TokTier: Exact Stateful CPU+GPU Tokenization for Agentic LLM Serving

arXiv:2607.29678 · cs.CL, cs.DC, cs.PF · Submitted 2026-07-31 · Read on arXiv

Zhenyu Zhang, Zhichao Cao

cs.CL, cs.DC, cs.PF

Submitted: 2026-07-31

Comments: 26 pages. Code: https://github.com/asu-idi/toktier

Code: https://github.com/asu-idi/toktier

License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/

Terminology

• Tokenization — 135× in this paper · explained in 2 episodes
• Core — 54× in this paper · explained in 2 episodes
• Trace — 40× in this paper · explained in 2 episodes
• Verifier — 15× in this paper · explained in 2 episodes
• KV cache — 14× in this paper · explained in 3 episodes
• Verification — 6× in this paper · explained in 2 episodes
• Accuracy — 4× in this paper · explained in 1 episode
• KV cache reuse — 3× in this paper · explained in 2 episodes
• Transformer — 3× in this paper · explained in 2 episodes
• Abstraction — 2× in this paper · explained in 1 episode
• Benchmark — 2× in this paper · explained in 2 episodes
• Calibration — 2× in this paper · explained in 7 episodes
• Symposium on Operating Systems Design and Implementation — 2× in this paper
• AI Agents — 1× in this paper · explained in 2 episodes
• Agentic Loop — 1× in this paper · explained in 1 episode
• Agentic serving — 1× in this paper · explained in 1 episode
• Alternation — 1× in this paper · explained in 1 episode
• Decode — 1× in this paper · explained in 2 episodes
• Fidelity — 1× in this paper · explained in 2 episodes
• Formal Verification — 1× in this paper · explained in 2 episodes
• Frozen model — 1× in this paper · explained in 2 episodes
• Generative Model — 1× in this paper · explained in 2 episodes
• Generative Models — 1× in this paper · explained in 2 episodes
• Geometric Distribution — 1× in this paper · explained in 2 episodes
• Provenance — 1× in this paper · explained in 3 episodes
• Pruning — 1× in this paper · explained in 4 episodes
• State Management — 1× in this paper · explained in 3 episodes
• Topological Order — 1× in this paper · explained in 2 episodes
• USENIX Conference on File and Storage Technologies — 1× in this paper

Sources

• Formalizing BPE Tokenization
• Characterizing CPU-Induced Slowdowns in Multi-GPU LLM Inference
• DeepSeek-V3 Technical Report
• SWE-Bench Pro: Can AI Agents Solve Long-Horizon Software Engineering Tasks?
• Incremental BPE Tokenization
• GPUTOK: GPU Accelerated Byte Level BPE Tokenization
• CacheWise: Understanding Workloads and Optimizing KVCache Management for Efficiently Serving LLM Coding Agents
• SWE-agent: Agent-Computer Interfaces Enable Automated Software Engineering
• ReAct: Synergizing Reasoning and Acting in Language Models
• BlockBPE: Parallel BPE Tokenization
• TraceLab: Characterizing Coding Agent Workloads for LLM Serving

Related papers

• Exploring Solution Divergence and Its Effect on Large Language Model Problem Solving
• Ishigaki-IDS-Bench: A Benchmark for Generating Information Delivery Specification from BIM Information Requirements
• Subliminal Steering: Stronger Encoding of Hidden Signals
• MedStruct-S: A Benchmark for Key Discovery, Key-Conditioned QA and Semi-Structured Extraction from OCR Clinical Reports
• The End of Transformers? On Challenging Attention and the Rise of Sub-Quadratic Architectures
• Untangling the Mechanisms of Misleading Context in Medical Question Answering