-03T18:26:09Z","timestamp":1780511169917,"version":"3.54.1"},"reference-count":29,"publisher":"IEEE","license":[{"start":{"date-parts":[[2026,1,19]],"date-time":"2026-01-19T00:00:00Z","timestamp":1768780800000},"content-version":"stm-asf","delay-in-days":0,"URL":"https:\/\/doi.org\/10.15223\/policy-029"},{"start":{"date-parts":[[2026,1,19]],"date-time":"2026-01-19T00:00:00Z","timestamp":1768780800000},"content-version":"stm-asf","delay-in-days":0,"URL":"https:\/\/doi.org\/10.15223\/policy-037"}],"content-domain":{"domain":[],"crossmark-restriction":false},"short-container-title":[],"published-print":{"date-parts":[[2026,1,19]]},"DOI":"10.1109\/asp-dac66049.2026.11420289","type":"proceedings-article","created":{"date-parts":[[2026,3,10]],"date-time":"2026-03-10T19:51:15Z","timestamp":1773172275000},"page":"1449-1455","source":"Crossref","is-referenced-by-count":1,"title":["Platinum: Path-Adaptable LUT-Based Accelerator Tailored for Low-Bit Weight Matrix Multiplication"],"prefix":"10.1109","author":[{"given":"Haoxuan","family":"Shan","sequence":"first","affiliation":[{"name":"Duke University,Department of Electrical and Computer Engineering"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Cong","family":"Guo","sequence":"additional","affiliation":[{"name":"Duke University,Department of Electrical and Computer Engineering"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Chiyue","family":"Wei","sequence":"additional","affiliation":[{"name":"Duke University,Department of Electrical and Computer Engineering"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Feng","family":"Cheng","sequence":"additional","affiliation":[{"name":"Duke University,Department of Electrical and Computer Engineering"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Junyao","family":"Zhang","sequence":"additional","affiliation":[{"name":"Duke University,Department of Electrical and Computer Engineering"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Hai Helen","family":"Li","sequence":"additional","affiliation":[{"name":"Duke University,Department of Electrical and Computer Engineering"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Yiran","family":"Chen","sequence":"additional","affiliation":[{"name":"Duke University,Department of Electrical and Computer Engineering"}],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"263","reference":[{"key":"ref1","article-title":"Llama: Open and efficient foundation language models","volume":"abs\/2302.13971","author":"Touvron","year":"2023","journal-title":"CoRR"},{"key":"ref2","article-title":"Llama 2: Open foundation and fine-tuned chat models","volume":"abs\/2307.09288","author":"Touvron","year":"2023","journal-title":"CoRR"},{"key":"ref3","article-title":"Megatron-lm: Training multi-billion parameter language models using model parallelism","volume":"abs\/1909.08053","author":"Shoeybi","year":"2019","journal-title":"CoRR"},{"key":"ref4","doi-asserted-by":"publisher","DOI":"10.1109\/micro56248.2022.00095"},{"key":"ref5","doi-asserted-by":"publisher","DOI":"10.1145\/3579371.3589038"},{"key":"ref6","article-title":"FP8-LM: training FP8 large language models","volume":"abs\/2310.18313","author":"Peng","year":"2023","journal-title":"CoRR"},{"key":"ref7","doi-asserted-by":"publisher","DOI":"10.1145\/3695053.3731024"},{"key":"ref8","doi-asserted-by":"publisher","DOI":"10.1109\/MCAS.2024.3476008"},{"key":"ref9","doi-asserted-by":"publisher","DOI":"10.23919\/AISE.2025.000002"},{"key":"ref10","article-title":"AWQ: activation-aware weight quantization for LLM compression and acceleration","volume":"abs\/2306.00978","author":"Lin","year":"2023","journal-title":"CoRR"},{"key":"ref11","article-title":"GPTQ: accurate post-training quantization for generative pre-trained transformers","volume":"abs\/2210.17323","author":"Frantar","year":"2022","journal-title":"CoRR"},{"key":"ref12","article-title":"Bitnet: Scaling 1-bit transformers for large language models","volume":"abs\/2310.11453","author":"Wang","year":"2023","journal-title":"CoRR"},{"key":"ref13","article-title":"The era of 1-bit 11ms: All large language models are in 1.58 bits","volume":"abs\/2402.17764","author":"Ma","year":"2024","journal-title":"CoRR"},{"key":"ref14","article-title":"TMAC: CPU renaissance via table lookup for low-bit LLM deployment on edge","volume":"abs\/2407.00088","author":"Wei","year":"2024","journal-title":"CoRR"},{"key":"ref15","doi-asserted-by":"publisher","DOI":"10.1109\/hpca61900.2025.00084"},{"key":"ref16","doi-asserted-by":"publisher","DOI":"10.1109\/MICRO.2016.7783722"},{"key":"ref17","doi-asserted-by":"publisher","DOI":"10.1145\/3695053.3731043"},{"key":"ref18","doi-asserted-by":"publisher","DOI":"10.1109\/sc41405.2020.00099"},{"key":"ref19","article-title":"LUT tensor core: Lookup table enables efficient low-bit LLM inference acceleration","volume":"abs\/2408.06003","author":"Mo","year":"2024","journal-title":"CoRR"},{"key":"ref20","article-title":"LUT-GEMM: quantized matrix multiplication based on luts for efficient inference in largescale generative language models","volume-title":"The Twelfth International Conference on Learning Representations, ICLR 2024","author":"Park"},{"key":"ref21","article-title":"Bitnet.cpp: Efficient edge inference for ternary 11 ms","volume":"abs\/2502.11880","author":"Wang","year":"2025","journal-title":"CoRR"},{"key":"ref22","doi-asserted-by":"publisher","DOI":"10.1109\/ISSCC49661.2025.10904761"},{"key":"ref23","doi-asserted-by":"publisher","DOI":"10.1145\/3695053.3731035"},{"key":"ref24","doi-asserted-by":"publisher","DOI":"10.1109\/HPCA61900.2025.00066"},{"key":"ref25","doi-asserted-by":"publisher","DOI":"10.1002\/j.1538-7305.1957.tb01515.x"},{"key":"ref26","doi-asserted-by":"publisher","DOI":"10.1145\/3085572"},{"key":"ref27","doi-asserted-by":"publisher","DOI":"10.1109\/isscc.2016.7418007"},{"key":"ref28","doi-asserted-by":"publisher","DOI":"10.1109\/isca45697.2020.00038"},{"key":"ref29","doi-asserted-by":"publisher","DOI":"10.1109\/lca.2020.2973991"}],"event":{"name":"2026 31st Asia and South Pacific Design Automation Conference (ASP-DAC)","location":"Lantau, Hong Kong","start":{"date-parts":[[2026,1,19]]},"end":{"date-parts":[[2026,1,22]]}},"container-title":["2026 31st Asia and South Pacific Design Automation Conference (ASP-DAC)"],"original-title":[],"link":[{"URL":"http:\/\/xplorestaging.ieee.org\/ielx8\/11420221\/11420229\/11420289.pdf?arnumber=11420289","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,3,11]],"date-time":"2026-03-11T05:37:54Z","timestamp":1773207474000},"score":1,"resource":{"primary":{"URL":"https:\/\/ieeexplore.ieee.org\/document\/11420289\/"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2026,1,19]]},"references-count":29,"URL":"https:\/\/doi.org\/10.1109\/asp-dac66049.2026.11420289","relation":{},"subject":[],"published":{"date-parts":[[2026,1,19]]}}}