Research

Paper

AI LLM March 16, 2026

LEXI: Lossless Exponent Coding for Efficient Inter-Chiplet Communication in Hybrid LLMs

Authors

Miao Sun, Alish Kanani, Kaushik Shroff, Umit Ogras

Abstract

Data movement overheads increase the inference latency of state-of-the-art large language models (LLMs). These models commonly use the bfloat16 (BF16) format for stable training. Floating-point standards allocate eight bits to the exponent, but our profiling reveals that exponent streams exhibit fewer than 3 bits Shannon entropy, indicating high inherent compressibility. To exploit this potential, we propose LEXI, a novel lossless exponent compression scheme based on Huffman coding. LEXI compresses activations and caches on the fly while storing compressed weights for just-in-time decompression near compute, without sacrificing system throughput and model accuracy. The codecs at the ingress and egress ports of network-on-chip routers sustain the maximum link bandwidth via multi-lane LUT decoders, incurring only 0.09 percent area and energy overheads with GF 22 nm technology. LEXI reduces inter-chiplet communication and end-to-end inference latencies by 33-45 percent and 30-35 percent on modern Jamba, Zamba, and Qwen LLMs implemented on a homogeneous chiplet architecture.

Metadata

arXiv ID: 2603.15589
Provider: ARXIV
Primary Category: cs.AR
Published: 2026-03-16
Fetched: 2026-03-17 06:02

Related papers

Raw Data (Debug)
{
  "raw_xml": "<entry>\n    <id>http://arxiv.org/abs/2603.15589v1</id>\n    <title>LEXI: Lossless Exponent Coding for Efficient Inter-Chiplet Communication in Hybrid LLMs</title>\n    <updated>2026-03-16T17:48:30Z</updated>\n    <link href='https://arxiv.org/abs/2603.15589v1' rel='alternate' type='text/html'/>\n    <link href='https://arxiv.org/pdf/2603.15589v1' rel='related' title='pdf' type='application/pdf'/>\n    <summary>Data movement overheads increase the inference latency of state-of-the-art large language models (LLMs). These models commonly use the bfloat16 (BF16) format for stable training. Floating-point standards allocate eight bits to the exponent, but our profiling reveals that exponent streams exhibit fewer than 3 bits Shannon entropy, indicating high inherent compressibility. To exploit this potential, we propose LEXI, a novel lossless exponent compression scheme based on Huffman coding. LEXI compresses activations and caches on the fly while storing compressed weights for just-in-time decompression near compute, without sacrificing system throughput and model accuracy. The codecs at the ingress and egress ports of network-on-chip routers sustain the maximum link bandwidth via multi-lane LUT decoders, incurring only 0.09 percent area and energy overheads with GF 22 nm technology. LEXI reduces inter-chiplet communication and end-to-end inference latencies by 33-45 percent and 30-35 percent on modern Jamba, Zamba, and Qwen LLMs implemented on a homogeneous chiplet architecture.</summary>\n    <category scheme='http://arxiv.org/schemas/atom' term='cs.AR'/>\n    <published>2026-03-16T17:48:30Z</published>\n    <arxiv:comment>7 pages</arxiv:comment>\n    <arxiv:primary_category term='cs.AR'/>\n    <author>\n      <name>Miao Sun</name>\n    </author>\n    <author>\n      <name>Alish Kanani</name>\n    </author>\n    <author>\n      <name>Kaushik Shroff</name>\n    </author>\n    <author>\n      <name>Umit Ogras</name>\n    </author>\n  </entry>"
}