TX
Tianhua Xia
cs.ARcs.LGcs.AIcs.CLcs.CV
On Valency
published · living versionsW_f354b536·v1 · currentpublished
CodeQuant2: Enhanced Outlier Mitigation and Adaptive Clustering for Low-Precision LRMs
with Xingyu Liu, Xiangyang Yin, Haiyu Wang, Sai Qian Zhang
1 version
Preprints & journals
11 papers in the corpus · 2023–2026Acceptance-Aware Draft Model Training for Speculative Decoding2609.24150v1 · Tianhua Xia, Mugilan Ganesan, Yifei Feng et al.2026 · 0 citationsarXiv
EPIC: A System Framework for Efficient Egocentric Perception on Embodied AR Glasses2606.15859v1 · Tianhua Xia, Haiyu Wang, Jiajing Zheng et al.2026 · 0 citationsarXiv
DREAM-S: Speculative Decoding with Searchable Drafting and Target-Aware Refinement for Multimodal Generation2606.00535v1 · Zining Liu, Yunhai Hu, Tianhua Xia et al.2026 · 0 citationsarXiv
DREAM-R: Multimodal Speculative Reasoning with RL-Based Refined Drafting, Precise Verification, and Fully Parallel Execution2605.28678v1 · Yunhai Hu, Zining Liu, Xiangyang Yin et al.2026 · 0 citationsarXiv
CodeQuant: Unified Clustering and Quantization for Enhanced Outlier Smoothing in Low-Precision Mixture-of-Experts2604.10496v2 · Xiangyang Yin, Xingyu Liu, Tianhua Xia et al.2026 · 0 citationsarXiv
Kelle: Co-design KV Caching and eDRAM for Efficient LLM Serving in Edge Computing2510.16040v1 · Tianhua Xia, Sai Qian Zhang2025 · 4 citationsarXiv
DREAM: Drafting with Refined Target Features and Entropy-Adaptive Cross-Attention Fusion for Multimodal Speculative Decoding2505.19201v3 · Yunhai Hu, Tianhua Xia, Zining Liu et al.2025 · 1 citationarXiv
Foveated Instance Segmentation2503.21854v1 · Hongyi Zeng, Wenxuan Liu, Tianhua Xia et al.2025 · 5 citationsarXiv
HAAN: A Holistic Approach for Accelerating Normalization Operations in Large Language Models2502.11832v1 · Tianfan Peng, Jiajun Qin, Tianhua Xia et al.2025 · 4 citationsarXiv
Huff-LLM: End-to-End Lossless Compression for Efficient LLM Inference2502.00922v1 · Patrick Yubeaton, Tareq Mahmoud, Shehab Naga et al.2025 · 0 citationsarXiv
Hyft: A Reconfigurable Softmax Accelerator with Hybrid Numeric Format for both Training and Inference2311.13290v2 · Tianhua Xia, Sai Qian Zhang2023 · 11 citationsarXiv
Career total: 17 works. 11 are in this corpus.
Profile built from the corpus for this byline.
Author records are still filling in while the Hub is in alpha. If this is your page, you'll be able to claim it soon. Spot a mistake? Tell us.