{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2025,12,18]],"date-time":"2025-12-18T12:41:06Z","timestamp":1766061666975,"version":"3.48.0"},"reference-count":32,"publisher":"IEEE","license":[{"start":{"date-parts":[[2025,10,19]],"date-time":"2025-10-19T00:00:00Z","timestamp":1760832000000},"content-version":"stm-asf","delay-in-days":0,"URL":"https:\/\/doi.org\/10.15223\/policy-029"},{"start":{"date-parts":[[2025,10,19]],"date-time":"2025-10-19T00:00:00Z","timestamp":1760832000000},"content-version":"stm-asf","delay-in-days":0,"URL":"https:\/\/doi.org\/10.15223\/policy-037"}],"funder":[{"DOI":"10.13039\/501100000266","name":"Engineering and Physical Sciences Research Council","doi-asserted-by":"publisher","id":[{"id":"10.13039\/501100000266","id-type":"DOI","asserted-by":"publisher"}]}],"content-domain":{"domain":[],"crossmark-restriction":false},"short-container-title":[],"published-print":{"date-parts":[[2025,10,19]]},"DOI":"10.1109\/iros60139.2025.11247019","type":"proceedings-article","created":{"date-parts":[[2025,11,27]],"date-time":"2025-11-27T18:54:45Z","timestamp":1764269685000},"page":"13141-13146","source":"Crossref","is-referenced-by-count":0,"title":["Computationally Efficient FPGA-based Large Language Model Inference for Real-Time Decision-Making in Robotic Systems"],"prefix":"10.1109","author":[{"given":"Huaizhi","family":"Zhang","sequence":"first","affiliation":[{"name":"University of Essex,School of Computer Science and Electronic Engineering,Colchester,United Kingdom,CO4 3SQ"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Tamim M.","family":"Al-Hasan","sequence":"additional","affiliation":[{"name":"University of Essex,School of Computer Science and Electronic Engineering,Colchester,United Kingdom,CO4 3SQ"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Xuqi","family":"Zhu","sequence":"additional","affiliation":[{"name":"University of Essex,School of Computer Science and Electronic Engineering,Colchester,United Kingdom,CO4 3SQ"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Jiacheng","family":"Zhu","sequence":"additional","affiliation":[{"name":"University of Essex,School of Computer Science and Electronic Engineering,Colchester,United Kingdom,CO4 3SQ"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Weiyong","family":"Si","sequence":"additional","affiliation":[{"name":"University of Essex,School of Computer Science and Electronic Engineering,Colchester,United Kingdom,CO4 3SQ"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Klaus D.","family":"McDonald-Maier","sequence":"additional","affiliation":[{"name":"University of Essex,School of Computer Science and Electronic Engineering,Colchester,United Kingdom,CO4 3SQ"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Xiaojun","family":"Zhai","sequence":"additional","affiliation":[{"name":"University of Essex,School of Computer Science and Electronic Engineering,Colchester,United Kingdom,CO4 3SQ"}],"role":[{"role":"author","vocabulary":"crossref"}]}],"member":"263","reference":[{"year":"2024","article-title":"GPT-4 technical report","key":"ref1"},{"year":"2019","author":"Devlin","article-title":"BERT: Pre-training of deep bidirectional transformers for language understanding","key":"ref2"},{"year":"2023","author":"Touvron","article-title":"LLaMA: Open and efficient foundation language models","key":"ref3"},{"key":"ref4","first-page":"4110","article-title":"FederatedScope-GNN: Towards a unified, comprehensive and efficient package for federated graph learning","volume-title":"Proceedings of the 28th ACM SIGKDD Conference on Knowledge Discovery and Data Mining","author":"Wang"},{"doi-asserted-by":"publisher","key":"ref5","DOI":"10.3390\/bdcc8040036"},{"key":"ref6","doi-asserted-by":"crossref","first-page":"8643","DOI":"10.18653\/v1\/2023.findings-emnlp.578","article-title":"Retrieval-based knowledge transfer: An effective approach for extreme large language model compression","volume-title":"Findings of the Association for Computational Linguistics: EMNLP 2023","author":"Liu","year":"2023"},{"doi-asserted-by":"publisher","key":"ref7","DOI":"10.55041\/ijsrem32878"},{"key":"ref8","doi-asserted-by":"crossref","first-page":"3316","DOI":"10.18653\/v1\/2021.findings-emnlp.282","article-title":"The low-resource double bind: An empirical study of pruning for low-resource machine translation","volume-title":"Findings of the Association for Computational Linguistics: EMNLP 2021","author":"Ahia","year":"2021"},{"key":"ref9","doi-asserted-by":"crossref","DOI":"10.18653\/v1\/2023.findings-emnlp.489","article-title":"Dynamic stashing quantization for efficient transformer training","author":"Yang","year":"2023"},{"doi-asserted-by":"publisher","key":"ref10","DOI":"10.1109\/lra.2024.3415931"},{"doi-asserted-by":"publisher","key":"ref11","DOI":"10.1002\/iis2.13262"},{"doi-asserted-by":"publisher","key":"ref12","DOI":"10.1109\/icra57147.2024.10610183"},{"doi-asserted-by":"publisher","key":"ref13","DOI":"10.1109\/HPCA47549.2020.00035"},{"doi-asserted-by":"publisher","key":"ref14","DOI":"10.1109\/wacv56688.2023.00386"},{"doi-asserted-by":"publisher","key":"ref15","DOI":"10.47363\/jeast\/2023(5)183"},{"doi-asserted-by":"publisher","key":"ref16","DOI":"10.1049\/cdt2.12038"},{"doi-asserted-by":"publisher","key":"ref17","DOI":"10.1109\/JPROC.2021.3098483"},{"year":"2021","author":"Lin","article-title":"Efficient micro-structured weight unification and pruning for neural network compression","key":"ref18"},{"year":"2023","author":"Wake","article-title":"GPT models meet robotic applications: Co-speech gesturing chat system","key":"ref19"},{"doi-asserted-by":"publisher","key":"ref20","DOI":"10.1109\/IROS58592.2024.10802322"},{"doi-asserted-by":"publisher","key":"ref21","DOI":"10.1109\/IROS58592.2024.10801932"},{"year":"2024","author":"Latif","article-title":"3P-LLM: Probabilistic path planning using large language model for autonomous robot navigation","key":"ref22"},{"doi-asserted-by":"publisher","key":"ref23","DOI":"10.1109\/icmra62519.2024.10809099"},{"doi-asserted-by":"publisher","key":"ref24","DOI":"10.1109\/ro-man60168.2024.10731312"},{"doi-asserted-by":"publisher","key":"ref25","DOI":"10.1016\/j.jai.2024.12.003"},{"year":"2023","author":"Dettmers","article-title":"SpQR: A sparse-quantized representation for near-lossless LLM weight compression","key":"ref26"},{"year":"2022","author":"Frantar","article-title":"GPTQ: Accurate post-training quantization for generative pre-trained transformers","key":"ref27"},{"key":"ref28","first-page":"12284","article-title":"Extreme compression of large language models via additive quantization","volume-title":"International Conference on Machine Learning","author":"Egiazarian"},{"doi-asserted-by":"publisher","key":"ref29","DOI":"10.1145\/3626202.3637562"},{"volume-title":"Workshop on Efficient Systems for Foundation Models II @ ICML2024","author":"Key","article-title":"HLSTransform: Energy-efficient LLaMA 2 inference on FPGAs via high level synthesis","key":"ref30"},{"doi-asserted-by":"publisher","key":"ref31","DOI":"10.1109\/WF-IoT62078.2024.10811385"},{"year":"2024","author":"Hasan","article-title":"Optimizing large language models through quantization: A comparative analysis of PTQ and QAT techniques","key":"ref32"}],"event":{"name":"2025 IEEE\/RSJ International Conference on Intelligent Robots and Systems (IROS)","start":{"date-parts":[[2025,10,19]]},"location":"Hangzhou, China","end":{"date-parts":[[2025,10,25]]}},"container-title":["2025 IEEE\/RSJ International Conference on Intelligent Robots and Systems (IROS)"],"original-title":[],"link":[{"URL":"http:\/\/xplorestaging.ieee.org\/ielx8\/11245651\/11245652\/11247019.pdf?arnumber=11247019","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2025,12,18]],"date-time":"2025-12-18T12:37:41Z","timestamp":1766061461000},"score":1,"resource":{"primary":{"URL":"https:\/\/ieeexplore.ieee.org\/document\/11247019\/"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2025,10,19]]},"references-count":32,"URL":"https:\/\/doi.org\/10.1109\/iros60139.2025.11247019","relation":{},"subject":[],"published":{"date-parts":[[2025,10,19]]}}}